Research what prominent hedge fund managers and asset managers…

Bill Ackman’s Pershing Square has concentrated ~30-40% of its portfolio in AI infrastructure leaders (Meta, Amazon, Alphabet, and a new $2.4 billion Microsoft stake initiated in February 2026), viewing heavy capex not as margin risk but as a J-curve investment in a competitive arms race that will reward scale winners at still-reasonable valuations.[1]

Ackman entered Microsoft after its stock dropped ~10-15% post-Q2 earnings on Azure growth concerns and $190 billion 2026 capex plans, buying at 21x forward earnings—below historical averages—and explicitly valuing Microsoft’s ~27% economic interest in OpenAI at ~$200 billion (7% of market cap). He frames AI as essential for corporate survival: firms ignoring it “are going to fall behind.” His prior buys in Meta, Amazon, and Alphabet followed the same logic of discounted long-duration cash flows augmented by AI upside.

• Portfolio snapshots show 38%+ allocation across three Magnificent Seven names as of late 2025, with Meta added in Q4 2025 at ~20-22.5x forward earnings.[2]
• Recent X posts and interviews emphasize that AI spending reflects structural competition rather than speculation, with upside in cloud, advertising, and productivity tools.[3]

For new entrants or competitors: Ackman’s approach shows that disciplined valuation entry into AI-exposed compounders (not pure hype plays) can still work even amid capex fears—focus on economic interests in foundational models and verifiable revenue moats rather than narrative momentum.

Ray Dalio has repeatedly labeled the AI boom as entering its “early stages of a bubble” (January 2026 X post), comparing euphoria levels to ~80% of the dot-com peak or 1929, while Bridgewater maintains selective long exposure to AI enablers like Nvidia, Lam Research, Salesforce, and Alphabet.[4]

Dalio’s mechanism insight: investors mistakenly bet on the technology itself rather than the companies that will survive intense competition; most will fail even as the underlying innovation endures. He notes the bubble may persist until the Fed tightens policy, and Bridgewater research highlights that AI capex is already supporting U.S. growth with second-order effects (productivity, capital allocation) not fully priced in.

• January 2026 retrospective explicitly tied 2025 tech gains to the early bubble phase while gold and non-U.S. assets outperformed.[5]
• March 2026 comments reinforced that “most companies won’t survive” the shakeout despite transformative potential.[6]

For competitors: Dalio’s stance illustrates the value of macro-aware hedging—own the picks-and-shovels (chips, software) but size positions modestly and prepare for a multi-year winner-take-most consolidation rather than broad AI equity outperformance.

Michael Burry has executed sizable put-option shorts on Nvidia and Palantir (disclosed in late 2025 13F filings, with notional values in the hundreds of millions but actual premium cost far lower, e.g., ~$9 million for the Palantir position) while publicly comparing the current AI mania to the final months of the 1999-2000 bubble.[7]

Burry’s core thesis: returns on invested capital for Big Tech AI spend are declining, valuations have detached from fundamentals, and the market ignores data (e.g., consumer sentiment) in favor of nonstop AI narrative. He maintains leveraged shorts against overvalued names and has warned of widespread AI-company bankruptcies and a 2026-2027 panic.

• Philadelphia Semiconductor Index (SOX) trajectory explicitly likened to the pre-March 2000 collapse.[8]
• Ongoing Substack and X commentary through spring 2026 urges investors to “reject greed” as momentum trades dominate.

For market participants: Burry demonstrates how asymmetric, low-premium options can express high-conviction bubble skepticism with limited downside—useful for hedging concentrated long AI exposure without needing to time the exact top.

Ken Griffin’s Citadel has aggressively added to Nvidia (nearly 120% increase, now largest disclosed equity holding at ~2.8% of portfolio) and Amazon while openly acknowledging “hype” and the capital-driven pressure to oversell AI’s near-term productivity gains.[9]

Griffin’s balanced mechanism: AI will deliver real but uneven results (strong in coding and call centers, mixed elsewhere), yet infrastructure spending ($500+ billion projected for U.S. data centers in 2026) must be justified by transformative promises. He bets on the hardware and cloud leaders positioned to capture that spend rather than dismissing the cycle.

• Q4 2025 13F activity shows Nvidia and Amazon as top convictions; other AI-adjacent names also appear in holdings.[10]
• Davos/WEF remarks (early 2026) explicitly flag hype while projecting sustained infrastructure demand.

Implication for entrants: Griffin’s playbook rewards identifying the capital-intensive bottlenecks (chips, data centers, cloud) where spending is already committed and hard to displace, even when narrative excess is evident.

Elliott Management (Paul Singer) has hedged AI exposure by shorting Nvidia via puts (early 2025 filings) after client letters labeled it “bubble land” and AI “overhyped,” while selectively moving into chip-design enablers like Synopsys and value-oriented names.[11]

Singer’s approach prioritizes cash-flow-anchored businesses and avoids Mag7 momentum; 2025 performance lagged broad indices partly due to this defensive stance. Recent activist steps into Synopsys reflect conviction that AI-driven chip complexity will reward specialized players in the supply chain.

• 2024-2025 letters warned that poor Nvidia results would “break the spell” and highlighted staying away from bubble valuations.[11]
• Portfolio tilt toward hard assets and infrastructure underscores skepticism of narrative-driven multiples.

For competitors: Elliott shows that pairing targeted shorts on the most stretched names with long positions in less-hyped parts of the AI stack (EDA tools, etc.) can protect capital during euphoria while still participating in secular growth.

Bull vs. Bear Split Summary (through May 2026):

The clearest bulls are Ackman and Griffin—concentrated longs in AI leaders at what they view as attractive entry points, treating capex as strategic investment. Dalio and Burry sit firmly on the bear side (early-bubble diagnosis and outright shorts). Elliott is bear-leaning with explicit shorts and client warnings but has pivoted to selective longs in the ecosystem. Bridgewater occupies the middle: Dalio’s public caution paired with institutional exposure to enablers.

This split highlights two distinct mechanisms at work: (1) data-moat and scale advantages that allow a few firms to monetize AI spend (Ackman/Griffin thesis), versus (2) historical pattern of most companies failing to capture transformative technology value, creating classic bubble conditions (Dalio/Burry/Singer view). For anyone building or competing in this space, the practical takeaway is to size AI exposure by conviction in specific cash-flow or competitive advantages rather than broad sector beta, and maintain hedges or dry powder for the inevitable shakeout that multiple veteran managers now anticipate.

---

Recent Findings Supplement (May 2026)

Ray Dalio views the AI boom as entering an early bubble phase but emphasizes the enduring value of the underlying technology. In January 2026, the Bridgewater founder warned on X that the AI-driven rally had pushed indices to records while entering “the early stages of a bubble,” comparing the euphoria level to roughly 80% of prior peaks like 1929 or 2000. By March 2026, in the All-In Podcast, he refined this to acknowledge that most individual AI companies will fail amid competition, even as the technology transforms the economy.[1]

• Dalio noted in March 2026 that investors mistakenly bet on companies rather than the technology itself, with only a small percentage of firms surviving the shakeout.[2]
• He highlighted that massive spending risks “eating itself” without adequate profits, though he did not advocate selling holdings outright.[3]

This positions Bridgewater as a cautious macro observer: the mechanism is classic bubble dynamics (pricing, ownership concentration, financing), but the implication is selective survival—winners will emerge from infrastructure and applications that deliver measurable ROI. For new entrants or competitors, it underscores the need to demonstrate concrete productivity gains rather than narrative-driven capex.

Michael Burry has intensified warnings that the AI rally resembles the final months of the 1999-2000 dot-com bubble, urging investors to trim exposure. In May 2026 Substack posts and X commentary, the famed short seller described nonstop AI coverage and parabolic stocks as unsustainable, advising to “reject greed” and “reduce positions almost entirely” in momentum names. He has maintained this stance since late 2025, including earlier put options on Nvidia and Palantir.[4]

• By April 2026, Burry doubled down, calling the bubble “too big to save” even if government intervention is attempted.[5]
• He compared recent Philadelphia Semiconductor Index (SOX) action directly to the pre-March 2000 collapse.

Burry’s mechanism relies on historical pattern recognition—euphoria detached from fundamentals—creating asymmetric downside risk. The implication is that broad AI exposure could face sharp mean reversion; competitors or allocators should stress-test portfolios for valuation compression and prioritize non-AI or value-oriented holdings as a hedge.

Bill Ackman sees AI as a productivity boom rather than a bubble, actively building positions in key beneficiaries like Microsoft and Meta. In a late April 2026 “This or That” exchange, he explicitly chose “boom” over bubble. Pershing Square’s February 2026 investor presentation highlighted Meta’s AI upside as underappreciated, calling the valuation “deeply discounted” despite heavy 2026 capex (up to $135 billion). Ackman added Microsoft exposure during 2026 sell-offs, viewing subscription resilience and OpenAI ties as underpriced.[6]

• He sold Google to fund Microsoft buys, citing AI/cloud strength amid broader sector weakness.
• Ackman has framed the environment as an “industrial-scale productivity boom,” betting billions that long-term earnings growth will justify spending.

The mechanism here is data-driven conviction in durable moats (cloud, advertising, compute demand). For market participants, it signals that selective long exposure to infrastructure leaders can outperform if productivity materializes faster than skeptics expect.

Ken Griffin acknowledges AI investment hype as a fundraising necessity while highlighting accelerating real-world productivity gains. At Davos in early 2026, the Citadel CEO noted that $500 billion in projected U.S. data-center spend requires “promising to profoundly change the world,” labeling much of the jobs-panic narrative “all garbage.” By May 2026 at the Stanford Leadership Forum, he described a “step change” in agentic AI over the prior nine months, with systems now handling weeks-to-months of high-skilled finance work (master’s/PhD-level analysis) in hours or days.[7]

• Griffin observed a power shift toward tech teams within corporations as AI tools mature.
• He remains pragmatic: hype fuels capital but actual deployment is delivering measurable efficiency.

This dual lens—narrative for capital raising versus tangible output—suggests the bubble debate is partly semantic. Competitors must separate infrastructure spend justification from operational integration to avoid over- or under-investment.

Paul Singer’s Elliott Management has maintained short exposure to AI leaders like Nvidia, consistent with earlier “bubble land” and “overhyped” characterizations, though fresh 2026 public commentary has been limited. The firm’s 2025 client letter and subsequent positioning targeted Nvidia puts and semiconductor momentum, viewing valuations as detached from realizable value. Recent portfolio commentary (early 2026) reinforces preference for hard assets and cash-flow anchors over AI growth stories.[8]

Elliott’s mechanism is classic activist/short-selling discipline: identify over-extrapolated multiples and hedge accordingly. The implication for the broader landscape is that dedicated shorts provide a natural counterweight, potentially pressuring valuations if fundamentals disappoint.

Overall bull-bear split among these prominent investors shows a clear divide, with bears focused on valuation extremes and survival rates while bulls emphasize selective opportunity and productivity realization. Dalio and Burry (plus Singer/Elliott) lean bearish, warning of early-to-late-stage bubble dynamics and the likelihood that most AI companies will not endure. Ackman stands out as strongly bullish, deploying capital into perceived AI winners. Griffin occupies a middle ground, discounting hype-driven narratives but confirming accelerating real capabilities.

This split implies that AI investment decisions now hinge less on blanket optimism or pessimism and more on distinguishing durable infrastructure and application leaders from speculative noise—favoring rigorous ROI tracking for any new allocation or competitive strategy.