--- type: Current-State Analysis title: The AI Demand Side (2023-2026) description: AI revenue is real and compounding at triple-digit rates, but several widely repeated adoption, utilization and revenue figures trace to no locatable primary source. resource: https://aibubblequestion.com/demand-question tags: [ai, revenue, adoption, inference, utilization, depreciation] sources: - id: geekwire-microsoft-ai-run-rate title: "GeekWire on Microsoft's AI run rate" resource: https://www.geekwire.com/2026/microsoft-tops-wall-street-expectations-reports-accelerating-azure-growth-and-37b-ai-run-rate/ author: "GeekWire" - id: zitron-openai-financials-jun2026 title: "Zitron, Exclusive: OpenAI Financials" resource: https://www.wheresyoured.at/exclusive-openai-financials/ author: "Ed Zitron" last_modified: 2026-06-16T00:00:00Z - id: bloomberg-openai-run-rate-aug2026 title: "Bloomberg on OpenAI's run rate topping $40bn" resource: https://www.bloomberg.com/news/articles/2026-08-13/openai-s-revenue-run-rate-tops-40-billion-ahead-of-ipo author: "Bloomberg" last_modified: 2026-08-13T00:00:00Z - id: techcrunch-anthropic-65bn title: "TechCrunch on Anthropic's $65bn run rate" resource: https://techcrunch.com/2026/08/17/anthropics-annualized-revenue-surges-to-65b/ author: "TechCrunch" last_modified: 2026-08-17T00:00:00Z - id: meta-8k-q1-2026-results title: "Meta 8-K, Q1 2026 results" resource: https://www.sec.gov/Archives/edgar/data/1326801/000162828026028364/meta-03312026xexhibit991.htm author: "Meta" last_modified: 2026-04-29T00:00:00Z - id: meta-8k-q2-2026-results title: "Meta 8-K, Q2 2026 results" resource: https://www.sec.gov/Archives/edgar/data/1326801/000162828026050596/meta-06302026xexhibit991.htm author: "Meta" last_modified: 2026-07-29T00:00:00Z - id: mckinsey-state-of-ai-2026 title: "McKinsey, The State of AI: Global Survey 2026" resource: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value author: "McKinsey & Company" last_modified: 2026-08-25T00:00:00Z - id: mochikabu-genai-survey-2026 title: "Mochikabu, Gen AI Survey 2026" resource: https://www.mochikabu.com/en/gen-ai-survey.html author: "MOCHIKABU Germany" - id: menlo-ventures-enterprise-genai-2025 title: "Menlo Ventures, State of Generative AI in the Enterprise 2025" resource: https://menlovc.com/perspective/2025-the-state-of-generative-ai-in-the-enterprise/ author: "Menlo Ventures" last_modified: 2025-12-09T00:00:00Z - id: stanford-hai-ai-index-2025 title: "Stanford HAI, AI Index Report 2025 (Ch. 1)" resource: https://hai.stanford.edu/assets/files/hai_ai-index-report-2025_chapter1_final.pdf author: "Stanford HAI" - id: cast-ai-kubernetes-2026 title: "Cast AI, 2026 State of Kubernetes Optimization Report" resource: https://cast.ai/press-release/2026-state-of-kubernetes-optimization-report/ author: "Cast AI" last_modified: 2026-08-07T00:00:00Z - id: msft-10k-fy2023 title: "Microsoft 10-K (FY2023)" resource: https://www.sec.gov/Archives/edgar/data/789019/000095017023035122/msft-20230630.htm author: "Microsoft" last_modified: 2023-07-27T00:00:00Z - id: googl-10k-fy2023 title: "Alphabet 10-K (FY2023)" resource: https://www.sec.gov/Archives/edgar/data/1652044/000165204424000022/goog-20231231.htm author: "Alphabet" last_modified: 2024-01-31T00:00:00Z - id: meta-10k-fy2025-demand title: "Meta 10-K (FY2025)" resource: https://www.sec.gov/Archives/edgar/data/1326801/000162828026003942/meta-20251231.htm author: "Meta" last_modified: 2026-01-29T00:00:00Z - id: amzn-10k-fy2025 title: "Amazon 10-K (FY2025)" resource: https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm author: "Amazon" last_modified: 2026-02-06T00:00:00Z - id: serverbuyback-gpu-resale-index title: "ServerBuyback GPU resale index" resource: https://serverbuyback.com/resources/gpu-resale-value-index/ author: "ServerBuyback" - id: thecube-depreciation-schedules title: "TheCUBE Research on depreciation schedules" resource: https://thecuberesearch.com/298-breaking-analysis-resetting-gpu-depreciation-why-ai-factories-bend-but-dont-break-useful-life-assumptions/ author: "theCUBE Research" generated: by: "claude/opus-5" at: 2026-08-30T00:00:00Z status: stable stale_after: 2026-11-30T00:00:00Z verified: - by: "claude/opus-5" at: 2026-08-30T00:00:00Z --- # Revenue (facts vs estimates) * Microsoft (disclosed): AI business passed a $37B annualized run rate, up 123% YoY, quarter ended March 2026 (vs $13B in January 2025). Management metric, not GAAP.[^geekwire-microsoft-ai-run-rate] * OpenAI (leaked, FT-corroborated): 2025 revenue $13.07B, costs ~$34B, operating loss $20.92B, net loss attributable $38.5B.[^zitron-openai-financials-jun2026] Run rate has since surpassed $40B annualized as of mid-August 2026, roughly doubling its late-2025 pace -- reported via an internal staff communication, not an audited or public figure, and superseding the ~$25B February 2026 snapshot.[^bloomberg-openai-run-rate-aug2026] * Anthropic (estimates, CONTESTED throughout): $316M ARR (Mar 2024) to $1.4B (Mar 2025) to $9B (end of 2025) to $47B (May 2026) to $65B (end of July 2026).[^techcrunch-anthropic-65bn] Every figure in this chain traces to third-party estimation or reporting -- Sacra's own model for the two oldest figures, Bloomberg's reporting of unspecified investor materials for the two newest -- never to an Anthropic statement, filing, or audited document; Anthropic declined to comment when last asked. * Meta (disclosed): Q1 2026 revenue $56.3B (+33% YoY);[^meta-8k-q1-2026-results] Q2 2026 revenue $60.8B (+28% YoY), advertising revenue $59.4B (+27% YoY).[^meta-8k-q2-2026-results] Management's figures for advertiser adoption (8M in Q1, 9M in Q2) and a >$75B Advantage+ run rate do not appear in either filed exhibit -- they are earnings-call disclosures, not filed figures. Q2 operating margin fell to 31% from 43% YoY (operating income -8% to $18.8B), partly on $2.40B of legal charges and $1.18B of severance -- a complication for a pure-upside "AI ad tools" framing. * No named, dated report with disclosed methodology was located behind a "~$140B global generative-AI revenue for 2026" figure; a check against several named forecasters (S&P Global Market Intelligence, ResearchAndMarkets, Exponential View) shows estimates in this space vary by more than 2x depending on scope and horizon, none matching $140B. The figure is dropped rather than restated without a source; see [infrastructure-overbuild](/mechanisms/infrastructure-overbuild.md) for the matching scope problem on the capex side of this comparison. # Adoption and economics * Four figures, four surveys, four scopes -- not one coherent 2026 enterprise picture. Enterprise gen-AI spending tripling from $11.5B (2024) to $37B (2025) is Menlo Ventures' own venture-capital estimate, not company-reported revenue.[^menlo-ventures-enterprise-genai-2025] The 24%-scaled-production/48%-in-pilot split is from Mochikabu's survey of DACH-region regulated enterprises, not a global sample.[^mochikabu-genai-survey-2026] 37% of organizations (not 39%) report at least some EBIT impact from AI use, per McKinsey's global 2026 survey (1,719 respondents, 97 nations, fielded 4 May-8 Jun 2026, published 25 Aug 2026) -- McKinsey itself calls this share "about the same" as the year before, so the 39%-to-37% shift looks like survey noise rather than a real move, but 37% is the current published figure.[^mckinsey-state-of-ai-2026] The ~44% year-one-ROI figure has no locatable primary report behind its usual "McKinsey & Forrester, 2025" attribution and is dropped. * Inference cost for fixed capability falls roughly 10x per year (GPT-3.5-class: $20 to $0.07 per million tokens, late 2022 to late 2024).[^stanford-hai-ai-index-2025] * Utilization outside hyperscalers is measured only narrowly: Cast AI's own sample of roughly 23,000 Kubernetes clusters it monitors found average GPU utilization near 5% -- its own customer base, not a fleet-wide or hyperscaler-wide measurement.[^cast-ai-kubernetes-2026] A widely repeated "86% of enterprises at or below 50% utilization" figure has no locatable source of record across two independent search passes and is dropped. No audited hyperscaler utilization figures exist. # GPU depreciation controversy * Filings, exact and by company (not a uniform "3 years to 5-6" move): Microsoft raised server and network equipment life from 4 to 6 years, effective FY2023;[^msft-10k-fy2023] Alphabet raised servers from 4 to 6 years and certain network equipment from 5 to 6, also FY2023;[^googl-10k-fy2023] Meta set most servers and network assets to 5.5 years, effective January 2025;[^meta-10k-fy2025-demand] Amazon shortened a subset of servers and networking equipment from 6 to 5 years in 2025, citing the pace of AI-driven hardware change -- the opposite direction from the others, and starting from 6 years, not 3.[^amzn-10k-fy2025] * Estimated aggregate effect: a derived analyst model (TheCUBE Research, echoing figures associated with Michael Burry) puts the impact near $18B less depreciation in 2024 and a cumulative $176B by 2028, flattering earnings 21-27% versus unchanged assumptions -- but the companies' own disclosed effects, in the years they actually reported them, do not sum to anything resembling that: Microsoft +$3.7B operating income (FY2023), Alphabet -$3.9B depreciation/+$3.0B net income (FY2023), Meta -$2.92B depreciation/+$2.59B net income (FY2025), and Amazon's FY2025 move running the other way (+$1.4B more depreciation, -$1.0B less net income). Read as a model, not as a sum of the filings.[^thecube-depreciation-schedules] * Critique (Burry): accounting life exceeds economic life; physical use is not economic value. Defense (Nvidia, analysts): changes predate the boom, were audited, and A100s shipped six years ago still run at full utilization. Amazon's 2025 reversal is itself a live, filed counterexample to a uniform "hyperscalers always extend life to flatter earnings" narrative. * Market data: ServerBuyback's dealer-quoted pricing shows used H100s retaining roughly 75-85% of value after 24 months as of mid-2026, with prior-generation prices down 10-20% against Blackwell availability -- dealer-quoted estimates, not transaction-level or audited data, and not independently reconciled with the accounting-life figures above.[^serverbuyback-gpu-resale-index] [^geekwire-microsoft-ai-run-rate]: GeekWire on Microsoft's AI run rate [^zitron-openai-financials-jun2026]: Zitron, Exclusive: OpenAI Financials [^bloomberg-openai-run-rate-aug2026]: Bloomberg on OpenAI's run rate topping $40bn [^techcrunch-anthropic-65bn]: TechCrunch on Anthropic's $65bn run rate [^meta-8k-q1-2026-results]: Meta 8-K, Q1 2026 results [^meta-8k-q2-2026-results]: Meta 8-K, Q2 2026 results [^mckinsey-state-of-ai-2026]: McKinsey, The State of AI: Global Survey 2026 [^mochikabu-genai-survey-2026]: Mochikabu, Gen AI Survey 2026 [^menlo-ventures-enterprise-genai-2025]: Menlo Ventures, State of Generative AI in the Enterprise 2025 [^stanford-hai-ai-index-2025]: Stanford HAI, AI Index Report 2025 (Ch. 1) [^cast-ai-kubernetes-2026]: Cast AI, 2026 State of Kubernetes Optimization Report [^msft-10k-fy2023]: Microsoft 10-K (FY2023) [^googl-10k-fy2023]: Alphabet 10-K (FY2023) [^meta-10k-fy2025-demand]: Meta 10-K (FY2025) [^amzn-10k-fy2025]: Amazon 10-K (FY2025) [^thecube-depreciation-schedules]: TheCUBE Research on depreciation schedules [^serverbuyback-gpu-resale-index]: ServerBuyback GPU resale index