Source Report

Analyst Survey Data Signals Widespread ROI Shortfalls

Gartner's surveys reveal a stark mechanism driving enterprise AI fatigue: organizations rush into agentic AI proofs-of-concept fueled by vendor hype, but scaling exposes immature models lacking sustained autonomy for complex tasks, leading to stalled pilots and outright abandonment. A April 2026 survey of 782 infrastructure leaders found only 28% of AI use cases fully succeed with ROI, while 20% fail completely; a May 2025 CIO poll showed 72% breaking even or losing money on investments.[1][2] Forrester echoes this, predicting 25% of 2026 AI spend deferred to 2027 as just 15% of decision-makers report EBITDA gains, forcing CFOs to demand measurable outcomes over experimentation.[3][4]

• Gartner's June 2025 prediction: >40% of agentic AI projects canceled by 2027 due to costs, unclear value, risk gaps—early experiments misapplied hype without production readiness.[5]
• MIT Project NANDA (2026): 95% of genAI investments yield zero measurable return; DDN survey: >50% of AI projects delayed/canceled in past two years over infrastructure complexity.[3]
• 63% of orgs lack AI-ready data practices (Gartner July 2024 survey), dooming 60% of unsupported projects by end-2026.[6]

Implication for entrants: Skip broad genAI pilots; target narrow, data-mature use cases like procurement analytics where ROI hits in <12 months—avoid the 72% breakeven trap by validating with existing infrastructure first.

Project Cancellations Reflect Hype-Reality Mismatch

Agentic AI—autonomous systems for multi-step enterprise tasks—promised workflow revolution but delivers via hype-driven pilots that crumble on cost/complexity at scale: early PoCs dazzle with demos, but production demands governance, data layers, and risk controls most lack, prompting pauses. Gartner forecasts 40%+ cancellations by 2027; half of 2026 US data centers (AI-powered) already delayed/canceled over supply/power shortages.[5][7]

• Over half of enterprise AI stalled last two years on infra mess (DDN/Google Cloud/Cognizant survey of 600+ leaders).[3]
• UBS: Only 11% of AI projects reach production in two years; most eye 2026/H2 2025.[8]
• No named firm-wide pauses, but patterns: 95% genAI pilots fail value (APQC); Deloitte: 25% convert pilots to prod.[9]

Implication for competitors: Incumbents pause via attrition (e.g., AI layoffs fund infra, not expansion); new entrants win by offering "production-ready" stacks with built-in governance—target the 60% abandoning hype plays.

Consumer Metrics Hint at Monetization Fatigue

ChatGPT's user base exploded to 900M weekly actives by Feb 2026 (from 400M Feb 2025), but growth slowed late-2025 vs rivals like Gemini; revenue hit $25B annualized (Feb 2026, up from $20B 2025), targeting $29.4B full-year—yet missed 2025 revenue/user goals (e.g., 1B users), with free users (95%) burning compute while paid subs stagnate at ~5%.[10][11] OpenAI CFO flags compute contracts at risk amid shortfalls.[12]

• Market share dip: 68% Jan 2026 (from 87% Jan 2025); mobile rev $1.35B 2025 (673% YoY) masks inference losses even on $200 Pro plans.[13]
• Enterprise pivot: API/subs drive growth, but consumer freemium caps margins.

Implication for market entry: Hyperscalers dominate via enterprise bundling (e.g., Gemini seats >8M); consumer plays face churn—focus B2B integrations where data moats yield sticky revenue.

Copilot Adoption Lags Amid Low Conversion

Microsoft's Copilot hit 15-16M paid M365 seats (Q2 FY26, +160% YoY) across 450M commercial base—mere 3.3-3.9% penetration—despite "multiples more" free Chat users; workplace conversion ~36%, with 44% lapsed citing distrust/accuracy (NPS -19.8 Jan 2026). No direct renewal/churn data, but low attach signals hesitation: Recon Analytics shows 8% prefer Copilot vs ChatGPT/Gemini when all available.[14][15]

• Azure AI: 39% growth Q2 FY26 (12pt AI contribution), cloud $51.5B +26%.[16]
• Bundling pressures: E3/E5 hikes July 2026 force "AI tax."[17]

Implication for rivals: Microsoft's moat is Graph data, but low usage exposes integration gaps—competitors like Gemini (8M+ seats) erode via superior accuracy/trust.

Hyperscalers Defy Softening with AI Acceleration

Cloud growth surges despite enterprise woes: AWS AI $15B run-rate Q1 2026 (+260x early pace), overall 24% Q4 2025; Azure 39% Q2 FY26; Google Cloud 48-50% Q4 2025/Q1 est., $17.7B Q4. Capex balloons ($600B+ 2026 across four), monetized as fast as built—no misses vs guidance, backlog surges.[18][19]

• Gemini: >8M enterprise seats (Q4 2025), 2,800+ firms; no challenges reported.[20]

Implication for challengers: Infra demand uncoupled from app-layer disillusion—bet on hyperscaler adjacencies, not standalone AI tools.

Trough of Disillusionment: Real but Sector-Skewed

Gartner's 2025 Hype Cycle places genAI in "Trough of Disillusionment" through 2026: post-peak, experiments fail, investments prune to survivors—yet global spend hits $2.52T (+44%). Evidence strongest in pilots/cancellations (40-60%), ROI surveys (28% success); isolated in infra (hyperscalers thrive). Widespread in non-tech (e.g., manufacturing 20% deploy-ready), narrower among cloud natives.[21][22]

For competition: Trough shakes weak hands—survive via ROI proofs (e.g., <3yr payback); hyperscalers' data moats widen gap, but app-layer niches open for specialists. Additional primary research on sector-specific churn would refine prevalence.

---

Recent Findings Supplement (April 2026)

Gartner Positions AI in "Trough of Disillusionment" for 2026

Gartner's Hype Cycle explicitly places generative AI in the Trough of Disillusionment throughout 2026, where early experiments fail to deliver promised ROI, forcing enterprises to prioritize predictable outcomes from incumbents over moonshot pilots; this explains why global AI spending surges to $2.5 trillion (up 44% YoY) despite stalled projects, as buyers demand enforceable baselines, targets, and accountability paths before scaling.[1][2]
- April 2026 Gartner survey (782 I&O leaders): Only 28% of AI use cases fully meet ROI; 20% fail outright; 57% report at least one failure due to misaligned expectations.[3]
- Oct 2025 Gartner (506 CIOs): 72% breaking even or losing money on AI; May 2025 update confirms trend.[4]
- S&P Global (early 2026): 42% of firms scrapped most AI initiatives in 2025 (up from 17% prior year); MIT (Aug 2025): 95% of GenAI pilots zero P&L impact.[5]
For competitors: This trough favors data-rich incumbents (e.g., Salesforce, ServiceNow) bundling AI; new entrants must prove integration/governance first or face 40% agentic AI cancellation by 2027.[6]

Forrester Echoes ROI Struggles Amid Siloed Adoption

Forrester's April 2026 report (1,500 AI decision-makers) reveals enterprises chase transformative value three years post-GenAI launch, but low AIQ (AI aptitude), productivity-only focus, and siloed tools block impact; mechanism: Employees lack fluency, metrics are weak, yielding fragmented pilots vs. enterprise-scale wins.[7]
- Few measure business outcomes; overemphasis on tactical use cases stalls compounding value.
- State of AI 2025 survey: 70%+ have AI in production, but minimal ROI tracking.
For competitors: Cross-functional AIQ training + outcome metrics (e.g., revenue lift) differentiate; avoid siloed pilots by centralizing governance.

OpenAI/ChatGPT Growth Slows, Missing Internal Targets

OpenAI missed 2025-end goals of 1B weekly ChatGPT users and full-year revenue, with Q1 2026 revenue also short amid Gemini/Anthropic gains; CFO Sarah Friar flagged compute commitments outpacing revenue, as token costs explode (e.g., Uber exhausted 2026 AI budget on Claude coding).[8][9]
- Users: ~900M weekly (Feb 2026, up from 400M prior year) but plateaued late 2025; market share fell from 87% (early 2025) to 64-68% (Jan 2026).[10]
- Revenue: $20B+ annualized (2025), but Q1 2026 miss; churn rising on subscriptions.
For competitors: Niche tools (e.g., coding agents) erode ChatGPT; hyperscalers win via bundled enterprise controls.

Microsoft Copilot: Low Active Usage Despite Seat Growth

Microsoft 365 Copilot hit 15M paid seats (Q2 FY2026, 3.3% of 450M M365 base, +160% YoY), but active usage lags at 35.8% conversion (Recon Analytics, 150K users); when competing with ChatGPT/Gemini, Copilot preference drops to 8%.[11]
- Median DAU: 30-38% at 90 days; top quartile hits 55-68%.
- No public renewal rates; low usage risks churn as ROI scrutiny rises.
For competitors: Bundled governance (e.g., Gemini Enterprise) boosts stickiness; focus change management over licenses.

Hyperscaler Cloud/AI Growth Moderates on Capacity Constraints

Azure growth slowed to 39% (Q2 FY2026, from 40% prior), guidance 37-38%; capex surges ($37.5B Q2) outpace revenue amid GPU shortages—45% of RPO from OpenAI.[12]
- AWS: 20% (Q4 2025, fastest in 13 quarters but lagging); Google Cloud: 26-35%+ expected.
- Combined hyperscaler capex: $250B+ (2026), but FCF pressure (e.g., Amazon -$17B).
For competitors: Edge in non-AI cloud or efficient inference (e.g., custom chips) exploits capacity gaps.

Trough Widespread but Not Fatal—Budgets Rise Amid Consolidation

Disillusionment is broad (42% scrapped pilots, 71% CIOs eye cuts sans ROI by H1 2026), yet 91% plan AI spend hikes (Deloitte); firms consolidate vendors, cut training (budgets +5% vs. AI +44%), fueling layoffs (27K AI-linked in 2026).[4][13]
- No Gemini-specific enterprise pauses; security flaws fixed, adoption accelerating (e.g., Deloitte 25K+ licenses).
- Strongest evidence: Gartner/Forrester surveys + project failures; isolated to pilots, not infrastructure.
For entrants: Target "value flywheel" (reinvest efficiency into growth); incumbents scale via enforcement.[14]