Hyperscaler capex Q3 2026: what Microsoft, Alphabet, Amazon and Meta guided

Microsoft’s late-October print is fiscal Q1 2027 (quarter ended September 30, 2026), released October 28, 2026. The binding fact pattern is not a spending pause. Capex including finance leases was $45.2 billion, below the company’s own “over $50 billion” outlook, because third-party capacity slipped—not because demand cooled. Azure still grew 48% to $33.2 billion, and management raised the next-quarter Azure guide to 49–50% in constant currency while saying demand still exceeds supply.

Capex: the miss is timing, and the mix is chips

Amy Hood’s definition of capital expenditures includes the full principal of finance leases at lease commencement. On that basis, Q1 spend was $45.2 billion, versus a prior outlook of more than $50 billion, because some third-party capacity expected in the quarter was delayed. Roughly three-quarters of that $45.2 billion went to short-lived assets—primarily CPUs and GPUs, including higher memory and storage costs. The rest was long-lived assets (buildings, power, shells).

Cash and the management capex number are not the same thing. GAAP cash paid for property and equipment was $39.367 billion, more than double the $19.394 billion a year earlier. Hood separately said finance leases in the capex figure were $7.4 billion. Those two lines do not foot to $45.2 billion ($39.4 + $7.4 = $46.8). That gap is timing: Microsoft books the full finance-lease principal in capex at commencement, while cash PP&E reflects when cash actually leaves. The 10-Q shows a related but different lease metric—right-of-use assets obtained in exchange for finance-lease obligations of $5.192 billion, versus $9.147 billion a year earlier—and operating-lease additions of $6.523 billion, versus $1.417 billion. The useful-life change (data-center buildings from 15 years to 25 years, effective July 2026) is pushing more new site leases into operating leases, which are outside the capex number. Reading $45.2 billion as “all-in infrastructure” understates the build.

Free cash flow stayed positive at $18.6 billion (operating cash flow of $58.0 billion minus cash PP&E). That is down about 27% from roughly $25.7 billion a year earlier, but it beat a Visible Alpha estimate of $8.54 billion by about $10 billion, which is why the stock traded as a cash-flow relief print rather than an AI-spend scare.

Azure and cloud: growth accelerated through a supply miss

This is the first quarter Microsoft reports Azure in dollars under the September 2026 segment redesign (Agents and Infra / Devices and Consumer). Azure revenue was $33.165 billion, up from $22.384 billion, or 48% both as reported and in constant currency—ahead of Microsoft’s own prior forecast of about 44–45% and ahead of the Street. Microsoft Cloud was $63.9 billion, up 30% (31% in constant currency), an acceleration from 27% in fiscal Q4 2026. Commercial remaining performance obligation was $677 billion, up 73%, with a weighted-average duration of 2.3 years; about 30% is due in the next 12 months (up 29%), and the portion beyond 12 months rose 98%.

The mechanism Hood described is efficiency, not a sudden flood of new halls. Foundry revenue more than tripled. GPU-fleet efficiency freed capacity that was allocated to customers, and token revenue per unit was higher than expected. On CPUs, engineering work monetized underutilized capacity inside the existing fleet. Those gains, plus model-mix ARPU and the timing of third-party GPU deliveries, are what she said will create quarter-to-quarter volatility in the growth rate. OpenAI commercial arrangements, including revenue-sharing, contributed $10.0 billion of revenue in the quarter—material, but not the whole Azure story. Company revenue was $92.279 billion, up 19%; non-GAAP EPS was $4.89, up 18%.

Q2 guidance (quarter ending December 31, 2026): revenue $96.0–$97.1 billion (18–19% growth) and Azure growth of 49–50% in constant currency, with efficiency gains expected to offset some GPU-as-a-service contracts pushed into the second half. Operating margin is guided slightly lower year over year, including an unquantified voluntary-exit cost.

FY2027 capex: no full-year dollar total—calendar 2026 cut to ~$170 billion, then spend reaccelerates

Microsoft did not guide a fiscal 2027 capex total. What Hood gave:

• Fiscal Q2 capex (including finance leases): over $50 billion.
• Calendar 2026 capex: approximately $170 billion, because of the third-party delays. That is a trim from the ~$175 billion path still in circulation after the July useful-life restatement. The July 29, 2026 call had put calendar 2026 at roughly $190 billion before that accounting shift moved more future data-center leases from finance leases (in capex) to operating leases (outside capex). Reuters, writing against the older $175 billion figure, put year-to-date calendar 2026 capex at $118.1 billion and said hitting $175 billion would require $56.9 billion in the December quarter. The call superseded that path with ~$170 billion. Some coverage (including an early CNBC line) mislabeled the $170 billion as calendar 2027; the transcript says “for the calendar year” immediately after the Q2 guide, i.e. calendar 2026.
• Second half of fiscal 2027: capex is expected to accelerate, mainly from CPUs (including higher memory and storage costs), GPUs, and the third-party deliveries that slipped from the first half.
• Fiscal 2027 free cash flow: expected to remain positive. Hood also said the company will tap additional funding when it is attractive. The next day Microsoft priced $25 billion of senior notes for general corporate purposes, which may include capex.

The implication for anyone modeling the build: the quarterly “miss” is a delivery deferral inside a still-rising plan, and the reported capex series is no longer comparable to pre-July 2026 prints because operating leases have been pulled out of the number.

What Nadella and Hood actually said on demand, constraints, and returns

Nadella’s constraint is physical and operational, not a demand air-pocket. Accelerating demand led Microsoft to add 24 data centers across five continents in the quarter. Dock-to-live times for new GB300s in the largest regions—most of the GPU deployments—fell about 20% quarter over quarter. Throughput on Copilot workloads rose more than 50% quarter over quarter by optimizing across data-center design, silicon, systems, software, and models. He framed the fleet as one fungible pool allocated to the highest-value use across compute, data, Foundry, and Copilot, and said returns compound across those layers rather than sitting in Azure utilization alone. Closing line: “based on the demand we see, the strength of our products, and the efficiencies we are driving across the stack, I have never been more confident in the long-term returns on our investments.” On the call he also drew a distinction between a handful of very large capacity buyers and the enterprise breadth motion he wants—agents built on Foundry that look, to him, like the database applications of prior cycles, and that take a decade rather than a few quarters.

Hood was more explicit that the constraint is still binding. “We still have demand that exceeds supply.” Where engineering frees capacity, she said they try to put it into Foundry. Q2’s 49–50% Azure guide assumes those efficiency gains continue, because some supply counted on for Q2 moved into the second half. On returns she did not publish an ROI: “as we continue to improve the efficiency of our infrastructure, each dollar of capex has the potential to support more revenue and higher returns over time,” and, in Q&A, “the return on every dollar of CapEx we’re spending, we’re confident in that.” The financing stance follows from that confidence plus positive free cash flow: leasing is already in use; the debt markets are open and, as of the call, she said they had not yet tapped them in this cycle; equity and other options exist if terms are attractive. She did not treat the Q1 underspend as a cut in the investment plan.

Figures

Item	Q1 FY2027 (ended Sep 30, 2026)	Comparison / note	Where it lives
Capex including finance leases	$45.2 billion	Below prior outlook of >$50 billion; ~$34.9 billion year-ago quarter	Earnings call (not a GAAP line)
Of which short-lived (CPUs/GPUs, incl. memory and storage)	Roughly three-quarters	Remainder long-lived assets. Press accounts put the prior three quarters near two-thirds; prepared remarks only state “roughly three quarters” this quarter	Earnings call
Finance leases in the capex figure	$7.4 billion	Full principal at commencement. Does not equal capex minus cash PP&E	Earnings call
Finance-lease ROU assets obtained	$5.192 billion	$9.147 billion in Q1 FY2026. Different metric from the $7.4 billion	10-Q
Operating-lease ROU assets obtained	$6.523 billion	$1.417 billion a year earlier; outside capex	10-Q
Cash additions to PP&E	$39.367 billion	$19.394 billion a year earlier	10-Q cash-flow statement; Hood rounded to $39.4 billion
Cash from operations	$58.000 billion	$45.057 billion a year earlier (+29%)	10-Q
Free cash flow	$18.6 billion	OCF minus cash PP&E; down ~27% YoY; beat Visible Alpha $8.54 billion	Earnings call; Reuters
Azure revenue	$33.165 billion	+48% reported and constant currency; $22.384 billion year ago	10-Q product table; earnings release growth rate
Microsoft Cloud revenue	$63.9 billion	+30% (+31% constant currency)	Earnings release
Commercial RPO	$677 billion	+73%; duration 2.3 years	Earnings release / call
Total revenue	$92.279 billion	+19%	Earnings release / 10-Q
Non-GAAP diluted EPS	$4.89	+18%; GAAP $4.92 includes a $0.03 OpenAI gain	Earnings release
Q2 FY27 capex outlook	Over $50 billion	Includes finance leases	Earnings call
Calendar 2026 capex outlook	Approximately $170 billion	Trimmed from ~$175 billion on 3P delays; July pre-accounting figure was ~$190 billion	Earnings call; prior-quarter call for the $190 billion
H2 FY27 capex	Expected to accelerate	CPUs (higher memory/storage), GPUs, delayed 3P capacity	Earnings call
FY27 free cash flow	Expected to remain positive	No full-year FY27 capex dollar guide	Earnings call
Q2 Azure outlook	49–50% constant currency	Demand still exceeds supply	Earnings call
Q2 revenue outlook	$96.0–$97.1 billion	18–19% growth	Earnings call

Primary documents

• Earnings release (8-K Exhibit 99.1, October 28, 2026): https://www.sec.gov/Archives/edgar/data/789019/000119312526436122/msft-ex99_1.htm and investor page https://www.microsoft.com/en-us/Investor/earnings/FY-2027-Q1/press-release-webcast
• Earnings call transcript (October 28, 2026): https://www.microsoft.com/en-us/investor/events/fy-2027/earnings-fy-2027-q1
• Form 10-Q for the quarter ended September 30, 2026 (filed October 28, 2026, accession 0001193125-26-436133): https://www.sec.gov/Archives/edgar/data/789019/000119312526436133/msft-20260930.htm

What this means if you are underwriting the spend. The investable signal is the mix and the constraint, not the $4.8 billion quarterly “miss.” About three-quarters of capex is short-lived silicon that can be turned into revenue inside a year if power and third-party halls show up; the miss was those halls, and management pushed both the silicon and the leases into the second half of fiscal 2027. Reported capex understates the footprint because operating leases jumped. Returns are being argued from utilization and token economics inside a still-short fleet—not from a disclosed ROI—and the day-after $25 billion note deal shows they will fund the gap in the debt market while free cash flow stays positive.

---

Recent Findings Supplement (October 2026)

Microsoft reported Q1 FY2027 results (quarter ended September 30, 2026) on October 28, 2026, with the earnings release, 10-Q filing, and earnings call transcript all dated that day.[1][2][3]

Key new data points (post-April 30, 2026) include accelerated Azure growth to 48%, Microsoft Cloud at $63.9 billion (+30%), total capex of $45.2 billion (below the >$50 billion guide due to third-party delays), and ~75% of capex allocated to short-lived assets. Management guided Q2 Azure growth at 49-50% (cc), Q2 capex >$50 billion, and ~$170 billion for calendar 2026, while affirming demand exceeds supply and expecting positive free cash flow for FY2027.[4][5][6]

Capex Breakdown and Allocation

Microsoft’s infrastructure buildout continued at scale but came in below the prior >$50 billion quarterly guide due to delayed third-party capacity deliveries (now expected in H2 FY2027). Total capital expenditures (including finance leases) reached $45.2 billion (+~30% YoY). Finance leases contributed $7.4 billion, while cash paid for property and equipment was $39.4 billion (more than double the prior-year $19.4 billion). Roughly three-quarters (~75%) of capex went to short-lived assets, primarily CPUs and GPUs (including higher memory/storage costs), up from ~66% in recent prior quarters; the balance was for longer-lived infrastructure.[7][5][6]

This mix reflects prioritization of quickly deployable, revenue-generating hardware amid ongoing AI demand. Free cash flow was $18.6 billion (down 27% YoY) after $58.0 billion in operating cash flow, but management reiterated expectations of remaining FCF-positive for FY2027.[2][4]

Azure and Cloud Revenue Performance

Azure revenue reached $33.2 billion (or $33.17 billion in some reports), up 48% YoY (48% in constant currency), accelerating from 42% in the prior quarter and beating Microsoft’s own 44-45% guidance despite third-party capacity delays. Microsoft Cloud revenue hit $63.9 billion (+30% YoY, +31% cc), with growth accelerating from 27% in Q4 FY2026.[5][8][9]

Agents and Infra segment revenue was $76.8 billion (+25%). Commercial remaining performance obligation rose 73% to $677 billion. These figures mark the first routine quarterly disclosure of exact Azure dollar revenue.[1][8]

Executive Commentary on Demand, Constraints, and Returns

CFO Amy Hood noted that results exceeded expectations across revenue, operating income, and EPS, driven by strong Microsoft Cloud demand. She highlighted capex at $45.2 billion (with the 75% short-lived asset split), finance leases at $7.4 billion, and cash PP&E at $39.4 billion; she attributed the capex shortfall to delayed third-party deliveries and guided acceleration in H2 from CPUs/GPUs (with higher memory/storage costs) plus shifted 3P capacity. Hood emphasized efficiency gains at the GPU/CPU level, enabling more capacity allocation (e.g., to Foundry), and stated Microsoft would “remain free cash flow positive” in FY2027 while accessing funding opportunistically.[4][5][6]

CEO Satya Nadella described a “fast start” to FY2027, with Microsoft Cloud surpassing $63 billion (+30% YoY) and “never been more confident in the opportunity ahead.” He pointed to broad demand across enterprises, digitally native companies, and AI infrastructure workloads, plus efficiency improvements allowing better utilization.[10][11]

Both executives stressed that demand continues to exceed supply, with capacity constraints persisting but being actively addressed through internal builds, third-party arrangements, and operational optimizations.[9][12]

FY2027 / Near-Term Guidance and Outlook

• Q2 FY2027 revenue: $96.0–97.1 billion (+18–19% YoY).
• Q2 Azure growth: 49–50% in constant currency (above consensus ~45%).
• Q2 capex (incl. finance leases): >$50 billion.
• Calendar 2026 capex: ~$170 billion (trimmed from prior ~$175B view due to delays; underlying plans unchanged).
• H2 FY2027 capex expected to accelerate from CPUs/GPUs/memory/storage and delayed 3P deliveries.
• Operating margin expected slightly lower YoY in Q2 (including unquantified voluntary exit costs); Microsoft Cloud gross margin ~65% in Q2.[12][4][5]

Overall revenue was $92.3 billion (+19%), operating income $45.9 billion (+21%), GAAP net income $36.6 billion (+32%), and non-GAAP EPS $4.89 (+18%).[1]

Summary Table of Key Figures (Q1 FY2027)

Metric	Figure	YoY Change / Notes	Source
Total Capex (incl. finance leases)	$45.2B	+~30%; below >$50B guide due to delays	Earnings call / release (Oct 28, 2026)[4][5]
Finance Leases	$7.4B	—	Earnings call (Hood)[6]
Cash Paid for PP&E	$39.4B	>2x prior year	Earnings call / 10-Q[2][7]
Short-Lived Assets Share (CPUs/GPUs + mem/storage)	~75% of capex	Up from ~66% recently	Earnings call (Hood)[7][5]
Azure Revenue	$33.2B / $33.17B	+48% (48% cc)	Earnings release / call[5][8]
Microsoft Cloud Revenue	$63.9B	+30% (+31% cc); accelerating	Earnings release[1]
Q2 Capex Guidance	>$50B	—	Earnings call / guidance[4]
CY2026 Capex Guidance	~$170B	Trimmed due to delays	Earnings call / guidance[5]

Primary sources: Microsoft earnings release (Oct 28, 2026), 10-Q (filed Oct 28, 2026), and Q1 FY2027 earnings call transcript (Oct 28, 2026).[3][13] Data reflects only post-April 30, 2026 publications and announcements.

Alphabet held 2026 capex at $195–205 billion after a $51.4 billion third quarter, told investors spending will rise “significantly” in 2027 without a number, and used Cloud’s doubling and a $561 billion backlog to argue the AI build is already earning its keep—even as free cash flow went negative and third-party compute starts to press Cloud margins.

Results were released October 28, 2026, for the quarter ended September 30. Alphabet does not publish a separate quarterly shareholder letter. The primary documents are the earnings release (8-K Exhibit 99.1), the CEO’s edited prepared remarks on the Google blog, the earnings-call transcript, and the Form 10-Q filed October 29, 2026.

Capex, the mix, and what 2026–27 actually says

The constraint is no longer “will they spend,” it is the mix and the cash timing. Purchases of property and equipment were $51.355 billion in Q3 (management rounded this to $51.4 billion), more than double the year-ago quarter, and $131.953 billion for the first nine months. Operating cash flow was $42.290 billion, so free cash flow—operating cash flow less capex—was negative $9.065 billion. Management also flagged a TPU inventory build and a $5.2 billion cash payment of a finalized European Commission fine as additional drains on cash.

Anat Ashkenazi’s split of the quarter’s technical-infrastructure spend, which she said was the vast majority of capex:

• About 65% servers and networking equipment, generally depreciated over six years.
• The remaining 35% data centers, with useful lives of seven to 40 years.

That is a flow mix, not the stock. The 10-Q says that as of September 30, 2026, about 60% of technical-infrastructure assets on the balance sheet were servers and network equipment, with the rest data-center land, buildings, and related assets.

Full-year 2026 capex guidance was left unchanged at $195 billion to $205 billion. With $132.0 billion already spent through September, landing inside that range implies roughly $63–73 billion of Q4 capex (arithmetic from the filing and the guidance range, not a company forecast).

On 2027, the company gave direction, not a number. Ashkenazi: “We continue to expect our CapEx to increase significantly in 2027, and we’ll provide further details on our Q4 call.” Figures circulating in the press that 2027 could exceed $300 billion are investor speculation, not guidance. Depreciation and data-center operating costs, including energy, are expected to keep weighing on the P&L as the new capacity comes online.

Longer-dated supply is already locked in. The 10-Q reports $802.1 billion of contractual obligations with remaining terms greater than one year as of September 30, 2026—mostly long-term supply agreements for technical infrastructure and inventory components, with the significant majority of those payments expected through 2031, plus energy agreements that can run into the 2050s.

Google Cloud: growth, margin, backlog

Cloud is the line where AI shows up as both revenue and profit at the same time. Google Cloud revenue was $30.709 billion, up 103% from $15.157 billion. Operating income more than tripled to $11.121 billion, and operating margin expanded to 36.2% from 23.7% a year earlier. Management said growth was led by Google Cloud Platform—enterprise AI infrastructure, enterprise AI solutions, and core GCP—and that TPU system sales were a meaningful contributor. On the call, Thomas Kurian said GCP was the vast majority of Cloud revenue, and that Cloud growth excluding TPU system sales was still very strong.

Backlog, from the 10-Q (Note 2), not the rounded call figure:

• Total remaining performance obligations: $567.7 billion.
• Of which Google Cloud: $560.8 billion (Ashkenazi rounded this to $561 billion).
• Just over 55% expected to be recognized as revenue over the next 24 months, up from “just over 50%” at June 30, when Cloud backlog was $513.9 billion of a $519.5 billion total.
• Includes $16.8 billion tied to a multi-year commercial agreement with Crux AI, the Blackstone joint venture and a related party.

Recognition timing depends on contract duration, Alphabet’s ability to deliver, and when customers actually use the capacity—so the 55% is an estimate, not a schedule.

The margin expansion is not yet fully stress-tested against the bridging strategy. Third-party compute capacity only came online at the very end of Q3, so it had limited negative impact on the quarter’s 36.2% margin. Ashkenazi said usage ramps in Q4 and will create near-term Cloud margin headwinds.

What Pichai said on demand, constraints, TPUs, and returns

Pichai’s framing is that demand is not the scarce input—capacity and allocation discipline are. From the earnings release and the blog/call remarks:

• Alphabet is at a half-trillion-dollar annual revenue run rate, with revenue up 25% to $127.577 billion, the 13th straight quarter of double-digit growth.
• “Our AI investments are driving overall Search query growth, including commercial queries. In fact, our year-over-year query growth rate is the highest it’s been in the last five years.” Search and other revenue was $63.350 billion, up 12%.
• On Cloud: customers see Google as the leading provider of AI infrastructure and solutions, “so there’s huge demand. We are working hard to bring more capacity online to meet this long-term demand.”
• TPU systems: “strong demand… across a wide range of customers, and we expect that to continue as we scale.”
• Capacity is already binding on the model side. Argon, launched about two weeks before the call, ramped on developer platforms three-and-a-half times faster than Gemini 3 in its first week, “even with capacity constraints.” First-party model APIs were processing nearly 32 billion tokens per minute, up 40% quarter over quarter and more than four times a year ago. The Gemini app reached one billion users.

On allocation, in Q&A, he said compute allocation is something they spend a lot of time on because of the forward planning required. He rejected the idea of a short-term rationing exercise: each business runs long-range plans under a “disciplined ROIC framework.” Models are the foundation across Search, YouTube, Cloud, and Waymo, and the full stack is meant to create both efficiency and optionality. Demand signals are “very, very robust,” and they are “investing for that moment.”

On economics of selling systems outside Google’s own data centers, he said the ability to deploy TPUs both in Google data centers and as systems in customers’ data centers scales the product and “directly translates into better economics.”

What Ashkenazi said on constraints, TPU cash timing, and returns

Ashkenazi separated three clocks that the market was collapsing into one number: capex, Cloud margin, and cash from TPU system sales.

Supply and margins. The environment is still supply-constrained. Alphabet started using third-party compute at the end of Q3 and will ramp it in Q4 as a bridge while internal capacity is built. That capacity is expected to pressure Cloud margins in the near term. Higher depreciation and data-center costs such as energy will also hit the P&L.

2026 and 2027 spend. The $195–205 billion 2026 range is unchanged. 2027 capex is still expected to increase significantly; the number comes on the Q4 call.

TPU systems are not traditional capex. A system sale is hardware, compute, storage, networking, and software for training and inference—not a chip. When a contract is signed it hits backlog first; Alphabet then builds inventory before it bills. Cash goes out before revenue comes in. Some system sales have already been recognized; sales are expected to keep scaling in Q4, with a meaningful revenue ramp in 2027. Kurian added that inventory is procured against signed contracts, and that placing systems in customer or partner data centers (Oracle; Crux AI with Blackstone) both widens the market—including customers who cannot move data to the cloud—and reduces the need for Google to pre-build hosting for those systems.

Returns. Before Q&A she addressed ROIC and free cash flow directly. The pitch is that the same infrastructure is monetized across products that already reach billions of users and enterprises: Search at its highest query growth in five years; Google Cloud with “tens of billions of dollars in AI revenues year-to-date”; Google One as the fastest-growing subscription, driven by AI plans. She said they expect an attractive ROIC that stays meaningfully above the cost of capital, and strong operating cash flow, even while technical-infrastructure investment and the TPU inventory build pressure free cash flow.

What the filings add that the call compressed

The 10-Q (filed October 29, 2026) is the cleaner source on backlog, cash, and commitments than the rounded call language.

• Revenue backlog $567.7 billion, Cloud $560.8 billion, just over 55% inside 24 months, including $16.8 billion with Crux AI.
• Q3 purchases of property and equipment $51.355 billion; nine-month total $131.953 billion; Q3 operating cash flow $42.290 billion.
• Inventory on the balance sheet rose sharply (the cash-flow statement shows a $10.959 billion inventory use of cash over nine months), consistent with the TPU build-ahead-of-bill story.
• Accrued purchases of property and equipment were $21.398 billion at quarter-end, versus $8.877 billion at year-end 2025—capex cash and capex accruals are not the same thing.
• Long-term contractual obligations with terms over one year: $802.1 billion.
• In Q3 the company issued senior notes for net proceeds of $28.7 billion. Long-term debt was $125.1 billion at quarter-end. The earnings release also notes ATM equity issuance in the quarter ($4.3 billion of Class A and $8.1 billion of Class C) and $27.6 billion still available under the program. There were no share repurchases in the first nine months of 2026, versus $39.810 billion in the year-ago nine months.
• The 10-Q states that Alphabet has signed a limited number of agreements to supply TPU systems to customers who need or provide on-premises infrastructure, began recognizing revenue from those agreements in Q2 2026, and expects the significant majority of that revenue in 2027—consistent with, and slightly more precise than, the call’s “meaningful revenue ramp in 2027.”

Consolidated operating income was $40.576 billion, up 30%, at a 31.8% margin (the release table rounds this to 32%). Net income available to common stockholders was $26.170 billion, or $2.12 diluted EPS, pulled down mainly by an $8.2 billion other-income loss, largely an unrealized loss on the SpaceX stake. Google Services operating margin was 40.7% on $96.302 billion of revenue.

How to read this if you are underwriting the spend

The company’s own case is that returns are already visible in query growth, Cloud AI revenue, and subscription mix, and that ROIC is being managed on a multi-year basis rather than against one quarter of negative free cash flow. The open items are the ones they deferred: the 2027 capex number, how large the Q4 third-party compute drag on Cloud margin is, and how fast TPU system inventory converts to cash in 2027. A competitor or investor should treat the $195–205 billion 2026 range as firm, treat “significantly higher in 2027” as directional only until the Q4 call, and not confuse the 65/35 quarterly technical-infrastructure spend split with the 60/40 stock of assets already on the books.

Sources:

1. Alphabet Q3 2026 earnings release (October 28, 2026) — https://abc.xyz/files/doc_financials/2026/q3/2026q3-alphabet-earnings-release.pdf
2. Alphabet Q3 2026 earnings call transcript (October 28, 2026) — https://s206.q4cdn.com/479360582/files/doc_events/2026/Oct/28/2026_Q3_Earnings_Transcript.pdf
3. Alphabet investor relations, Q3 2026 earnings call page — https://abc.xyz/investor/events/event-details/
4. Sundar Pichai, “Q3 2026 earnings call: Remarks from our CEO” (October 28, 2026) — https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q3-2026/
5. Alphabet Form 10-Q for the quarter ended September 30, 2026 — https://www.sec.gov/Archives/edgar/data/1652044/000165204426000084/goog-20260930.htm
6. CNBC, Alphabet Q3 2026 earnings coverage (October 28, 2026) — https://www.cnbc.com/2026/10/28/alphabet-googl-earnings-q3-2026.html
7. Reuters, Alphabet Q3 2026 (October 28, 2026) — https://www.reuters.com/business/alphabets-google-cloud-unit-beats-quarterly-revenue-estimates-steady-ai-demand-2026-10-28/

---

Recent Findings Supplement (October 2026)

Alphabet reported Q3 2026 results on October 28, 2026 (quarter ended September 30), with total revenue of $127.58 billion (+25% YoY, reported and constant currency), marking its 13th consecutive quarter of double-digit growth and reaching a ~$500 billion annual run rate.[1][2][3]

Google Cloud drove outsized growth amid heavy AI infrastructure investment, while core Search showed moderating but still solid expansion. The results featured a significant acceleration in capex and negative free cash flow, with management maintaining 2026 capex guidance while signaling a sharp step-up in 2027 alongside TPU system sales ramping meaningfully.[4][5]

Capex reached $51.4 billion in Q3 (more than double YoY), primarily for technical infrastructure supporting AI; FY 2026 guidance held at $195–205 billion, with expectations of a significant increase in 2027 (one reference to exceeding $300 billion).[4][6][7]

Approximately 65% of technical infrastructure spend went to servers and networking equipment (useful lives ~6 years), with ~35% to data centers. Free cash flow turned more negative at –$9.07 billion (vs. expectations near breakeven), driven by capex, TPU inventory build, and other factors like a European fine payment; TTM FCF fell 73% to $19.7 billion.[5][8]

This reflects Alphabet’s aggressive buildout to address sustained AI demand, shifting from prior years’ lower spending levels. Competitors or entrants face pressure to match scale or find differentiated efficiency, as the capex intensity highlights the capital barriers in AI infrastructure.

Google Cloud revenue surged 103% to $30.71 billion (beating estimates), with operating income more than tripling to $11.1 billion and margin expanding to 36.2% (from 23.7%).[9][8][10]

Growth was led by Google Cloud Platform (GCP) strength in enterprise AI infrastructure/solutions and core services; TPU system sales were a “meaningful contributor.” Backlog reached $561 billion (Alphabet total ~$568 billion), with just over 55% expected to be recognized over the next 24 months.[6][9]

TPU inventory rose sharply (to $13.3 billion in one analysis) as the company builds ahead of deliveries, with ~100 customers including AI labs like Anthropic. This positions Cloud as the primary growth engine, with TPU systems extending the moat beyond internal use into external sales. New entrants or smaller players must navigate supply constraints and compete on full-stack AI offerings (silicon to software) to capture similar enterprise commitments.

No detailed updated full-year 2026 revenue guidance was provided beyond capex; Q4 faces larger FX headwinds and lapping of prior Search acceleration.[11][12]

Initial 2027 commentary emphasized capex rising significantly, TPU system sales continuing to scale in Q4 with a “meaningful revenue ramp” in 2027, and AI-driven productivity gains (Pichai described 2027 as a potential “golden year of productivity” from models like Argon and Carbon, citing internal gains such as 35% higher code deliveries).[5][13]

Near-term Cloud margin pressure is expected from third-party compute capacity (ramping in Q4 after starting end of Q3) as a bridge amid supply constraints.[9][12]

Sundar Pichai highlighted robust demand signals and AI momentum: AI investments drove the highest Search query growth rate in five years (including commercial queries); Cloud saw “extraordinary demand” for AI infrastructure/solutions; TPU systems showed strong customer demand across AI labs, enterprises, and others, with scaling expected to continue.[14][15]

He noted ongoing work to bring more capacity online and careful compute allocation between internal frontier needs, consumer/enterprise serving, and external opportunities (e.g., partnerships like Oracle and Crux AI JV).[8]

Anat Ashkenazi emphasized capex discipline within the unchanged 2026 range while flagging 2027 increases, TPU inventory build pressuring FCF, and near-term margin headwinds from third-party capacity; she underscored the $561 billion Cloud backlog and >55% recognition timeline.[16][12]

Thomas Kurian (Google Cloud CEO) reinforced that GCP drove the majority of Cloud growth (excluding TPU systems impact), with broad demand from digital natives, AI labs, enterprises, and financial services; TPU systems sales are scaling as a distinct offering.[9][17]

Overall, the earnings underscore accelerating AI-driven demand (especially in Cloud and via TPUs) offset by elevated capex intensity and near-term cash flow/margin pressures from supply bridging and inventory. This creates a high-bar environment where scale, custom silicon (TPUs), and full-stack capabilities differentiate leaders, while pressuring returns timelines into 2027+.

Amazon’s Q3 2026 print is a capacity story financed with debt: AWS grew 39% to a $184 billion run rate with a $578 billion backlog, while cash capex hit $75.9 billion in the quarter and the company still expects about $220 billion for 2026—and “meaningfully more” in 2027. The release attributes the free-cash-flow swing to AI infrastructure, but management did not break spend into Trainium, Nvidia GPUs, and power.

Quarterly cash capex, and what is not in that number

Amazon’s cash capex is purchases of property and equipment net of proceeds from sales and incentives—not gross cash outlays, and not finance-lease additions. On that definition, CFO Brian Olsavsky said Q3 cash capex was $75.9 billion and “primarily relates to AWS and AI.” The cash-flow statement shows the gross figure underneath that: purchases of property and equipment of $77.259 billion, against $1.381 billion of proceeds from sales and incentives, which nets to about $75.9 billion.

Finance leases are disclosed separately and are not in cash capex. Property and equipment acquired under finance leases, net of remeasurements and modifications, was $1.472 billion in the quarter (versus $977 million a year earlier). Principal repayments of finance leases were $433 million and sit in financing cash flow, not investing. Operating-lease assets acquired were $7.407 billion. Adding finance-lease commencements to cash capex would overstate cash spent; the economic commitment is real, but the cash leaves later as lease principal.

• Gross purchases of PPE: $77.259 billion in Q3 2026 versus $35.095 billion in Q3 2025.
• Nine-month gross purchases: $175.670 billion; proceeds from sales and incentives $3.482 billion, so net cash capex year-to-date is about $172.2 billion.
• Trailing-twelve-month net purchases of PPE: $210.7 billion, up $94.8 billion year over year. The release says that increase “primarily reflects investments in artificial intelligence.”
• Trailing-twelve-month free cash flow swung to an outflow of $42.4 billion from an inflow of $14.8 billion a year earlier. Operating cash flow still rose 29% to $168.3 billion. The gap was funded in part by $48.102 billion of long-term debt proceeds in the quarter. Long-term debt was $173.1 billion at September 30, 2026, versus $65.6 billion at December 31, 2025.

What this means for anyone modeling Amazon: use the $75.9 billion cash-capex figure when comparing to the $220 billion full-year guide. The $77.3 billion gross number that circulated in headlines is before incentives. Finance leases add roughly $1.5 billion of non-cash asset additions this quarter, not another $1.5 billion of cash capex.

Full-year 2026 guide and the 2027 direction

The company reiterated, in both the release and on the call, that it continues to expect 2026 cash capital expenditures of approximately $220 billion, “the majority for AWS.” That figure was raised earlier in the year from about $200 billion, with management previously tying the increase to higher memory costs. It was not raised again on October 29.

The arithmetic of the guide is the tension. About $172 billion of net cash capex is already in the first nine months, so holding $220 billion implies roughly $48 billion in Q4—a sharp slowdown from $75.9 billion in Q3. Management did not walk through that deceleration. Outside commentary (XTB) flagged the guide as conservative relative to the run rate and noted the company gave no indication of when the cash-burn trend reverses. Amazon did not raise the 2026 number on this call, so the official figure remains approximately $220 billion.

For 2027 the language is directional only. Jassy: “We still expect 2026 cash CapEx of approximately $220 billion, the majority for AWS, and expect to spend meaningfully more in 2027 with a clear line of sight to attractive financial returns, as I laid out on our last call.” No dollar range, no growth rate, and no split of incremental spend between chips, power, and buildings. Fourth-quarter operating guidance is $28.5–$33.5 billion of operating income on $240–$247 billion of sales; that does not constrain 2027 capex.

What this means for competing hyperscalers: Amazon is the only one of the three large clouds, on this week’s prints, that both held a very high 2026 number and explicitly pointed 2027 higher without a ceiling. Microsoft trimmed its fiscal-2026 capex view to about $170 billion; Alphabet held $195–$205 billion and also said 2027 would be significantly higher. Amazon’s incremental message is that the binding constraint is still supply, not a desire to protect free cash flow.

AWS growth, backlog, and where AI shows up in the P&L

AWS net sales rose 39% (39.3% in the release) to $46.0 billion, the fastest growth in nearly five years, to a $184 billion annualized run rate. Operating income was $17.8 billion, a 38.7% margin, versus $11.4 billion a year earlier. That is about 63% of company operating income of $28.1 billion. Core compute and storage grew at their fastest rates in several years, so the acceleration is not only an AI overlay on a flat base.

Backlog is the forward indicator management wants investors to use. Jassy said AWS backlog is $578 billion, growing triple digits year over year and “well diversified”—more than double a year earlier, per contemporaneous reporting of the call. He tied the multi-year growth claim to that book: “we believe that AWS can grow north of 40% in each of 2027 and in 2028, and be a trillion-dollar revenue business in the next 5 to 10 years.” The earnings release used “over 40%” for both years and the $184 billion run rate.

On the mix of AI revenue, the only quantitative split disclosed is inference versus training, not chip vendor versus custom silicon. Jassy: enterprises are “still very early in using inference at scale, and yet inference already accounts for well over half of AWS’s AI revenue and growing,” with “a lot of training revenue as well.” Bedrock token volume grew nearly 17 times year over year; Fortune 500 spend on Bedrock was nearly 7 times higher. Chips revenue is growing triple digits, and revenue commitments for Graviton and Trainium each increased more than 20 times year over year. Customers using custom silicon saved more than $2 billion in Q3 alone. Trainium will be offered outside AWS late in 2027.

What this means for the growth math: 40%+ growth on a $184 billion run rate for two years is a capacity-addition statement as much as a demand statement. The $578 billion backlog is contracted future revenue, not recognized revenue, and management said the largest blocks—hundreds of megawatts, sometimes up to a gigawatt—are reserved well in advance, which is why they claim visibility into 2027, 2028, and some of 2029.

AI share of spend: Trainium, Nvidia, power

Amazon did not disclose a percentage of capex going to Trainium, Nvidia GPUs, or power. The closest official characterizations are qualitative and should not be turned into a mix model.

• The year-over-year increase in purchases of property and equipment “primarily reflects investments in artificial intelligence” (earnings release).
• Q3 cash capex of $75.9 billion “primarily relates to AWS and AI” (Olsavsky).
• The majority of the ~$220 billion 2026 cash capex is for AWS (Jassy).
• The company is “installing an AI accelerator about every 14 seconds,” and capacity is being absorbed about as fast as it is installed (Jassy). That phrase covers accelerators generally; it is not a Trainium-versus-Nvidia count.
• AWS expanded its Nvidia GPU portfolio in the quarter (CNBC), and new agreements cited in the release include Meta, NVIDIA, and OpenAI, among others. No dollar allocation was given.
• Power is treated as a constraint and a cost, not a disclosed capex line. Guidance assumptions explicitly exclude energy-derivative remeasurements. Component-cost pressure called out on the call was memory, not a power-price figure.

The mechanism management is selling is vertical integration, not a GPU rental book. Jassy contrasted AWS with “primarily hosting GPUs or renting short-term data center capacity,” and said most capacity is owned or long-leased, which he argued carries better operating margins than colocation-heavy models. Trainium and Graviton are the cost lever inside that owned fleet: price-performance of 30–40% better on Graviton versus x86 alternatives, TorchNeuron so existing PyTorch runs on Trainium without a rewrite, and external Trainium sales only in late 2027—so near-term chip spend still mostly serves AWS’s own capacity, not a merchant chip business.

What this means for anyone trying to size the AI bill: “primarily AI” and “majority AWS” are not the same as “90% GPUs.” Fulfillment, retail logistics, and non-AI AWS still sit inside the $220 billion. Without a segment capex table, the honest range is: most of the dollars are AWS infrastructure, the incremental dollars versus last year are mostly AI, and the Trainium/Nvidia/power split was not disclosed on October 29.

Capacity, power, and returns: what Jassy and Olsavsky actually said

On the October 29 call, capacity comments were about demand absorption and reservation lead times, not a restatement of the “sold out” line. Jassy said overall demand for AWS capacity “remains exceptionally strong,” that an accelerator is going in about every 14 seconds, and that customers are taking capacity about as fast as it is installed. In Q&A he said the big blocks—“hundreds of megawatts or sometimes up to a gigawatt”—are reserved well in advance because “you can’t leave a gigawatt of capacity hanging around,” which is the basis for “very good visibility into ’27 and ’28 and some of ’29.” He also said Amazon keeps some capacity for on-demand EC2 and Bedrock rather than selling every megawatt forward; that reservation policy was stated more fully six days earlier, not as a new numeric disclosure on the earnings call.

The sharper power and sold-out language is from October 23, on the a16z podcast, reported by Reuters—not from the earnings release. Jassy said 2026 capacity is sold out, 2027 is “largely sold out,” and “a good chunk of 2028 is sold out,” while AWS still holds some capacity for on-demand, Bedrock, and EC2. On bottlenecks: “There’s only so much data center and power today. There’s only so much memory, and I don’t think that shortage ends in 2027.” He also said Amazon “could invest a lot more and it wouldn’t be a bad use of capital.” Treat that as management’s pre-print view of the constraint set (space, power, memory), confirmed in direction by the call’s “meaningfully more in 2027” and the gigawatt-reservation comments, but not as a figure audited in the 8-K.

On returns, both executives argued payback is shortening even as depreciation and memory costs rise. Jassy: operating margins on AI are “rapidly improving… well ahead of where they were on core at a similar time of evolution, even with heightened component costs like memory and significant depreciation from the CapEx we’re investing. We continue to accelerate our CapEx payback” and have “strong return on invested capital.” He called AI “a once-in-a-lifetime opportunity” and said the company will invest to remain the leader rather than ration a 40% growth path. Olsavsky: the investments, “especially in AI,” are meant to drive “long-term revenue and free cash flow, as well as return on invested capital.” He also said the 38.7% AWS margin is the product of efficiency, capacity optimization, fixed-cost control, and a rising share of custom chips—and that margins will still fluctuate with AI mix and component costs. Neither gave a payback period, an IRR, or an ROIC percentage on this call. (On the prior quarter’s call, Jassy had said servers and networking break even in a little less than three years on average; that specific timing was not repeated as a new Q3 metric.)

What this means for capital allocation: the returns argument is “payback is moving left, backlog is contracted, custom silicon improves the cost stack,” not a published ROIC bridge. The cash-flow statement is the counterweight—TTM free cash flow of negative $42.4 billion and $48 billion of new long-term debt in one quarter—so the return case only works if the $578 billion backlog converts and the 40% growth years actually show up in revenue before depreciation and interest dominate.

How to use the figures

Item	Figure	Source treatment
Q3 cash capex (net of incentives)	$75.9 billion	Olsavsky on the call; equals gross PPE purchases minus proceeds
Q3 gross purchases of PPE	$77.259 billion	Cash-flow statement in the Oct. 29 release
Q3 finance-lease asset additions	$1.472 billion	Supplemental cash-flow disclosure; not cash capex
2026 cash capex	Approximately $220 billion, majority AWS	Release and Jassy; unchanged
2027 capex	“Meaningfully more” than 2026; no number	Jassy
AWS revenue / growth	$46.0 billion, +39.3%	Release
AWS backlog	$578 billion, triple-digit growth	Jassy on the call
AWS 2027–28 growth	North of / over 40% each year	Jassy; also in the release
AI share of capex	Not quantified by Trainium / Nvidia / power	“Primarily” AI for the YoY PPE increase; majority of 2026 cash capex is AWS

Confidence is high on the cash-flow and guidance figures because they are in the October 29 release and cash-flow statement. Confidence is high on the call quotations from multiple transcripts (Investing.com, AlphaStreet, Benzinga), which agree on the $75.9 billion, $578 billion, 40%+, and “meaningfully more” lines. Confidence is lower on any implied Q4 capex of ~$48 billion—that is arithmetic from the guide minus year-to-date net spend, not a company forecast. There is no sourced split of spend across Trainium, Nvidia GPUs, and power.

---

Recent Findings Supplement (October 2026)

Amazon reported Q3 2026 results on October 29, 2026, highlighting accelerated AWS growth amid heavy AI-driven capex that turned trailing twelve-month (TTM) free cash flow deeply negative.[1][2]

Quarterly cash capex (purchases of property and equipment): $77.3 billion in Q3 (vs. $35.1 billion YoY and well above analyst estimates around $57–58 billion). TTM net purchases of PP&E reached approximately $210.7 billion (up 82% YoY), driving TTM free cash flow to an outflow of $42.4 billion (vs. +$14.8 billion prior year). The surge is attributed primarily to AI infrastructure investments.[3][4]

Full-year 2026 expectations: Reaffirmed at approximately $220 billion in cash capex (majority directed to AWS and AI). Q4 operating income guided at $28.5–33.5 billion.[1][2]

2027 directional commentary: Management expects to spend “meaningfully more” than $220 billion on capex, with continued acceleration of payback periods and a “clear line of sight to attractive financial returns” and strong ROIC on the investments.[5][6]

AWS revenue growth and backlog: AWS net sales reached $46.0 billion (+39.3% YoY, fastest pace in nearly five years), achieving a $184 billion annualized run rate. Contracted backlog stood at $578 billion (more than double YoY, triple-digit growth, well diversified).[1][7]

AI-related spend and commentary: Capex is heavily weighted toward AI infrastructure (including Trainium custom chips, Nvidia GPUs, power, data centers, and memory). Specific percentage breakdowns of spend allocation (e.g., Trainium vs. Nvidia vs. power) were not disclosed. Jassy noted installing an AI accelerator roughly every 14 seconds, with capacity snapped up immediately by customers. Inference already accounts for well over half of AWS AI revenue (and growing), with substantial training revenue as well. Custom chips (Graviton and Trainium) delivered >$2 billion in customer savings in Q3; Graviton offers 30–40% better price-performance. Strong Trainium demand prompted plans for external sales late in 2027.[8][9]

Capacity constraints, power availability, and ROIC comments (Jassy and Olsavsky): Jassy stated that 2026 capacity is sold out, 2027 is largely sold out (reserving some for on-demand/EC2/Bedrock), and a good chunk of 2028 is committed. Key bottlenecks include data center space, power availability, and memory (HBM for AI vs. DRAM tradeoffs); he does not expect the shortage to end in 2027. On the earnings call, Jassy described AI as a “once-in-a-lifetime opportunity” that will transform Amazon’s scale, revenue, and free cash flow, with strong conviction on returns despite near-term cash burn. Olsavsky highlighted efficiency gains, capacity optimization, custom chips improving cost structure/margins, and accelerating breakeven/payback periods, while noting AI mix exerts some margin pressure. Both emphasized disciplined investment with visibility into attractive long-term ROIC.[10][11][6]

These details come exclusively from the October 29, 2026 earnings release, 8-K, and call commentary (plus contemporaneous reporting). No other significant post-April 30, 2026 updates on these specific metrics were identified.

Meta’s Q3 2026 print is the quarter the AI build crossed the cash line. Capex including finance-lease principal was $34.91 billion, above $32.79 billion of operating cash flow, so free cash flow was negative $2.12 billion. Management narrowed 2026 capex to $135–145 billion (floor raised $5 billion) and refused a 2027 number, while telling investors the direction is still “aggressively” higher. Returns are framed through ads, Muse, and enterprise inference—not a cloud segment.

Q3 capex crossed operating cash flow

Meta defines the number investors track as purchases of property and equipment plus principal payments on finance leases, because both procure the same servers and data-center kit. In the quarter ended September 30, 2026, that total was $34.91 billion: $33.942 billion of equipment purchases and $964 million of finance-lease principal. Operating cash flow was $32.786 billion. Free cash flow, on Meta’s own definition, was negative $2.12 billion, versus positive $10.63 billion a year earlier, when the same capex line was $19.37 billion.

The ad engine did not stall. Revenue was $64.601 billion, up 26% reported and constant-currency. Advertising was $63.06 billion, also up 26%, with impressions up 10% and average price per ad up 14%. GAAP operating income fell 37% to $12.988 billion only because of a $9.89 billion charge for the August agreement with 52 state attorneys general. Excluding that charge, operating income would have been $22.88 billion, up 11%, and diluted EPS $7.40 instead of $3.83.

• Cash and marketable securities ended the quarter at $82.45 billion; Susan Li put debt at $83.7 billion (long-term debt on the balance sheet was $80.92 billion).
• Headcount was 66,775, down 15% year over year after the May reduction.
• Nine-month 2026 capex on the same definition was $85.824 billion ($83.055 billion of equipment plus $2.769 billion of finance-lease principal).

What this means: the constraint is no longer whether ads can fund a normal capex cycle. One quarter of spending now exceeds a full year of 2024 capex. Anyone modeling Meta has to treat infrastructure cash, not GAAP earnings, as the binding constraint.

2026 capex range: floor up, ceiling unchanged, Q4 implied at $49–59 billion

The October 21 release narrowed full-year 2026 capital expenditures, including finance-lease principal, to $135–145 billion, from the July range of $130–145 billion. The ceiling did not move. The floor has ratcheted all year: $115–135 billion in January, $125–145 billion in April, $130–145 billion in July, $135–145 billion in October.

Subtracting the $85.824 billion already spent through September implies $49.2–59.2 billion of capex in the fourth quarter if the annual range is hit—about 40–70% above the Q3 run rate. Meta did not issue a separate Q4 capex forecast. Full-year 2026 total expenses were raised to $175–179 billion solely to absorb the $9.9 billion legal charge; excluding it, management still expects 2026 operating income above 2025. Q4 revenue was guided to $73–76 billion, with about a 1% FX headwind. Q4 tax rate: 15–17%, absent tax-law changes. Reality Labs losses for 2026 are now expected at $18–19 billion, down from 2025.

• Midpoint of the new capex range ($140 billion) sat in line with the Street’s roughly $140 billion consensus at the time of the print.
• 2025 capex was about $70–72 billion on the prior year’s outlook, so 2026 is roughly a doubling.

What this means: the “narrowing” is a higher floor, not restraint. The second-half bill is back-loaded into a quarter that, on the low end, already exceeds any prior quarter.

Early 2027 guidance: direction, not a number

Meta did not give a 2027 capex figure, a 2027 expense-growth rate, or a full-year 2027 revenue outlook. Li said the company is “still developing our budget and capacity plans for next year” and would share specifics on the Q4 call, as usual. The directional language in the release and on the call is unambiguous: Meta expects to continue investing aggressively in both the core business and emerging AI initiatives. Current capacity plus what is already scheduled to come online is described as flexible enough to shift among model training, core-business growth, and new products.

The only quantified 2027 P&L item is Reality Labs. Losses are expected to narrow to around $15 billion next year, from $18–19 billion in 2026, as spend is reallocated “primarily” to AI. Zuckerberg said VR hardware will ship only on major breakthroughs so the company can pace that investment and prioritize personal superintelligence. On the call, answering Doug Anmuth, Li said planning is early, many proposals have “attractive return profiles over different time horizons,” and the goal is to fund as many as possible inside the overall budget and financial goals. She also said new AI businesses can begin contributing incremental revenue in 2027—subscriptions, transactions, advertising, and enterprise—and scale more meaningfully after that. Serving costs for Muse are expected to rise in absolute dollars even as unit compute efficiency improves.

• No 2027 expense-growth percentage was provided. Do not infer one from the “aggressively” language.
• Revenue framing for 2027 is qualitative only: AI benefits in the core business plus new products “position us to sustain strong overall revenue growth throughout 2027.”

What this means: the Q4 call is the real 2027 budget event. The October message was that the spending cycle is not topping out in 2026, and that the only opex line they are willing to cut is Reality Labs, with the savings redirected into AI.

How Meta prices AI infrastructure without a cloud segment

Hyperscalers with a cloud segment can point at contracted backlog and external GPU rental. Meta cannot. Management’s substitute is a stack of internal and near-internal uses for the same training and inference fleet, so that a large fixed training cost is spread across as much high-margin demand as possible.

Zuckerberg’s capacity frame is consumer, not cloud: Meta is “building out the capacity to serve more than a billion people over the next few years.” Muse is not global yet largely because it is too compute-intensive; he called compute efficiency and the ability to serve a billion-plus users a limiting factor over the next year. Both he and Li said they expect Muse’s compute cost efficiency to improve more than 10x over the next year, via off-peak and spare capacity, smaller models, and inference optimization. Li added that longer term, models should finish more tasks with fewer tokens. Once revenue lines are “at scale,” they expect Muse to be profitable; if efficiency keeps improving for several more years, “highly profitable.”

The return stack they described, in order of maturity:

1. Core ads and recommendations, already paying. The release says AI is already delivering “substantial incremental advertising revenue” through engagement, recommendations, and monetization. On the call, the JEM ranking model drove a 9.6% increase in app ad conversions; generative retrieval on Facebook feed drove nearly 2% more conversions in targeted segments; Andromeda retrieval improvements added 1.4% conversions. Moving content-understanding models onto the first-party Muse family cut per-unit processing costs by more than 80% in offline evaluations. Partnership ads grew over 97% and passed a $15 billion annual run rate. Nearly 10 million advertisers used at least one generative creative tool.
2. Muse, the consumer agent, pre-revenue at scale but with a commerce proof. More than 10 million people in the US and Canada set up Muse in about six weeks; about 25% use it on a given day. Shopping is the top use case—over 25% of users, average purchase above $200. Zuckerberg said Family of Apps revenue for existing Muse users is several hundred dollars a year, and he expects Muse ARPU to exceed that over time because of commerce. The model: free for most people, with businesses paying transaction fees; subscriptions for advanced models and extra tokens; business subscriptions; optional connection into ads and content personalization.
3. Enterprise inference, the “high-margin demand” filler. Because training is a large fixed cost, Zuckerberg said Meta wants to serve as much high-margin inference as possible. Near-term enterprise products are the model API, Muse Code, and business agents, under the Meta Enterprise platform led by CJ Desai. Li pointed to at least 500 million self-identified small businesses already using Meta apps as the installed base. Longer term, a “Muse Network” of consumer and business agents transacting directly. She described this as a potential diversified recurring stream—subscriptions, usage-based API, messaging—that develops over multiple years, not a cloud P&L this year.
4. MSL / frontier models, explicitly longer-dated. Investments in Meta Superintelligence Labs “will take longer to generate returns than those in our core business,” but management cited the breadth of monetization paths plus internal efficiency as the reason for confidence.

What this means: Meta is asking the market to underwrite infrastructure the way a cloud company would, but to score the return the way an ads company plus a consumer-agent company would. The bridge is utilization—ads ranking, Muse inference, and enterprise API sharing one fleet—not a disclosed cloud gross margin. If Muse efficiency misses the 10x, the same capacity either stays idle, gets sold as API, or forces another financing round.

Financing: debt, joint ventures, then a non-voting share class

Operating cash flow is no longer the sole pillar. The quarter closed with cash and securities ($82.4–82.45 billion) essentially matched by debt ($83.7 billion). Long-term debt was $80.92 billion, up from $58.74 billion at year-end 2025, after roughly $24.9 billion of bond issuance during 2026. Buybacks have stopped relative to 2025.

Two structures sit largely off the income statement:

• Data-center joint ventures. Meta holds 20% of two unconsolidated ventures—Louisiana (formed October 2025, about $27 billion of development cost; Hyperion, partnered with Blue Owl) and Texas (about $14 billion; a BlackRock partnership for the El Paso-area campus). It leases the campuses back (starts in 2028 and 2029) and guarantees residual value. Maximum residual-value guarantees rose to $41.49 billion at September 30; total maximum exposure to loss in the two ventures rose to $68.40 billion. Guarantees are carried at zero because payment is “not probable.” Partners fund construction up front; Meta pays rent for up to 20 years and covers any shortfall versus the guarantee at lease end.
• Signed but not yet spent. Bloomberg, adding the 10-Q lines, put future lease and contract commitments at about $825 billion: $454.5 billion of non-cancellable contracts (mostly third-party cloud, servers, network, energy, and Reality Labs hardware) plus $355.7 billion of leases not yet commenced. In July the comparable figure was almost $700 billion. A further contingent cloud obligation of up to $14.7 billion over five years can shrink if the provider sells that capacity to others.

On the call, the new instrument was equity, not another bond. The board approved non-voting Class C common stock, effective January 19, 2027, distributed as a one-for-one stock dividend on Class A and Class B—economically a two-for-one split, with trading expected in late January 2027. Voting power is unchanged. The special committee’s stated purpose is capital raising, employee compensation, and acquisitions “amidst the AI transformation,” while keeping Zuckerberg’s control. Li was explicit: “We do not have any specific financing plans to announce at this time,” but the class was built so Meta has “access to a diverse pool of funding sources.”

Five days later, on October 26, Meta announced it will sell up to $95 billion of stock to finance data centers, starting with about $35 billion via an at-the-market program beginning early next year. That was not on the October 21 call. It is the first use of the framework the earnings release created.

What this means: the capex number in the P&L understates the commitment book. Joint ventures move construction off the balance sheet and replace it with rent plus residual guarantees. Class C, and then the ATM, is how management plans to raise equity without diluting the vote. Credit investors are already in the structure; equity investors are next.

What Zuckerberg and Li said, and how the stock traded

Zuckerberg (release): “We hit major milestones this quarter toward delivering personal superintelligence for everyone—including launching Muse and Muse Wonder, our first frontier model. Our investments are paying off, the core business is strong, and we’re seeing velocity across the board.”

Zuckerberg (call): Muse went from whiteboard to product in about eight months. More than 10 million US and Canada setups in six weeks, number one in the App Store almost every day, ~25% daily use, retention improving with tenure—more than 2x where Threads started. Shopping is the lead use case; Stripe credit limits had to be raised in week one. He tied the business case to commerce fees plus subscriptions, and the capacity case to serving more than a billion people, with a greater-than-10x efficiency gain next year as the condition for profitability at scale. He also said Meta has “not yet found the limit” of applying compute to its own ads and recommendation systems, including using Muse Wonder to improve those systems autonomously. Reality Labs losses should trend toward about $15 billion in 2027 and then toward profitability as VR is paced and glasses become a software and superintelligence business rather than a hardware-margin business.

Li (call): Expense growth in the quarter was driven by legal costs, infrastructure (third-party cloud, depreciation, data-center operating costs), and third-party AI token costs. Capex of $34.9 billion was “driven by investments in servers, data centers, and network infrastructure.” She repeated the $135–145 billion 2026 range, the aggressive 2027 investment stance, and the absence of a 2027 revenue or capex number. On Muse unit economics she would not give a per-user cost, but confirmed total serving costs will rise while efficiency improves more than 10x, using off-peak capacity, smaller models, and inference optimization. On capital structure she described Class C as a tool for raising capital, paying employees, and doing deals without changing the vote, and said there were no specific financing plans that night. On 2027 she said they are evaluating many high-return proposals and will try to fund as many as the budget allows, with quantitative detail deferred to the Q4 call. New AI revenue, she said, can start in 2027 and scale after that.

Market reaction: The tape split in two. After the release on October 21, extended trading was positive—CNBC reported about +2.5%; other coverage described an initial move of more than 4%, with Seeking Alpha citing a gain of as much as 7% as the revenue beat ($64.60 billion vs. about $63.3–63.4 billion) and Q4 guide ($73–76 billion, midpoint above a roughly $73.6–73.7 billion consensus) landed, and the capex ceiling stayed at $145 billion. The regular session the next day reversed. META closed October 21 at $697.52 and October 22 at $648.85, down 7.0%, after trading as low as $639.69. Intraday coverage put the drop at about 7–7.6%. By October 29 the close was $624.52, roughly 10% below the pre-earnings close. Commentary attributed the reversal to negative free cash flow, the higher capex floor and the implied $49–59 billion fourth quarter, the lack of a 2027 capex cap, the Class C governance structure, and a same-day New Mexico court order imposing a $25 billion civil penalty (which Meta said it would appeal). The legal charge had already been flagged; the spending trajectory and the financing architecture were what the next-day sellers priced.

What this means for anyone underwriting the stock or competing for the same power, chips, and capital: Meta just told the market that 2026 capex floors only go up, 2027 will not be quantified until February, and the return case rests on ads already working plus a consumer agent that is not yet global because the company does not yet have cheap enough inference. The financing response—JVs, debt that now matches cash, then a non-voting share class and a subsequent equity program—is the operational answer to having no cloud segment to pre-sell the capacity.

---

Recent Findings Supplement (October 2026)

Meta’s Q3 2026 results (ended September 30, 2026), released October 21, 2026, showed revenue of $64.60 billion (+26% YoY), driven by advertising strength, alongside a $9.89 billion legal charge that produced negative free cash flow for the first time in recent quarters as AI-related capex surged.[1][2]

The company narrowed its full-year 2026 capex range (including principal payments on finance leases) to $135–145 billion (raising the floor from the prior $130–145 billion outlook) while signaling continued aggressive spending into 2027 to support core operations and AI initiatives, without providing full 2027 numbers yet.[3][4]

Q3 capex (including finance-lease principal payments) reached $34.91 billion (+80% YoY), exceeding operating cash flow of $32.79 billion and producing free cash flow of –$2.12 billion.[2][2]

• This compared to $19.37 billion capex and positive $10.63 billion FCF a year earlier; purchases of property and equipment were $33.942 billion plus $964 million in finance-lease principal payments.[5]
• Nine-month capex totaled ~$85.82 billion, implying $49–59 billion for Q4 to hit the updated annual range.[2]
• Spending focused on servers, data centers, and network infrastructure to support AI model training, inference, and capacity expansion (including plans for multiple >1 GW superclusters, with overall infrastructure capacity nearly doubling in 2026).[6]

Meta updated its 2026 capex outlook upward at the low end and provided directional 2027 commentary emphasizing ongoing aggressive investment, while narrowing Reality Labs loss expectations.[4]

• Full-year 2026 capex guidance: $135–145 billion (including finance-lease principal payments).[3]
• Q4 2026 revenue guidance: $73–76 billion.[1]
• Full-year 2026 expenses: $175–179 billion (raised to incorporate the legal charge); excluding the charge, operating income is still expected to exceed 2025 levels.[7]
• Reality Labs operating losses: Expected at $18–19 billion in 2026, narrowing to ~$15 billion in 2027 as investment reallocates toward AI priorities.[4]
• 2027 outlook: No full-year revenue guidance provided yet (still developing budgets and capacity plans), but the company expects to “continue investing aggressively” in core business and emerging AI initiatives, with AI businesses beginning to contribute incremental revenue in 2027 and scaling over time. Strong overall revenue growth is anticipated for 2027.[4][8]

Without a cloud segment, Meta frames AI infrastructure spending as internal capacity for its own models (e.g., Muse frontier model), products, and efficiencies rather than external sales; it highlights early returns via core advertising and longer-term monetization potential.[9]

• AI investments are already delivering “substantial incremental advertising revenue through stronger engagement, better recommendations, and more efficient monetization.”[2]
• New products like Muse (personal AI agent/superintelligence focus) and Muse Wonder saw rapid adoption (10 million users in the US/Canada within ~6 weeks of launch, with ~25% daily usage); these are currently compute-intensive, with expected >10x cost-efficiency gains in 2027 enabling profitability at scale.[3]
• Enterprise opportunities include model APIs, Muse Code, business agents, and longer-term “Muse Network” agent-to-agent transactions; these are positioned as diversified recurring revenue streams (subscriptions, usage-based APIs, messaging) developing over multiple years.[10]
• Broader commitments: Meta disclosed ~$825 billion in future spending obligations (non-cancelable contracts ~$454.5 billion + operating/finance leases not yet on balance sheet ~$355.7 billion + up to $14.7 billion cloud capacity), up from ~$700 billion in July, largely tied to data centers, AI infrastructure, cloud capacity, and energy (many extending through 2036).[11][12]

Financing structures include a new non-voting Class C common stock (effective January 19, 2027) for capital raising, employee compensation, and acquisitions without diluting voting control, alongside prior debt issuance and balance-sheet cash.[1]

• Class C shares will be distributed one-for-one to existing Class A/B holders (economically akin to a 2-for-1 split with no voting-power change); Zuckerberg’s sales of his Class C shares are subject to a Founder Alignment Agreement (20% freely saleable, with matching requirements for the rest).[13]
• The move supports the “significant” scale of AI and growth investments while preserving founder-led control.[2]
• Long-term debt stood at ~$80.9 billion (up substantially YoY, following $25 billion notes issued in May); no share repurchases occurred in the first nine months of 2026.[14]
• No specific joint ventures were highlighted in the earnings materials.

Zuckerberg emphasized AI milestones and investment returns, while CFO Susan Li addressed the timing of AI monetization and 2027 spending direction.[15]

• Zuckerberg: “We hit major milestones this quarter toward delivering personal superintelligence for everyone—including launching Muse and Muse Wonder, our first frontier model. Our investments are paying off, the core business is strong, and we’re seeing velocity across the board.” He noted Muse’s product-market fit and the shift to a scaling/operational phase, with capacity plans aimed at serving more than 1 billion people over time.[16]
• Li: AI/MSL investments “will take longer to generate returns than those in our core business,” but “the breadth of monetization opportunities and internal efficiencies gives us confidence in their long-term potential.” On 2027, the company is “quite early in developing our budget and capacity plans” but expects aggressive investment; more details are anticipated on the Q4 call.[16][10]

Market reaction was mixed: an initial after-hours gain on the revenue beat, strong Q4 guidance, and AI progress (Muse adoption) gave way to a sell-off the next day amid capex/negative FCF concerns, echoing patterns from prior quarters.[17]

• Shares rose ~2.5–3%+ in after-hours trading on October 21.[17]
• The stock then declined sharply on October 22 (reports of ~7–11% drops), with later trading around the mid-$600s amid broader volatility; some analysts viewed the reaction as overdone given core ad momentum and AI traction, while noting pressure on near-term FCF and margins.[18][19]
• Positive sentiment focused on ad pricing/impressions growth (+14% and +10% YoY) and Muse momentum; concerns centered on the pace of AI infrastructure spend outstripping cash generation in the quarter and elevated future commitments.[9]

For competitors or entrants in AI infrastructure or social/ad tech: Meta’s results underscore the capital intensity of building proprietary frontier-scale capacity (no reliance on a cloud revenue segment), the value of data/moat advantages in monetizing AI via core products first, and creative financing/governance tools (e.g., non-voting equity) to sustain multi-year builds without ceding control. Early AI returns are appearing in advertising efficiency, but new revenue streams are framed as multi-year opportunities, with 2027 expected to show incremental contributions alongside continued heavy investment. The $825 billion commitment pipeline signals sustained demand for related hardware, power, and capacity.

The four hyperscalers are Amazon, Microsoft, Alphabet, and Meta. Q3 2026 did not show a demand cut. It showed a delivery split: Amazon absorbed a wave of equipment the Street had not modeled, while Microsoft and Alphabet slipped below estimates because third-party capacity and chips arrived late. Full-year plans barely moved. The 2027 step-up is still the base case, but only if the tally’s universe and the lease definition are kept straight.

Q3 was a timing print, not a capex peak

Amazon’s cash purchases of property and equipment hit $77.3 billion in the September quarter, about $18–20 billion above pre-earnings estimates, and the company still left 2026 cash capex at about $220 billion. That combination only works if Q4 steps down hard: nine-month purchases were $175.7 billion, so the unchanged guide implies roughly $44 billion in Q4, versus $77.3 billion just printed. CEO Andy Jassy separately said 2027 spend will be “meaningfully more” than 2026, so the quarterly spike is a delivery bulge, not the new run-rate. [1] [2] [3]

Microsoft and Alphabet went the other way. Microsoft’s capital expenditures including finance leases were $45.2 billion, below StreetAccount’s $50.67 billion and a Financial Times-cited $51.7 billion, because third-party capacity expected in the quarter was delayed into the first half of 2027. Cash paid for property and equipment was $39.4 billion; finance leases were $7.4 billion. Alphabet’s purchases of property and equipment were $51.36 billion, about $3.7 billion under the $55.09 billion analysts expected, and free cash flow was negative $9.1 billion. Both companies refused to raise 2026 guides. Microsoft cut calendar-2026 capex to about $170 billion from about $175 billion on those delays; Alphabet held $195–205 billion and said 2027 will “increase significantly,” with a number only on the Q4 call. [4] [5] [6] [7] [8]

Meta sat in between. Company-defined capex (purchases of property and equipment plus finance-lease principal) was $34.91 billion — $33.94 billion of cash purchases plus $964 million of lease principal — and free cash flow turned negative at $2.12 billion. The 2026 range was narrowed to $135–145 billion from $130–145 billion, a higher floor, not a cut. The midpoint stayed in line with estimates around $138–140 billion. [9] [10] [11]

Quarterly table: cash purchases vs. company-defined capex

Cash purchases of property and equipment are the only line that is comparable across all four. Microsoft’s and Meta’s headline “capex” add lease items, and those items are not the same thing.

$ billions	Definition	Q3 2025	Q4 2025	Q1 2026	Q2 2026	Q3 2026	Q3 vs Q2	Q3 vs Q3’25	Trailing 4Q (Q4’25–Q3’26), cash
Amazon	Cash PPE purchases	35.1	39.5	44.2	54.2	77.3	+23.1	+42.2	215.2
Alphabet	Cash PPE purchases	24.0	27.9	35.7	44.9	51.4	+6.4	+27.4	159.8
Microsoft	Cash PPE	19.4	29.9	30.9	35.8	39.4	+3.6	+20.0	136.0
Microsoft	Capex incl. finance leases	—	—	31.9	~41.0	45.2	+4.2	—	—
Meta	Cash PPE	18.8	21.4	19.0	30.1	33.9	+3.8	+15.1	104.4
Meta	Capex incl. finance-lease principal	19.4	22.1	19.8	31.1	34.9	+3.8	+15.5	—

Sources for the cells: Amazon cash path and the $77.3 billion print; Alphabet’s five-quarter PPE series ($23.95 / $27.85 / $35.67 / $44.92 / $51.36 billion, rounded above); Microsoft cash from cash-flow history and the $39.37 / $45.2 billion split; Meta cash from cash-flow history and company-defined series of $19.37 / $22.14 / $19.84 / $31.08 / $34.91 billion. [1] [3] [12] [13] [5] [14] [15] [16] [17]

Trailing four quarters of cash PPE sum to about $616 billion. That is not a run-rate for 2027. Amazon’s Q3 is an outlier inside that sum, and Microsoft’s trailing cash figure excludes several billion dollars a quarter of finance leases that its own guidance includes.

What the definitions actually do. Amazon guides “cash capital expenditures,” which tracks purchases of property and equipment. Alphabet’s quarterly figure and its $195–205 billion guide are the same cash PPE line. Meta’s guide adds principal payments on finance leases — a financing cash outflow, not the initial lease addition. Those payments were only $2.77 billion of the $85.82 billion spent in the first nine months, so the gap is small, on the order of 3 percent. Microsoft’s headline capex adds finance-lease additions at inception ($5.6 billion in fiscal Q4 2026, $7.4 billion in fiscal Q1 2027). In July it also lengthened datacenter useful lives from 15 to 25 years, which reclassifies future datacenter leases from finance leases (inside capex) to operating leases (outside capex). That accounting change, not less building, is why the calendar-2026 figure fell from about $190 billion in April to about $175 billion in July. The October trim from $175 billion to $170 billion is the separate delivery delay. Adding the four guides as if they were one GAAP line overstates a pure cash total by roughly Microsoft’s finance-lease slice and slightly overstates Meta. [18] [19] [20] [21]

Consensus: two underspends, one large overspend, one in-line year

Company	Q3 actual (basis analysts use)	Pre-earnings consensus	Variance	Full-year guide vs. prior
Amazon	$77.3B cash PPE	$58.7B FactSet; about $20B above LSEG; one tally at $57.47B	Overspend of roughly $18–20B	Held at ~$220B cash
Alphabet	$51.36B cash PPE	$55.09B (CNBC); $55.1–55.2B in other tallies	Underspend of about $3.7B	Held at $195–205B
Microsoft	$45.2B incl. finance leases	$50.67B StreetAccount; $51.7B cited via FT	Underspend of about $5.5–6.5B	Cut to ~$170B calendar from ~$175B
Meta	$34.91B incl. lease principal	No widely cited quarterly capex consensus in retrieved coverage	Not scored as a quarterly beat or miss	Narrowed to $135–145B from $130–145B; midpoint in line with ~$138–140B estimates

[1] [22] [2] [6] [4] [7] [10] [23]

Markets treated the overspend and the underspends differently, and the cash-flow line explains why. Amazon’s trailing-12-month free cash flow fell to an outflow of $42.4 billion. Alphabet’s quarterly free cash flow was negative $9.1 billion, worse than a near-break-even expectation. Microsoft still produced about $18.6 billion of free cash flow because operating cash flow of $58 billion covered $39.4 billion of cash capex, and the stock rose. Meta’s quarterly free cash flow was negative $2.12 billion. The pattern: a capex “miss” that is a delay is tolerable when free cash flow stays positive; a capex “beat” that consumes operating cash is not. [1] [6] [5] [9]

2026 total: about $730 billion on mixed definitions

Latest company guidance, post Q3 prints:

Company	Latest 2026 guide	Basis	Implied Q4 if the guide is hit
Amazon	~$220B	Cash capex	~$44B after $175.7B in nine months
Alphabet	$195–205B (mid $200B)	Cash PPE	~$63–73B after ~$132B in nine months
Microsoft	~$170B calendar	Capex incl. finance leases	~$52B after ~$118B in Q1–Q3 on that basis; company also guided fiscal Q2 (calendar Q4) at over $50B
Meta	$135–145B (mid $140B)	Incl. finance-lease principal	$49.2–59.2B after $85.8B in nine months
Four-company sum	$720–740B; midpoint $730B	Mixed	Q4 must rise at Alphabet and Meta and fall at Amazon

[24] [1] [6] [4] [21] [3]

Outside tallies for 2026 do not match that $730 billion, because they do not cover the same companies or the same line:

• Bloomberg, reported October 30, put AI spending by America’s tech giants at about $750 billion, funded by ad, consumer, and retail cash flows. A Janus Henderson summary put Amazon, Alphabet, Meta, and Microsoft calendar-2026 capex tracking near $750 billion. Both sit a bit above the post-trim midpoint, consistent with figures compiled before Microsoft’s $5 billion cut or with a slightly broader spend definition. [25] [26]
• S&P Global Market Intelligence consensus, cited in an October 28 Amazon preview, had calendar-2026 hyperscaler capex at $762 billion, up from $362 billion. That set is wider than the big four. [27]
• Goldman Sachs Research, September 23–25, had the five largest U.S. hyperscalers — Amazon, Alphabet, Microsoft, Meta, and Oracle — on track for about $800 billion in 2026. [28] [29]
• Morgan Stanley’s mid-October global model, covering hyperscalers and neoclouds and adding lease investment, put 2026 total investment near $978 billion. [30]
• UBS, via Reuters on October 30, estimated hyperscaler capex up 76 percent this year to $673 billion. That is the low outlier versus company guides and should not be averaged in without noting the gap. [31]

The cleanest 2026 number for the four named companies is the guidance sum: $720–740 billion, midpoint $730 billion, not additive on a single cash basis. A pure cash-PPE total would be a little lower, almost entirely because Microsoft’s $170 billion includes finance leases and because the July reclassification already removed future operating leases from that guide.

2027: company language is “up a lot”; sell-side is $1.1–1.3 trillion depending on who is counted

None of the four has issued a 2027 dollar guide.

• Amazon: spend “meaningfully more” in 2027 than in 2026, with a line of sight to returns, and AWS growth projected above 40 percent in both 2027 and 2028. [32] [3]
• Alphabet: capex to “increase significantly” in 2027; details on the Q4 call. A $300 billion scare-number circulating in coverage is an investor comment, not company guidance. [6] [33]
• Meta: still building the 2027 budget, but “continue investing aggressively.” [34]
• Microsoft: fiscal 2027 capex expected to grow year over year; spend accelerates in the second half of the fiscal year as delayed third-party capacity and higher component costs land. Hood also said the company expects to stay free-cash-flow positive in fiscal 2027. Pre-print analyst expectations had fiscal 2027 capex above $190 billion. [35] [36] [37]

Sell-side aggregates, newest first where the date is known:

Source	Date	2026	2027	Who is in the number	Definition note
CNBC group comment	Oct 29, 2026	—	“Top $1 trillion”	“Hyperscalers,” unspecified	Qualitative
UBS via Reuters	Oct 30, 2026	$673B (+76%)	+25% (implies ~$840B)	“Hyperscalers”	Low vs. company guides; +6% in 2028
S&P consensus in Amazon preview	Oct 28, 2026	$762B	$1.3T	Hyperscalers, set not fully specified in the preview	Consensus estimates
Morgan Stanley	~Oct 15, 2026	~$978B	$1.58T	Global hyperscalers and neoclouds	$1.388T cash capex + $195B lease investment; leases lift the total ~14%
Goldman (Hammond / Snider)	Sep 23–25, 2026	~$800B	$1.2T	Five U.S. hyperscalers incl. Oracle	Above then-consensus of $1.1T; $1.4T in 2028
S&P Global six-company model	Early Sep 2026	$870B	$1.3T	AMZN, GOOGL, MSFT, META, ORCL, SpaceX	Company-level 2027 figures in that model: Alphabet $357B, Amazon $319B, Microsoft $189B, Meta $164B, Oracle $95B, SpaceX $197B

[8] [31] [27] [38] [30] [28] [29] [39]

The spread is mostly scope, not disagreement about Amazon or Google. Strip Oracle, SpaceX, and neoclouds out of the $1.2–1.3 trillion figures and the big four alone are not a published post-earnings consensus in the sources retrieved. The early-September S&P company splits (Alphabet $357 billion, Amazon $319 billion, Microsoft $189 billion, Meta $164 billion) sum to about $1.03 trillion for the four, but those estimates predate Q3 and Microsoft’s guide cut, and Microsoft’s $189 billion in that table is not on the same calendar or lease basis as Amazon’s cash guide. Treat them as a pre-print snapshot, not an updated total.

Morgan Stanley’s lease adjustment is the cleanest illustration of the definition problem at 2027 scale: cash capex of about $1.39 trillion plus $195 billion of lease investment equals $1.58 trillion, and the bank said hyperscalers already hold more than $1.1 trillion of lease commitments that have not started. Those commitments never hit the cash PPE line in the year they are signed. A model that uses only cash capex will look “light” against a model that books the lease. [30] [40]

What this means if you are selling into the build, or trying to fund one

The binding constraint in this print was delivery, not orders. Microsoft cut $5 billion from calendar 2026 because GPUs and third-party halls slipped into the first half of 2027, and it still guided the December quarter above $50 billion and a faster second half. Alphabet underspent in Q3 and still needs $63–73 billion in Q4 to hit an unchanged guide. Amazon overspent by ~$20 billion and did not raise the year, which means Q4 cash capex is guided, by arithmetic, well below Q3. Suppliers should read Q4 purchase orders, not the Q3 headline, and should not treat Microsoft’s lower reported capex as less demand: part of the 2026 “cut” is leases that moved from finance to operating and therefore left the capex line without leaving the build. [35] [18]

Cash coverage is now the competitive filter inside the group. In the September quarter, purchases of property and equipment ran at about 1.8 times operating cash flow at Amazon ($77.3 billion vs. $42.4 billion), 121 percent at Alphabet ($51.4 billion vs. $42.3 billion), and about 104 percent at Meta ($33.9 billion vs. $32.8 billion). Microsoft was at 68 percent ($39.4 billion vs. $58 billion) and remained free-cash-flow positive. That is why a capex undershoot helped Microsoft’s stock and a capex overshoot hurt Amazon’s. Anyone underwriting 2027 capacity — a neocloud, a lessor, a chip supplier — should assume the incremental dollar is increasingly debt- and equity-funded, not operating-cash-funded, at three of the four. [24] [41]

For a 2027 planning number, use two stacks and do not blend them. Big-four company language supports another leg up from a ~$730 billion 2026 midpoint, but no company has put a dollar on it. The sell-side cluster of $1.1–1.3 trillion is a five- or six-name figure (Oracle, and in some models SpaceX). The $1.58 trillion Morgan Stanley figure is global hyperscalers plus neoclouds, with leases added back. Quote the universe in the same sentence as the number, or the total is not usable.

---

Recent Findings Supplement (October 2026)

The four major hyperscalers—Amazon (AWS), Microsoft (Azure), Alphabet (Google Cloud), and Meta—reported Q3 2026 (or equivalent fiscal periods ending ~Sept 30, 2026) results in late October 2026, revealing sharply higher capex driven by AI infrastructure, with mixed beats/misses versus estimates, negative or sharply lower free cash flow at three of the four, and updated 2026 guidance that keeps the big-four total near $720–745 billion.[1][2][3][4]

Most report cash purchases of property and equipment (PPE) from the cash flow statement; Meta explicitly includes principal payments on finance leases in its capex definition and guidance, while Microsoft distinguishes cash additions to PPE (~$39.4B in the quarter) from a higher total capex figure (~$45.2B including leases in some reports). Amazon emphasizes “cash capital expenditures.” These distinctions matter because lease accounting changes (e.g., Microsoft reclassifying some to operating expenses) and financing can shift reported numbers without altering underlying spending.[5][6]

Amazon’s Q3 capex of $77.3 billion (PPE purchases) beat analyst estimates of ~$57–59 billion by a wide margin, but drove TTM free cash flow to a $42.4 billion outflow; the company held its 2026 cash capex guidance at ~$220 billion while signaling “meaningfully more” spending in 2027.[2][7]

• Q3 2026: $77.259 billion PPE purchases (vs. $35.095 billion in Q3 2025 and $54.208 billion in Q2 2026).[7][8]
• Beats: Significantly exceeded FactSet/Street estimates (~$58.7 billion or $57.47 billion).[2]
• 2026 guidance: Reiterated ~$220 billion cash capex (previously raised from ~$200 billion).[1]
• 2027: CEO Andy Jassy stated the company expects to “spend meaningfully more.”[2]
• Context: AWS revenue +39% YoY to ~$46 billion; heavy AI-driven investment outpaced operating cash flow (~1.8x ratio in the quarter).[7]

Microsoft’s calendar Q3 2026 (fiscal Q1 FY2027) cash PPE additions reached $39.4 billion (total capex ~$45.2 billion including leases), below its own ~$50 billion projection due to third-party capacity delays, while free cash flow of $18.6 billion beat estimates; calendar 2026 guidance was trimmed to ~$170 billion (from prior ~$175 billion or higher) partly due to lease reclassifications.[9][5]

• Q3 2026: Cash additions to PPE $39.367 billion (vs. $19.394 billion YoY); total capex cited as $45.2 billion in some reports.[10]
• Miss vs. own guidance/projection: Below ~$50 billion expectation (delays in third-party deliveries).[9]
• 2026 guidance: ~$170 billion calendar year (adjusted lower via accounting changes, not reduced plans); Q2 FY2027 expected >$50 billion.[11]
• Context: Azure +48% YoY; strong FCF beat (vs. ~$8.5 billion Visible Alpha est.) despite spending.[9]

Alphabet’s Q3 PPE purchases of $51.4 billion missed analyst estimates (~$55 billion), widened negative free cash flow to $9.1 billion, and prompted the company to hold 2026 guidance at $195–205 billion while flagging a “significant” increase in 2027.[3][12]

• Q3 2026: $51.355 billion PPE purchases (vs. $23.953 billion in Q3 2025 and $44.924 billion in Q2 2026).[12]
• Miss: Below ~$55.1–55.2 billion estimates.[3]
• 2026 guidance: Unchanged at $195–205 billion (raised earlier in the year from $180–190 billion).[13]
• 2027: CFO Anat Ashkenazi stated capex will “increase significantly”; some market speculation around $300+ billion.[3]
• Context: Google Cloud +~103% YoY to $30.7 billion; capex ~121% of operating cash flow in the quarter.[14]

Meta’s Q3 capex (including finance lease principals) reached $34.91 billion (~$33.94 billion PPE + $0.96 billion leases), pushing quarterly free cash flow negative at $2.12 billion; the company narrowed 2026 guidance upward to $135–145 billion.[4][15]

• Q3 2026: $34.91 billion total capex incl. finance lease principals (PPE $33.942 billion); vs. ~$19.37 billion total in Q3 2025 and ~$31 billion in Q2.[4]
• 2026 guidance: $135–145 billion (narrowed/raised floor from prior $130–145 billion range).[16]
• 2027: No specific guidance yet (still developing plans); expects continued aggressive investment.[17]
• Context: Revenue +26% YoY; capex outran operating cash flow (~106% ratio).[4]

Big-four 2026 capex guidance aggregates to roughly $720–745 billion (Amazon ~$220B cash; Alphabet $195–205B; Microsoft ~$170–175B incl. leases post-adjustment; Meta $135–145B incl. leases), with sell-side estimates for broader hyperscaler/AI spend (often including Oracle or others) ranging from ~$650–800+ billion in 2026 and $1.1–1.58 trillion in 2027.[18][19]

• Company-implied total (midpoints): ~$732.5 billion.[18]
• Goldman Sachs (Sept 2026): ~$800 billion in 2026 and $1.2 trillion in 2027 for five largest US hyperscalers (above consensus ~$1.1T for 2027).[20]
• Morgan Stanley (Oct 2026): ~$980 billion in 2026 and ~$1.58 trillion in 2027 for AI compute investment (cash capex baseline ~$1.388T + leases); cumulative shortfall ~$600 billion through 2028 for top four.[19]
• Broader context: Multiple reports cite >$700 billion collective Big Tech/hyperscaler spend in 2026; 2027 seen as peak cash-burn year before potential moderation.[9]

Prior-quarter context (selected, from recent filings and analyses) shows rapid sequential and YoY ramps: Amazon Q2 2026 ~$54.2B (Q3 2025 $35.1B); Alphabet Q2 2026 ~$44.9B (Q3 2025 ~$24.0B); Meta Q2 ~$30–31B range; Microsoft sequential cash PPE also rising sharply (e.g., prior quarter ~$35.8B in some reports). Full prior-four-quarter series would require pulling individual 10-Q/8-K cash flow statements, but the pattern is consistent acceleration into 2026.[18][8]

For competitors or new entrants, the scale and pace imply massive barriers (data center/power/supply constraints), but also opportunities in financing, efficiency tools, alternative chips/energy solutions, or leasing structures; however, negative FCF at three firms and heavy reliance on debt/equity raises (~$325B noted for three in 2026) highlight execution and return risks if AI monetization lags.[21]

These Q3 reports (published Oct 21–29, 2026) represent the most recent verified updates; 2027 details will likely emerge in Q4 earnings. Aggregate figures can vary by exact scope (cash vs. leases, four vs. five+ firms) and should be cross-checked against primary filings.

The post-Q3 2026 earnings tape is the clearest evidence yet that markets are no longer treating the AI capex boom as self-funding. Cloud growth is still extraordinary, but free cash flow has flipped negative at several of the biggest spenders, debt and lease commitments are rising faster than disclosed balance-sheet leverage, and the revenue needed to earn a normal return on the buildout is not yet in sight. Skepticism is concentrated on financing, depreciation timing, and customer concentration—not on whether chips are selling.

Cash flow is the new grading metric

After the late-October reports, investors stopped rewarding “we are spending more” and started punishing cash burn. Alphabet’s September quarter is the cleanest illustration: revenue rose 25% to $127.6 billion and Google Cloud more than doubled to about $30.7 billion, yet capital expenditure of $51.4 billion exceeded operating cash flow of $42.3 billion and free cash flow was negative $9.1 billion—the second consecutive quarter of cash burn. The stock fell roughly 11–12.6% on October 29, its worst day on record, erasing on the order of $500 billion in market value. Search advertising, still roughly half the company, grew at its slowest pace in five quarters, so the cash engine that is supposed to fund the buildout is cooling while the buildout accelerates. Alphabet held 2026 capex guidance at $195–205 billion and said 2027 spending will rise significantly. [1] [2] [3] [4] [1]

Microsoft was the contrast case that proved the market’s new rule. Fiscal Q1 revenue was $92.3 billion, up 19%; Azure grew 48% to $33.2 billion; commercial remaining performance obligations reached about $677 billion. Cash capex of about $39.4 billion (roughly $45 billion including finance leases) still left free cash flow of $18.6 billion, well above expectations, partly because third-party chip deliveries slipped. The company cut its calendar-2026 capex forecast to $170 billion from $175 billion. Shares held up. The next day it priced a $25 billion senior-note deal—its first U.S. investment-grade bond sale in more than nine years—that was more than four times oversubscribed, with proceeds available for capital expenditure. Cash-flow discipline, not lower ambition, is what the market paid for. [3] [1] [5] [6] [7]

Meta and Amazon landed on the wrong side of the same line. Meta’s Q3 capex, including finance-lease principal, was $34.9 billion against operating cash flow of about $32.8 billion, producing negative free cash flow of roughly $2.1 billion. It raised the floor of 2026 capex to $135–145 billion. Shares fell in the days after the October 21 report. Amazon spent $77.3 billion on capex in the quarter, kept a roughly $220 billion 2026 plan, and said 2027 will be “meaningfully more.” Trailing free cash flow is negative; the stock sold off even as AWS posted its fastest growth in nearly five years. Across the large spenders, the pattern is consistent: operating profits can still rise while cash after investment does not. [8] [9] [10] [11]

What this means for anyone competing in the stack: the binding constraint has shifted from “can you get GPUs” to “can you fund the next gigawatt without a multiple compression.” Positive free cash flow is now a competitive advantage, not a residual.

Debt, leases, and the financing gap are the real balance sheet

Reported long-term debt understates the economic commitment. On October 27 the Bank for International Settlements published Bulletin 140, “Building AI on credit.” It notes that capital expenditure by Alphabet, Amazon, Meta, Microsoft, and Oracle is projected to exceed $800 billion in 2026—around 2.5% of U.S. GDP—and that those five had issued $222 billion of bonds by September 22. Market sources put 2026 securities linked to off-balance-sheet vehicles at $75–100 billion. Moody’s figures cited by the BIS put large-hyperscaler lease commitments at about $1.2 trillion as of mid-2026, including roughly $820 billion for leases not yet started. Spillovers to Treasury and investment-grade markets have so far been small, but the authors warn that sustained issuance at this pace could change that. [12] [13] [14]

Company-level footnotes are larger still. Meta has disclosed commitments approaching $825 billion—non-cancelable contracts of about $455 billion, uncommenced leases of about $356 billion, plus contingent purchases—an increase of roughly $125 billion in three months. Its five-year CDS premium jumped to about 119 basis points after the disclosure. Alphabet has told investors of more than $1 trillion in future purchase and lease obligations in some readings of the latest call. S&P’s Mark Habib said on October 19 that the biggest hyperscalers “likely” carry more AI-linked liabilities than are publicly known, and that the top six (including SpaceX) are set to spend more than $7 trillion on data centers and related capex from 2025 to 2030, with negative free cash flow until at least next year. Nuveen estimates AI-related borrowing already above $460 billion this year and a cumulative financing gap of $4.5 trillion by 2030. Independent tallies of purchase commitments plus leases range from roughly $1.65 trillion (a July Nikkei figure for five firms) to multi-trillion advocacy estimates that mix contractual obligations with true debt; the conservative, ratings-agency numbers are already large enough to matter. [15] [16] [17] [18] [5]

The fragile end of the credit spectrum is already pricing distress. CoreWeave’s unsecured bonds have traded with yields around 15% and prices below 80 cents on the dollar; five-year CDS spreads have exceeded 1,100 basis points. CoreWeave, Nebius, and IREN each lost more than 30% in October in some tallies, and neocloud shares fell again after Microsoft and Alphabet declined to raise capex forecasts. Gartner expects 80% of neoclouds to cease independent operations by 2030. Hyperscalers can still borrow cheaply—Microsoft’s deal proved that—but the marginal AI builder cannot. [19] [20] [21] [22]

Implication: a large share of “AI demand” on hyperscaler income statements is contracted capacity that smaller counterparties must keep funding. If credit tightens further, the first capacity to slip is the leased and project-financed layer, not Azure or Google Cloud’s core.

Depreciation is deferred earnings risk, not a solved accounting issue

The earnings boom partly reflects a timing mismatch. One company’s capex is another’s revenue immediately; the buyer depreciates servers over five to six years. Jefferies strategist Christopher Wood has called this “probably the most earnings-accretive capex boom in history.” Gerard Minack’s work, cited by Reuters, shows AI-linked names accounting for a disproportionate share of the rise in S&P 500 earnings forecasts since early 2025. Nuveen estimates that about 70% of 2026 capex goes to GPUs and servers with short useful lives, and that rising depreciation will weigh on hyperscaler returns on equity steadily into the end of the decade. [23] [24]

Reuters Breakingviews calculated on October 29 that trimming a single year from estimated useful lives would erase some $36 billion of combined net income next year at Alphabet, Amazon, Meta, and Microsoft, theoretically about $800 billion of market value at current multiples. Most of the group now uses five to six years (Alphabet and Microsoft six; Meta 5.5; Amazon five to six, after shortening a subset from six to five in 2025 and taking a higher depreciation charge). Sixteen of seventeen public-company life changes since 2020 lengthened schedules. Michael Burry has argued the extension understates depreciation by on the order of $176 billion from 2026 to 2028 relative to a two- to three-year Nvidia product cycle. The counter-argument is empirical rental life: older GPUs are still earning, and Jensen Huang has said economic life can run to eight years, with a 1-gigawatt cluster costing roughly $50 billion and generating about $40 billion of annual revenue. Accounting lives and economic lives can both be “right” for different purposes; the risk is a synchronized shortening if utilization or residual values disappoint. [25] [26] [27] [28] [29]

For entrants, the practical point is that reported cloud margins overstate cash returns until the depreciation wall arrives in 2027–2030. Models that treat six-year straight-line as conservative are making a residual-value bet, not a free lunch.

Power and supply are capping the boom even if demand is real

The buildout is not only a financing story; it is physically late. Microsoft’s CFO said third-party AI-chip delays will push some capacity into the first half of 2027, which is why quarterly capex undershot and the full-year forecast was trimmed. Alphabet said outside computing power “came online only at the very end of Q3.” Those are supplier constraints, not cancelled orders—but they still defer the revenue that is supposed to justify the spend. [1] [1]

Power is the slower constraint. Goldman Sachs, in an October briefing reported by Axios, called power “probably the forever bottleneck,” expects global data-center power demand to rise 170% from 2025 to 2030, and puts capital spending across computing, data centers, and power at roughly $7.6 trillion from 2026 to 2031. Morgan Stanley has estimated a U.S. net power shortfall on the order of 32–38 gigawatts by 2028 even after behind-the-meter generation and other workarounds. North American data-center vacancy hit a record low of 1.4% in the first half of 2026, with most capacity under construction already pre-leased (CBRE). Gas-turbine slots at GE Vernova, Mitsubishi, and Siemens are quoted into 2029–2031. Generator step-up transformer lead times have stretched past 160 weeks. Goldman has also said only 50–60% of planned data-center capacity will come online as expected over the next two years. PJM capacity auctions have cleared at the price cap and still missed reliability targets. [30] [31] [32] [33] [34]

The non-obvious implication: a power-constrained boom can look like overbuilding in financial statements (assets under construction, leases signed, chips ordered) while still being supply-short in the market that matters. GPU rental rates have continued to rise—UBS noted older GPUs up nearly 30% year to date and two-year-old GPUs up about 35%—which is evidence of scarcity, not glut. The overbuild risk is that capital is committed to sites that cannot be energized on the schedule embedded in the return model. [35]

Customer concentration and the revenue-versus-spend gap

The revenue that is supposed to close the loop is growing fast and is unusually concentrated.

Microsoft recorded $10 billion of revenue from commercial arrangements with OpenAI in the September quarter—about 11% of total revenue—and $24.1 billion in the fiscal year ended June 2026. OpenAI owed Microsoft $7.2 billion at quarter-end. OpenAI has committed to an additional $250 billion of Azure purchases. Earlier disclosures showed Microsoft’s commercial backlog growth collapsing from the mid-80s percent to about 25% once OpenAI was excluded. That is not a side customer; it is a related party (Microsoft owns roughly a quarter) whose spending and revenue-share both flow through Redmond’s income statement. [36] [37] [38]

Anthropic’s S-1, filed in late October ahead of a planned IPO, makes the mismatch explicit. First-half 2026 revenue was about $16.3 billion, up roughly 1,300% year over year, with Q2 revenue of $11.6 billion. Unit economics improved: about $2.08 of revenue per dollar of compute cost in the second quarter, versus $0.40 a year earlier. At the same time, future compute and chip commitments run from at least $475 billion to as much as $750–800 billion depending on which contracts are counted as firm, including large non-cancellable minimums with Amazon, Google (at least $114 billion in one tally), Microsoft (about $45 billion), Broadcom, and others. Much of Anthropic’s revenue is consumption-based and cancellable on short notice; much of its cost is not. Top-five customers were about 21–25% of recent revenue, down from higher levels earlier in the year, but Ramp data show the top 1% of firms accounting for about 80% of commercial spending on OpenAI and Anthropic, with token dollar spending nearly flat in August and September. Press reports put OpenAI’s September annualized revenue near $50 billion, below earlier signals approaching $70 billion; definitional differences (gross versus net of marketplace fees, revenue-share treatment) make lab-to-lab comparisons noisy. [39] [40] [41] [42] [43] [44] [45]

Independent estimates of the gap are stark. Gartner said on October 14 that AI data-center owners would need $8.5 trillion of cumulative revenue through 2030 to cover announced spending, financing costs, and a 12% return, against a forecast of no more than $3.6 trillion of attributable enterprise and consumer AI revenue—a $4.9 trillion shortfall. More than $3.5 trillion of infrastructure spending could be exposed to impairment by the end of 2030, and more than $4 trillion of asset value by 2032, if the spending path continues. Bain, in late September, said the industry needs $6 trillion of annual revenue by 2031 to justify data-center capital, with existing services perhaps supplying $1.8 trillion and $4.2 trillion still to be created. A Goldman equity note argued the five hyperscalers plus SpaceX need an extra $300 billion a year in AI services revenue in 2026 and 2027 just to break even on capital spending. One research tally put confirmed AI revenue run-rates at roughly 20% of annualized big-four capex, or about 26% using later press figures for the labs. [46] [47] [48] [49] [50]

Apollo’s Torsten Slok framed the bear case in economic rather than accounting terms in notes around October 22–24. The “core economic flaw,” he argued, is that the marginal cost of producing compute exceeds the marginal benefit delivered to users, so demand is not unlimited; the compute demand curve looks flat and price-sensitive. Outside the Magnificent Seven, he said AI has delivered “zero financial payoff”: revenue per worker and profit margins are rising for the mega-caps and flat for the rest of the S&P 500, with no clear lift yet in utilization-adjusted productivity. Nuveen adds that only about 15% of U.S. employees use AI daily and that more than 90% of generative-AI pilots have yet to show a measurable profit impact. [19] [23] [51] [52]

Circular flows amplify the measurement problem. Hyperscalers and chip vendors invest in labs; labs sign multi-hundred-billion-dollar take-or-pay cloud deals; those deals show up as backlog and, eventually, as the investors’ own cloud revenue. NVIDIA sits on more than one side of several of these arrangements. The dollars are real. The question is how much is incremental end-demand versus recycled supplier financing.

Bear case, stock reaction, and the serious counterpoints

The bear case, as of October 30, 2026, is not that AI demand is fake. It is that:

• Returns are being pushed out while the cost of capital has risen (10-year yields have been cited above 5% in recent market commentary).
• A large fraction of current S&P earnings growth is the capex cycle itself and will reverse if spending growth merely slows, not if it collapses. Goldman has put hyperscaler capex near $800 billion in 2026, up on the order of 90% from 2025, heading toward $1.1–1.3 trillion in 2027, and has linked roughly half of index earnings growth to AI infrastructure. [53] [23]
• Core ad, commerce, and device businesses that fund the bet are slowing and, in places, being cannibalized by the products the capex is building. [4] [54]
• Credit is already sorting winners from leveraged intermediaries. Alphabet’s record one-day decline, Amazon’s post-earnings drop, Meta’s negative free cash flow, and neocloud CDS at distress levels are the market’s scorecard. Microsoft’s flat-to-up reaction and oversubscribed bond deal are the exception that defines the rule.

The counterpoints are equally specific, and they are why this is not a clean “bubble popped” story:

• Capacity is still scarce. Chip delays and power queues are pushing revenue right, not destroying it. Azure, Google Cloud, and AWS growth rates remain historically high, and management teams describe AI capacity as sold out or supply-constrained.
• Backlogs are enormous relative to near-term capex. Microsoft’s commercial backlog is $677 billion; Alphabet has cited cloud backlog in the hundreds of billions, with a large share converting inside 24 months. That is contracted demand, not a slide-deck TAM.
• The largest spenders still have cash buffers and investment-grade access. Consensus work cited by wealth managers shows combined free cash flow for the group turning negative in 2026–27 and recovering later as spend growth slows and revenue catches up; Microsoft is the outlier that stays positive. Alphabet and Amazon’s net cash still covers a year or two of projected shortfalls in those models. [55]
• Lab economics are improving even as absolute losses and commitments explode. Anthropic’s revenue-to-compute ratio has moved through 1x. If that holds as models get cheaper to serve, the coverage ratio can rise without capex falling.
• Historical capex manias often leave useful infrastructure even when equity holders in the peak vintage lose money. The relevant risk for public shareholders is an industrial overbuild—technology works, late capital does not earn its cost—not a demand collapse.

Bottom line for anyone allocating capital or trying to compete. The evidence that the boom is poorly returning so far, relative to the capital committed, is strong and recent: negative free cash flow at Alphabet, Meta, and Amazon; multi-trillion lease and purchase commitments; Gartner and Bain shortfalls measured in trillions; lab revenue still an order of magnitude below the spend rate; and a credit market that has already marked neoclouds as the weak link. The evidence that it is simply overbuilt relative to end demand is weaker: rental prices are up, cloud growth is accelerating, and the binding constraints in the October prints were chips and power, not empty racks. The trade after Q3 is selectivity—funders with cash flow and energized power win; leveraged capacity that depends on one or two labs, and on depreciation lives that must stay long, is where the poor returns show up first.

---

Recent Findings Supplement (October 2026)

Post-Q3 2026 earnings (reported late October 2026) show mounting evidence of strain in the AI capex cycle, with hyperscalers adjusting or holding aggressive spending plans amid delivery delays, negative free cash flow at several firms, and sharp stock selloffs.[1][2]

Microsoft lowered its full-year 2026 capex forecast to $170 billion (from $175 billion) citing third-party AI chip delays that push capacity into H1 2027; Alphabet held its $195–205 billion range but delivered negative quarterly FCF exceeding $9 billion after capex outpaced operating cash flow.[1] Meta raised its 2026 capex range to $135–145 billion while reporting its first negative FCF in over a decade ($2.12 billion burn in the quarter). Amazon maintained ~$220 billion guidance but showed slowing core e-commerce and ad growth.[2][3]

These updates followed a pattern where both raising and cutting capex guidance triggered investor punishment (3–7%+ drops in several cases), reflecting uncertainty over the “right” spending level to balance buildout speed against returns.[4]

• Alphabet shares fell ~10% (extended trading) and later reports noted a record plunge wiping out nearly $500 billion in market value; Amazon down ~7% post-earnings; Meta down 7–11% since its report.[1][2]
• Broader context includes prior-year cumulative AI capex by major players already matching or exceeding post-ChatGPT totals, with 2026 hyperscaler spend projected at $1.1–1.3 trillion.[5]

This dynamic raises the bar for proof of monetization: investors now penalize both over- and under-spending, shifting focus from narrative momentum to demonstrable ROI on infrastructure that is increasingly debt- or commitment-financed.

Free cash flow compression is materializing exactly as capex surges, with multiple hyperscalers reporting or projecting negative or sharply lower FCF in 2026, forcing greater reliance on balance sheets, new equity classes, or external borrowing.[6][7]

Alphabet’s Q3 capex reached 121% of operating cash flow (negative FCF ~$9 billion); Meta turned negative at $2.12 billion; collective hyperscaler FCF is projected to drop ~90% in 2026.[8][6] Microsoft remained positive (~$18.6 billion FCF in its fiscal Q1) but still guided elevated spend. AI-related borrowing has already exceeded $460 billion in 2026, with estimates of a cumulative financing gap reaching $4.5 trillion by 2030.[8]

Meta disclosed plans for a new class of non-voting stock partly to fund AI investments and reported rising credit default swap premiums; neocloud providers face higher costs of capital amid debt concerns.[5][3] Off-balance-sheet elements include large lease commitments (e.g., Meta referenced hundreds of billions in leases not fully on-balance-sheet) and non-cancelable cloud compute contracts.

• Consensus models show several of the five major cloud firms facing a combined ~$93 billion cash-flow gap in 2026 (capex $800+ billion vs. operating cash flow).[9]
• Projections indicate FCF turning negative or remaining pressured into 2027 before potential recovery in 2028 as revenue catches up.[10]

For competitors or entrants, this compresses the window for self-funded growth; only the strongest balance sheets (or those with diversified cash cows) can sustain the pace without diluting or increasing leverage, favoring incumbents with scale while pressuring pure-play or debt-dependent players.

Depreciation accounting choices are drawing renewed scrutiny as a potential amplifier of future earnings pressure, with analysts noting that modest extensions of useful life for AI servers have already muted the income-statement hit—but any reversal could erase tens of billions in combined net income.[11]

A Reuters Breakingviews analysis highlighted that trimming one year from estimated useful lives of AI assets could erase ~$36 billion in combined net income for the Big Tech group next year (implying substantial market-cap impact at current multiples). Meta alone amortized >$14 billion on servers/gear in the first nine months of 2026.[11]

Longer amortization schedules are reducing the current earnings drag from heavy capex, but this invites questions about whether the schedules align with actual technological obsolescence or utilization.[12]

• Damodaran’s commentary (around early October 2026) noted capex-driven distortions where book equity and debt have risen sharply, with earnings growth outpacing FCF growth.[12]
• As more assets come online, rising depreciation charges are expected to weigh on returns on equity through the end of the decade.[8]

This creates a timing mismatch risk: current reported profits may overstate sustainable economics if hardware life cycles prove shorter than assumed, complicating valuation and capital allocation decisions for firms and investors.

Power and grid constraints have emerged as a binding, longer-lead-time bottleneck that is already causing project delays, cancellations, and queue pauses—shifting the limiting factor from chips to electricity supply and transmission.[13]

A Delphi Digital report (Oct 29, 2026) titled “The AI Boom Has a Power Problem” argues electricity and grid infrastructure now cap AI scaling more than semiconductors. ERCOT’s large-load queue surged to 474 GW by June 2026 (~90% data centers); PJM capacity auctions cleared at the price cap with shortfalls of 6.6–6.8 GW (first such shortfalls in history).[13][14]

Goldman Sachs warns power is the “forever bottleneck,” projecting global data-center power demand up 170% by 2030 (majority U.S.-driven) and ~$7.6 trillion in related capex through 2031; dozens of projects canceled or delayed since January 2026, many due to local opposition.[15][16] Grid interconnections take 24–72 months (sometimes longer), with transformer and turbine lead times extending into the early 2030s.

• North American data-center vacancy hit record lows (1.4% H1 2026) while under-construction capacity is heavily pre-leased; ~7 GW of planned U.S. capacity for 2026 already delayed/canceled due to power.[17][18]
• Developers are turning to behind-the-meter generation (e.g., gas turbines) and alternative sites, but manufacturing capacity for turbines and related equipment is constrained.[14]

This extends the payback period on capex and raises the risk of stranded or underutilized assets if power availability lags demand; it also advantages players with access to favorable power contracts or co-located generation.

Customer concentration risks are explicit in Anthropic’s October 2026 S-1 filing ahead of its planned IPO, revealing massive non-cancelable compute commitments alongside revenue concentration and hyperscaler “frenemy” dynamics.[19]

Anthropic disclosed at least $475 billion (up to ~$750 billion including leases) in future cloud compute purchases, mostly non-cancelable, with major deals including Amazon (~$110B+), Google (~$114.5B+), Microsoft (~$44.9B+), and others (SpaceXAI up to $120.5B). Top-five clients represent ~25% of revenue; H1 2026 revenue reached $16.28 billion (up massively YoY) but net losses widened to $80.7 billion (from $13 billion).[19][20]

The filing flags that hyperscalers (AWS, Google Cloud, Azure) are simultaneously suppliers, investors, and direct competitors developing their own models, creating risk of compute rationing or repricing. Estimates suggest OpenAI and Anthropic together underpin 30–50% of certain hyperscaler cloud backlogs/revenue.[21]

• Anthropic’s quarterly sales growth slowed to 50% (from 140% earlier in the year); prior $2 trillion valuation talk has met skepticism (some investors eyeing ~$1.5 trillion).[22]
• Gartner estimated potential $6 trillion revenue shortfall vs. infrastructure spend by 2032, with ~$4 trillion at risk of write-off; hyperscalers + SpaceX may need an extra ~$300 billion in annual AI services revenue in 2026–2027 just to break even on capex.[22]

Bear-case arguments center on overbuild relative to monetizable demand, financing stress, and concentration creating correlated risk; counterpoints emphasize contracted (but unrecognized) revenue backlogs exceeding $2 trillion, strong core-business cash flows at some firms buffering spend, and the possibility that utilization or new applications close the gap over time.[8][10]

Former SEC Chair Gary Gensler warned (late October 2026) that history suggests a pullback within 18–30 months, with debt levels determining severity.[23] Market sentiment has shifted toward selectivity and proof of returns rather than capex announcements alone.[24]

These October 2026 developments—particularly the earnings reactions, Anthropic disclosures, power reports, and accounting scrutiny—provide the most concrete recent signals of potential overbuild or suboptimal returns in the AI infrastructure boom. Earlier 2026 data showed acceleration; the post-Q3 period highlights the emerging tension between spend and realized economics.