Impact SuiteSundayPyjamas Impact Foundation
Browse

AI Economy Metrics

Hyperscaler capex, AI venture capital flows, market sizing, sovereign AI funds, productivity projections, and the adoption gap: the numeric spine of the AI-economy analysis. Every value cites its source.

Share
Vintage: 2026-10. Refreshed October 6, 2026 (daily briefing): added Anduril’s $4.7B defense-contract week ($1.8B Army NGC2 + $2.9B Navy submarine deal; source 67), the Pentagon’s 7-vendor deals to deploy frontier AI on classified networks (source 68), SpaceX’s $1.11B/month AI hosting deal (~$13B annualized; source 69), the $526M two-day AI-drug-discovery raise wave — Enveda’s $311M Series E at $2B and BigHat’s $75M Series C (source 70), ByteDance’s Anew Labs $290M spin-off raise at $1.5B (source 71), and the 2026 Crop Robotics Landscape (400+ companies, +25% since 2024; source 72); updated Harvey to $15.5B (Sep 9 round; source 73). Refreshed October 5, 2026 (daily briefing): added BCG’s Applied AI Index 2026 (corporate AI spending doubled to 3.3% of revenue, banks lead at 5.7%; sources 59–60), NVIDIA’s healthcare AI ROI survey (63% actively using AI, $3.20 per $1 in ~14 months; source 61), India’s sovereign GPU expansion (~30,000 of 45,000 committed GPUs accessible; source 62), the Granite Asia/PSA $50M ports-and-logistics AI fund (source 63), OneByZero’s $20M Series A for governed enterprise AI agents (source 64), and the Treasury Dragons deepfake-fraud index with the Fideuram €95M case (sources 65–66). Refreshed October 5, 2026 (manual ontology test run): added Instinct’s $1B Series C at a $10B valuation alongside EliseAI ($350M at $4B) and Armadin ($255.5M at >$2.5B) — 9 of the week’s 10 biggest US rounds were AI (source 57) — and GMI Cloud’s $445M debt + $223M equity raise, another datapoint in the neocloud debt-financing wave (source 58). Refreshed October 4, 2026: added Amazon’s ~15% EC2 chip-rental price hike alongside Jassy’s $220B 2026 capex plan and ‘striking’ 2028 demand (source 54), JPMorgan AM’s count of ~$250B in bond-market plus ~$250B in bank-loan funding for hyperscaler capex — global cash underwriting the buildout, with off-balance-sheet and concentration warnings (source 55) — and Goldman’s Q3 preview: AI infra companies driving >50% of S&P 500 EPS growth, market breadth at dot-com-bubble lows (source 56). Refreshed October 2, 2026: added Tencent's ~$7B five-year lease of ~100K AI chips from Oracle data centers in Southeast Asia (id 43), JERA/Dell/RHAELM's >$15B, 400 MW Chiba AI data center MoU with FT's broader $140B Japan estimate (id 44), the Brookings BPEA cost model — ~$8.2B per 200 MW campus, plus the off-balance-sheet financing warning (id 45) — and Anthropic's IPO-prospectus risk disclosures (id 46). Refreshed October 1, 2026: added Samsung's $1B commitment to the Nvidia/KKR-backed Helix Digital Infrastructure vehicle (total committed capital >$11B) (id 41) and Morgan Stanley's count of $286B in US data-center projects delayed or cancelled since 2025 (id 42). Refreshed September 30, 2026: added McKinsey's estimate that ~11M US workers (~7% of the labor force) may need to switch occupations by 2035 as AI reshapes work (sources 36-37), Workday's second 2026 layoff round — ~500 jobs (2.5%), mostly Product & Technology, disclosed in a Sep 29 Form 8-K (source 38), Nvidia's +$150B share-repurchase authorization (source 39), and the FT's report that Nvidia is in early talks with insurers on residual-value coverage for chip-backed loans to neoclouds (source 40). Refreshed September 28, 2026: added Goldman's $1.42T revenue-hurdle math for 2026–27 AI compute capex (ids 29–30), fresh SkillSyncer/Layoffs.fyi tech-layoff totals (id 31), and ByteDance's reported ~1M-unit RTX Pro 5500 order consideration under China's potential Nvidia chip approval (id 32). Refreshed September 27, 2026: added Moody's US hyperscaler capex forecast (>$700B in 2026, 6x 2022; $870B in 2027, electricity constraining capacity) (id 27) and Crunchbase's 2026 tech-layoff tracker (94,046 US cuts Jan-Aug, +16.8% YoY; AI cited in 33% of events vs 1% in 2024) (id 28). Refreshed September 26, 2026: added The Economist's net-jobs estimate (~1M US jobs created vs ~200K AI-attributed layoffs since mid-2023) (id 26). September 25 (midday): added SoftBank $11.1B record bond sale for AI financing (id 23), Basecamp Research $140M Series C (id 24), and Google Project Suncatcher orbital-AI test satellite (id 25). Morning: Brookings $10.3T AI-buildout projection (id 17), Goldman 1.8% GDP + hyperscaler capex/monetization math (ids 18-20), FactSet $4.2T debt-funded capex (id 21), Gartner 30% AI-layoff rehire (id 22). Generative-AI market estimates for 2026 range from $121B to $395B depending on methodology. The spread is itself a finding, not a resolvable disagreement. Refreshed September 29, 2026: added Goldman's $7.6T cumulative 2026–31 AI infra capex and the debt-financing picture (35% of 2027 hyperscaler capex, ~$500B issued by Aug 2026) (sources 33), AMD's $8.2B all-stock World Labs acquisition (source 35), and the New York Fed's AI-and-jobs survey — 4% of service firms laid anyone off because of AI, 13% hired to deploy it, a third of AI-using firms retrained workers (source 34).

87 of 87 rows

EntityMetricValuePeriodConfidenceSource
United States
Aggregate spending
Data-center + AI infrastructure spending (Brookings projection)New
Stijn van Nieuwerburgh (Columbia/Brookings), reported by WSJ Sep 23, 2026: ~3.6% of GDP per year — the largest single-industry buildout as a share of GDP in US history (railroads 2.24%, interstates 1.13%, electrification 0.5%). Projections could come in substantially lower.
~$10.3 trillion2025-2032Press-reported[17]
World
Aggregate spending
AI spending (Gartner forecast)
+47% YoY; potentially $3.3T in 2027.
~$2.59 trillion2026Official[1]
United States
Aggregate spending
AI investment share of GDP (Goldman Sachs)New
Goldman Sachs Research: US AI capex ~1.8% of GDP in 2026 (global 0.9%), rising to 2.5% in 2027 and 2.8% in 2028 — inside the 2-5% range of prior general-purpose-technology buildouts, not beyond it.
~1.8-1.9%2026Official[18]
United States
Aggregate spending
AI capex share of US GDP
~1-2.2%2025-2026Press-reported[2]
United States
Aggregate spending
AI capex vs. consumer spending contribution to H1 2025 GDP growth
Rare: consumption is ~68% of GDP and rarely dethroned as growth driver.
AI capex contributed more than consumer spendingH1 2025Press-reported[2]
World
Aggregate spending
Infrastructure (chips / servers / networking / cloud) share of AI spend
45%+2026Official[1]
United States
Aggregate spending
AI investment share of S&P 500 EPS growthNew
Goldman Sachs Research (Sep 17, 2026): AI capex is the largest of three temporary earnings drivers (with chip margins and equity gains). The boost fades as spending growth slows and depreciation charges rise. Q2 2026 S&P 500 EPS grew 51% YoY.
Almost half2026Official[15]
Largest US hyperscalers
Hyperscaler capex
2026 capex (Goldman Sachs estimate)New
+94% YoY. Goldman (Sep 17, 2026 research): $1.2T in 2027, $1.4T in 2028 — but depreciation drag is expected to fully offset the capex earnings tailwind by 2028.
~$800 billion2026Press-reported[19]
Hyperscalers
Hyperscaler capex
Annual AI revenue needed to justify capex (Goldman Sachs)New
Goldman Sachs: the monetization hurdle behind the buildout. The application layer may need ~$1T in annual revenue for both hyperscalers and software providers to earn satisfactory returns.
~$300B to break even; >$500B for 20% return2026Press-reported[20]
Alphabet + Amazon + Meta + Microsoft + Oracle
Hyperscaler capex
Combined capex through 2029 (FactSet)New
Rising share debt-funded, partly via off-balance-sheet SPVs with banks and private credit. Brookings estimates ~$3.7T in annual AI revenue by 2032 would be needed to generate sufficient returns on the buildout.
~$4.2 trillion2026-2029Press-reported[21]
SoftBank
Hyperscaler capex
Largest high-yield corporate bond sale on recordNew
Dollar- and euro-denominated, to fund SoftBank's $64.6B OpenAI commitment (~13% stake). LSEG: hyperscaler bond sales have more than doubled in 2026 to over $200B — the debt leg of the AI buildout is accelerating.
$11.1 billionSep 2026Press-reported[23]
US hyperscalers (Moody's Ratings)
Hyperscaler capex
US hyperscaler capex (Moody's forecast)New
6x 2022's capex. Moody's (Sep 2026): AI demand is ‘substantially exceeding supply’, but capex is eroding tech giants' historically strong free cash flow and prompting higher borrowing; bond spreads are widening and median equity prices declining. 2027's $870B will be capped by lack of readily available electricity — demand outruns capacity through 2027. A data center takes 12–24 months from initial spend to revenue; creditworthiness could be reassessed if profit growth fails to materialize.
>$700B in 2026; ~$870B in 20272026-2027Press-reported[27]
Amazon + Microsoft + Alphabet + Meta + Oracle
Hyperscaler capex
2026 combined capex (planned)
Up ~35%+ from 2025's ~$450-500B; exceeds Switzerland's GDP.
~$600-725 billion2026Press-reported[3] [4] [5]
Amazon
Hyperscaler capex
2026 capex guidance
~$200 billion2026Press-reported[3]
Microsoft
Hyperscaler capex
2026 capex guidance (CY)
~$190 billionCY2026Press-reported[3]
Alphabet
Hyperscaler capex
2026 capex guidance
$175-190 billion2026Press-reported[3]
Meta
Hyperscaler capex
2026 capex guidance
$115-145 billion2026Press-reported[3]
Global
Hyperscaler capex
Data-center buildout needs (projection)
McKinsey-class estimate.
~$6.7 trillion by 2030 (~70% AI-driven)By 2030Proxy / estimate[1]
World
Hyperscaler capex
Data center capex growth (YoY)New
Dell'Oro Group: surging AI demand plus rising memory/storage prices lifting server ASPs. Neocloud providers and AI model builders grew fastest. Dell'Oro's 2026 global data-center capex forecast exceeds $1 trillion.
+92%Q2 2026Official[16]
United States
Venture capital
AI share of US VC
~61% (~$259B of $427B)2025Press-reported[2]
World
Venture capital
AI share of global VC (Q1 2026)
80% (~$242B in the quarter)Q1 2026Press-reported[1]
OpenAI
Venture capital
Largest venture round ever
$40 billion2025Press-reported[1] [2]
World
Venture capital
Aggregate AI revenues vs. capex
The bubble/buildout question in one number.
~$50B revenues vs. capex 10×+ that2026Proxy / estimate[1] [2]
Generative AI
Market sizing
2026 market estimate (methodology spread)
Spread itself is the finding, no methodology consensus on what counts.
$121B - $395B2026Proxy / estimate[1] [6] [7] [8]
Abu Dhabi (MGX)
Sovereign AI
AI fund closed
$49 billionJuly 2026Press-reported[9]
United Kingdom
Sovereign AI
Sovereign AI Fund
£500 millionApril 2026Official[10]
UAE + Japan
Sovereign AI
Share of disclosed sovereign AI investment
> 2/32025-2026Press-reported[9]
World
Sovereign AI
New national AI strategies published in 2024
First low-income countries with formal plans include Ethiopia and Rwanda.
122024Press-reported[9]
Global organisations
Adoption & productivity
Use AI in ≥1 function / see enterprise-level EBIT impact
The central adoption-gap number of the AI-economy analysis.
88% / 39%2025-2026Press-reported[1] [11]
United States
Adoption & productivity
Projected generative-AI GDP effect (Penn Wharton)
+1.5% by 2035By 2035Official[11]
United States
Adoption & productivity
Projected generative-AI GDP effect (Penn Wharton)
~+3% by 2055By 2055Official[11]
Global
Adoption & productivity
Long-run GDP effect (Goldman-style range)
Range reflects diffusion-speed assumption disagreement, not measurement error.
+1.5% to +15%Long runProxy / estimate[2]
World
Adoption & productivity
AI-displaced employees needing rehire (Gartner)New
Gartner (Sep 9, 2026): employees laid off because of AI replacement will need to be rehired, often at significantly higher cost — ~40% for large enterprises. AI-productivity layoffs were <1% of 2025 layoffs and fell in H2 2025.
~30% by 2029By 2029Official[22]
World
Adoption & productivity
Roles created / displaced by 2030 (WEF)
Netting hides distributional pain, per framework, never report only the net.
+170M / -92M (net +78M)By 2030Press-reported[12]
United States
Adoption & productivity
AI-created jobs vs. AI-attributed layoffs (The Economist)New
The Economist (Sep 2026): AI has created roughly 1 million US jobs against ~200,000 AI-attributed layoffs since mid-2023. Engineers, software developers, mathematicians and data scientists added ~730,000 jobs above the professional-employment trend since 2022; data-center and power-infrastructure construction lifted trades hiring. Challenger, Gray & Christmas: 116,175 announced US cuts cited AI through Aug 2026 (~22% of all announced cuts). Customer-service roles fell ~10% and administrative-assistant roles ~15% since Jan 2023. The two headline measures aren't directly comparable — announced layoffs vs. trend-estimated creation — and aggregate studies find no economy-wide displacement yet. BLS August: +162K jobs, unemployment 4.1%.
~1M created vs. ~200K cut since mid-2023 (~5:1)Mid-2023–Sep 2026Press-reported[26]
US tech employers
Adoption & productivity
Tech layoffs, AI cited (Crunchbase / Layoffs.fyi)New
Crunchbase's tracker: AI cited in 33% of tech layoff events in 2026 vs 1% in 2024; Layoffs.fyi attributes 92,913 layoffs globally (72% of its 2026 total) to AI. Layoffs.fyi founder Roger Lee: ‘little evidence AI is actually replacing the work of the human employees let go’ — firms are funding AI capex and cutting elsewhere to lift productivity with smaller workforces. The cuts came in bursts: 31,513 in May (incl. Meta's 8,000-job reduction), easing to 2,347 in August.
94,046 Jan-Aug 2026 (+16.8% YoY)Jan-Aug 2026Press-reported[28]
Google
Sector / company signals
Project Suncatcher orbital-AI test satelliteNew
First milestone toward a planned 2027 follow-on: tests whether AI hardware survives launch stress, radiation, and temperature swings running Gemini workloads. Google's answer to SpaceX's orbital-AI push.
4 TPUs aboard a refrigerator-sized satelliteOct 1, 2026 (Transporter-18 rideshare)Press-reported[25]
Basecamp Research
Sector / company signals
Series C for AI-designed medicinesNew
London frontier-AI company training biological models on proprietary biological data — not public protein/genomic datasets — to design medicines. Capital is flowing to AI biology built on private data moats.
$140 millionSep 2026Press-reported[24]
Abridge
Sector / company signals
US health systems deployed
Strongest at-scale vertical AI deployment signal outside pure software dev.
150+2026Press-reported[13]
Abridge
Sector / company signals
Valuation
$5.3 billion2026Press-reported[13]
Harvey (legal AI)
Sector / company signals
ValuationNew
Raised $550M (co-led by Diffusion and Lightspeed; Sequoia, Kleiner Perkins, a16z, Coatue, Goldman Sachs Alternatives participated) at $15.5B — up 41% since March's $11B. Used by 80% of the 100 highest-grossing US law firms; acquired agent-security startup Guardrails AI (4th acquisition of 2026). Vertical AI in International Legal Services is repricing on real enterprise penetration, not pilots.
$15.5 billionSep 2026Press-reported[73]
Legora (legal AI)
Sector / company signals
Valuation
$5.55 billion2026Press-reported[13]
Cursor
Sector / company signals
ARR growth
Among the fastest software growth ever measured.
$0 → $2 billion in under 3 years2023-2026Press-reported[13]
Lovable
Sector / company signals
ARR growth
~$400M in 14 months2025-2026Press-reported[13]
US Healthcare
Sector / company signals
Projected annual savings from AI by 2030
Methodology-dependent; McKinsey-class estimate.
~$150 billion / yearBy 2030Proxy / estimate[11]
Global Banking
Sector / company signals
Projected sector AI savings
$447 billion+ by 2030By 2030Proxy / estimate[1]
Global Manufacturing
Sector / company signals
AI contribution to global GDP by 2035
$3.8 trillion (projection)By 2035Proxy / estimate[14]
Six US hyperscalers
Hyperscaler capex
Cumulative AI revenue needed to justify 2026–27 AI compute capex (Goldman Sachs)New
Goldman (Sep 2026 research): at a 15% annualized ROIC, Alphabet/Microsoft/Amazon/Meta/Oracle/SpaceX need ~$1.42T cumulative revenue — ~$11.6B of annual revenue per GW of compute — to earn back 2026–27 'Phase II' AI compute capex. ~$300B/year just to break even on average annual capex; ~$1T/year for satisfying returns across the stack. The offset case: the three big public clouds sat on ~$1.69T of contract backlog as of Q2 2026, and cloud revenues are running ~$70B above the pre-AI trend line — real, but far short of either threshold.
~$1.42 trillion (2028–2030)2028–2030Press-reported[29] [30]
US tech employers
Adoption & productivity
2026 tech layoff totals across trackers (SkillSyncer / Layoffs.fyi)New
SkillSyncer's tracker: 225,122 tech layoffs across 519 events as of Sep 28, 2026; Oracle's March 31 cut of 30,000 is the largest single entry. Layoffs.fyi counted 128,536 cuts across 299 companies by Sep 10 — already past all of 2025 (122,606) with nearly four months left. Trackers differ in scope, so totals vary — the direction is consistent: 2026 is running well ahead of 2025.
225,122 across 519 events (SkillSyncer, Sep 28)Jan–Sep 2026Press-reported[31]
China / Nvidia
Sector / company signals
ByteDance's reported ~1M-unit RTX Pro 5500 order considerationNew
China's MIIT reportedly asked ByteDance, Alibaba and others to report planned purchases of Nvidia's new Blackwell-generation RTX Pro 5500 (84 GB GDDR7, 21,760 CUDA cores, $6,000+) — outside the data-center accelerator category targeted by US export controls — and signaled intent to approve some orders. ByteDance is reportedly weighing ~1M units (~two quarters of planned China supply). Reported by The Information Sep 27; Reuters carried with 'could not independently verify.'
~1 million units (reported)2026Press-reported[32]
United States
Aggregate spending
Cumulative AI infrastructure capex, 2026–2031 (Goldman Sachs)New
Goldman (Sep 2026): annual AI capex from $765B in 2026 to $1.6T in 2031 across compute, data centers and power. For scale: hyperscaler capex alone runs ~$800B in 2026 → $1.2T in 2027 (+50%) → $1.4T in 2028. The spending figure and the ~268 GW grid supply-demand gap by 2030 are the same buildout measured two ways — money and megawatts.
~$7.6 trillion2026–2031Press-reported[33]
Global hyperscalers
Hyperscaler capex
AI-related debt issued and debt share of 2027 hyperscaler capexNew
Nearly $500B in AI-related debt issued globally as of Aug 5, 2026, with hyperscalers accounting for 40%; debt expected to finance 35% of hyperscaler capex in 2027. The buildout is increasingly borrowed — bond-market scrutiny (widening spreads, record CDS demand) is where the financing risk lands. UBP separately flagged the same squeeze: 2026 hyperscaler capex (~$820B) tops operating cash flow, with $5.6T of investment projected by 2030.
~$500B issued by Aug 2026; 35% of 2027 capex debt-financed2026–2027Press-reported[33]
US firms (NY Fed district)
Adoption & productivity
AI layoffs vs AI hiring vs worker retraining (New York Fed survey)New
Federal Reserve Bank of New York regional survey (published Sep 1, 2026): 61% of service firms / 51% of manufacturers use AI (up from 40%/26% a year earlier); only 4% of service firms laid anyone off because of AI in the past six months, and no manufacturers reported AI-driven layoffs across all of 2025–2026. Meanwhile 13% of service firms hired more people specifically to deploy and use AI, while 15% hired fewer because of it. More than one-third of AI-using service firms retrained workers rather than replacing them — the authors: 'Existing workers are much more likely to be retrained than replaced by AI.' Caveat: adoption is broad but shallow — median share of workers using AI is 17% at service firms, 7% at manufacturers.
4% cut jobs due to AI; 13% hired to deploy it; >1/3 retrainedAug 2026 surveyPress-reported[34]
AMD / World Labs
Sector / company signals
AMD's all-stock acquisition of World LabsNew
AMD announced Sep 28, 2026 (expected close by end of 2026, pending regulatory approval): buys Fei-Fei Li's spatial-intelligence startup — 'world models' that generate and reason about physical 3D environments, for robot training, factory simulation and scientific discovery (products: Marble, Atlas). Li becomes AMD EVP & Chief Scientist. The chip race is moving from language models to physical AI: understanding frontier workloads shapes the chip roadmap. AMD's latest quarter: record data-center revenue, sales +50% to $11.5B.
$8.2 billion2026Press-reported[35]
US labor force
Adoption & productivity
Workers who may need to change occupations due to AI (McKinsey)New
McKinsey report released Sep 29, 2026: automation could reduce labor demand by 36M jobs by 2035 while creating 40M — net positive, but ~11M workers (range 6–16M depending on adoption pace and aging) face outright occupation switches, roughly triple the current ~215K/year. Most at risk: office/admin support, retail/sales, transport/logistics (customer service reps, cashiers, warehouse workers). Low-wage workers ~8x more likely to face a switch than high-wage workers. The AI jobs story is retraining-plus-mobility, not just cuts: compare the NY Fed's finding that firms retrain far more than they replace (source 34).
~11M by 2035 (~7% of US labor force)2026–2035Press-reported[36] [37]
Workday
Sector / company signals
Second 2026 layoff round (Form 8-K, Sep 29)New
Disclosed in a Form 8-K on Sep 29, 2026: cuts concentrate in the Product and Technology organization, with $65–85M in restructuring charges (severance, benefits, office space). Second round this year after February's ~400 cuts. The stated reason: 'align team structures with strategic growth priorities' — the familiar software pattern of 2026: rebuild products around AI while shrinking headcount, and investors reward both (the S&P 500 software index rose 3.5% the same day).
~500 jobs (~2.5%), mostly Product & TechnologySep 2026Press-reported[38]
Nvidia
Sector / company signals
Share repurchase authorization increase (Sep 2026)New
Announced Sep 28, 2026: Nvidia's board added $150B to the buyback authorization, leaving ~$235B remaining through fiscal 2028 — management signaling confidence in future cash flow even as it explores residual-value support for neocloud financing (source 40). On the same beat: Nvidia launched its Open Agent Safety Platform, SpaceX is reportedly buying more GPUs, and Rosenblatt reiterated a $390 price target.
+$150B (remaining ~$235B through fiscal 2028)Sep 2026Press-reported[39]
Nvidia / neoclouds
Hyperscaler capex
Insurer talks on chip-backed AI loans (FT, Sep 2026)New
Financial Times (reported Sep 29): Nvidia has held early talks with insurers about absorbing part of the financing risk on chip-backed loans to smaller cloud companies ('neoclouds') — coverage that protects lenders if a borrower defaults and the chips pledged as collateral can't be resold for enough. Talks may not lead to deals. Context from Huang's Aug 11 blog: Nvidia may provide residual-value support for up to 25% of a financing opportunity, assessed project by project, as part of platforms meant to mobilize more than $500B of outside capital. The buildout's debt share keeps growing (source 33) — someone has to underwrite the risk, and the chipmaker is now volunteering.
Early talks; residual-value support up to 25%Sep 2026Press-reported[40]
Samsung / Helix Digital Infrastructure
Aggregate spending
Samsung affiliates' commitment to Nvidia/KKR-backed AI infra vehicleNew
Six Samsung affiliates (Samsung Electronics $500M) invested in Helix Digital Infrastructure — the KKR-created, Nvidia-backed vehicle building data centers plus power, transmission, and fiber for hyperscalers. Capital is stacking behind integrated 'power + land + compute' plays: the bottleneck is being treated as an infrastructure-financing problem, not just a chip problem.
$1B (total committed capital >$11B)Sep 2026Press-reported[41]
United States
Hyperscaler capex
US data-center projects delayed or cancelled (Morgan Stanley)New
Morgan Stanley research (Sep 28, 2026): ~$286B of US data-center projects delayed or cancelled since 2025, mostly over local opposition — alongside a 57 GW power shortfall 2026–28 and hyperscaler/neocloud capex growth decelerating 93% (2026) → 61% (2027) → 14% (2028) as the cycle shifts from hardware to software. The buildout isn't stopping; the friction is community consent and grid capacity.
$286B since 20252025–2026Press-reported[42]
Tencent / Oracle
Hyperscaler capex
Five-year AI-chip lease (FT via Reuters, Sep 30)New
Tencent's largest overseas lease: ~30% paid upfront, and the deal already weighed on Tencent's Q2 free cash flow. The chips never enter China — leasing compute abroad exploits the gap in US export controls, which govern physical transfer, not remote use. For Oracle, another multi-year pre-commitment at a fragile moment: negative $5.4B free cash flow last quarter, stock down ~50% from its highs, +$700M restructuring charges in September.
~$7B for ~100,000 advanced AI chips (Oracle data centers, SE Asia)Sep 2026Press-reported[43]
Japan (JERA / Dell / RHAELM)
Aggregate spending
Chiba hyperscale AI data center MoU (Oct 1)New
JERA, Dell, and RHAELM signed an MoU for a repeatable AI-data-center model, starting with a 400 MW hyperscale campus beside JERA's Chiba thermal power station — behind-the-meter power that skips the grid-connection queue entirely (phased ops from 2028, full capacity 2029). Multi-GW rollout planned for the 2030s with Apollo as financing partner. The $140B figure is FT's estimate for the broader buildout; JERA confirmed multi-GW ambitions but attached no number. Blackstone separately pledged $30B for Japan AI data centers in June.
>$15B, 400 MW (FT: broader Japan push up to $140B)Oct 2026Press-reported[44]
United States
Aggregate spending
AI campus build cost (Brookings BPEA conference draft)New
Van Nieuwerburgh (Columbia/Brookings, BPEA fall conference): a representative 200 MW AI training campus costs ~$8.2B — ~2/3 IT equipment, 1/3 real estate and power infrastructure. Central scenario: 182.8 GW online by 2032 (57 GW operating today, 509 GW in the project pipeline, 227 GW never built). Key warning: financing risk is migrating from transparent on-balance-sheet hyperscaler capex to opaque off-balance-sheet structures — JVs, private credit, securitization, SPVs, leases, loan guarantees — that 'may make correlated exposures hard to observe before a downturn.'
~$8.2B per 200 MW campus (~$41B per GW)2026Official[45]
Anthropic
Sector / company signals
IPO prospectus risk disclosures (Reuters, Oct 2)New
The prospectus warns that government attitudes toward the company and its technology could spill beyond direct dealings into commercial customer and partner relationships — while also warning that advanced AI could pose 'catastrophic or existential risks to humanity.' Government contracts are <1% of annual revenue. Filed days after CEO Amodei's call to slow development, reported rogue-agent incidents (misaligned agent activity notified to 100+ third parties), three researchers fired for mishandling information, and an FTC industrywide AI probe.
Potential ~$2T valuation; warns of 'catastrophic or existential risks'Oct 2026Press-reported[46]
United States
Adoption & productivity
US job cuts attributed to AI, YTD Sep 2026 (Challenger)New
Challenger, Gray & Christmas (Oct 1): for the first time, AI is the single leading stated reason for US layoffs — 120,136 announced cuts through nine months, ~21% of 573,195 total. The honest read: September itself was quiet (43,281 cuts, −18% MoM, the lowest September since 2022; AI ranked only fifth that month with 3,961, ~9%) — AI topped the table because every other reason fell away faster, which reads as slow substitution at the margin rather than a mass-redundancy wave. The hiring side is frozen: 90,787 announced hiring plans, the weakest September since 2011. Tech led industries with 165,925 cuts YTD (+54% YoY).
120,136 announced cuts — ~21% of all US cuts, #1 stated reason YTDJan–Sep 2026Press-reported[47] [48]
United Kingdom / global
Hyperscaler capex
AI-related debt issuance (Bank of England FPC, Sep 2026)New
The Bank of England's Financial Policy Committee (record of its Sep 25 meeting, via FT, Oct 1): global AI-related debt issuance reached ~$450B from January through early September 2026 — more than double the total for all of 2025 (Morgan Stanley estimate cited). JPMorgan estimates ~$4.1T of AI capex could be debt-financed between 2026 and 2030; AI hyperscalers accounted for 47% of sterling corporate bond issuance this year. The Committee warned that 'circular arrangements' and opaque financing (SPVs, guarantees) complicate risk assessment and could amplify losses if expectations disappoint — debt, not just equity, is now underwriting the AI boom.
~$450B YTD through early Sep — more than 2x all of 20252026 YTD (Sep)Official[49]
United States
Adoption & productivity
AI's effect on software-engineering productivity (NBER WP 35793)New
Blumenfeld / Hazell / Lian / Schaab (NBER Working Paper 35793, Sep 2026): financial markets price AI as adding the equivalent of a permanent 32.6% increase in software-engineering productivity — firms with larger software-engineering payroll shares gain more when AI news hits, and a model maps that cross-sectional relationship into productivity gains. Mapped to GDP: +3.6% baseline, +6.5% if higher software-engineering productivity also raises R&D productivity. By mid-2026, amid rapid progress in coding agents, the effect had more than doubled vs. end-2025. The other side of the layoffs ledger: the same tools displacing jobs are priced as the biggest productivity lever in a generation.
Equivalent of a permanent +32.6% productivity increase (Nov 2022–Dec 2025)Nov 2022–Dec 2025Official[50] [51]
Global
Aggregate spending
Annual AI revenue needed to justify data-center capex (Bain 2026)New
Bain's annual global technology report (via Bloomberg, Oct 3): the AI industry must earn $6T in annual revenue by 2031 to justify the data-center capital being deployed — existing consumer and enterprise AI services may supply at most $1.8T, leaving $4.2T of new revenue to be created by autonomous machines, robotics, drug discovery, mental health, and energy generation. Crawford: 'AI infrastructure is being built well ahead of the demand curve and funding it sustainably will require adding approximately 1% to the annual global GDP growth rate.' Also: data-center sizes and costs doubling every 12–16 months; $5–6.5T of data-center spend by 2030; $68B of US projects blocked or delayed in the June quarter on local opposition.
$6T/yr by 2031; existing services supply at most $1.8T — $4.2T gap2031Press-reported[52]
Amazon
Hyperscaler capex
$8B Nvidia chip leaseback vehicle (FT/Reuters, Oct 2)New
Amazon is in talks (FT via Reuters, Oct 2) to move ~$8B of Nvidia Grace Blackwell chips — thousands of units across 12+ US data centers in five states — into an outside-investor vehicle (debt plus up to 10% equity) and lease them back: ownership moves off Amazon's balance sheet while it keeps using the hardware. Context: Amazon's trailing-twelve-month free cash flow was −$7.6B with $96.3B of H1 2026 capex; Michael Burry argues the hyperscalers spread hardware costs over too many years. A live example of the off-balance-sheet migration Brookings warned about — the most expensive input in AI is becoming collateral, not just equipment.
~$8B of Grace Blackwell chips → investor SPV, leased backOct 2026 (exploratory)Press-reported[53]
Amazon
Hyperscaler capex
EC2 Capacity Blocks chip-rental price hike (MarketWatch, Oct 2)New
Amazon raised prices ~15% (effective next week) on EC2 Capacity Blocks — rented access to Nvidia chips from A100 to B300 — after an earlier July hike; A100 prices hadn’t moved since January. Jassy: even with $220B of 2026 capex (nearly double 2025), “we will still not have enough capacity to meet all the demand we have in 2026... this dynamic will also be true in 2027 too. In fact, the demand we already have for 2028 is striking.” AI compute is priced like a shortage good — rising rental prices are the market rationing GPUs. The hike also punctures the under-depreciation argument (Burry): if the chips were overvalued on the books, customers wouldn’t pay more to rent them.
~+15% on Nvidia A100→B300 rentals; $220B 2026 capex (~2x 2025)Oct 2026Press-reported[54]
Global / United States
Hyperscaler capex
Hyperscaler debt funding via global cash (JPMorgan AM, SuperReturn Singapore)New
Charles Wu (JPMorgan Asset Management, head of APAC alternatives institutional client strategy), at SuperReturn Singapore: pensions, sovereign wealth funds, insurers, and foreign banks (Asia, Gulf) are increasingly the capital behind North American AI compute — “compute training assets remain predominantly in North America... the capital facilitating these assets is increasingly becoming global.” Hyperscalers fund most capex from operating cash flow but are raising ~$250B from bonds and ~$250B from bank loans. Wu’s warnings: off-balance-sheet lease obligations and concentration risk — “many funding securities... seem to pivot around five key hyperscaler entities” as guarantor, tenant, funder, or customer. Pairs with the Bank of England’s $450B AI-debt warning: the debt side of the buildout is now a global-capital story.
~$250B bond markets + ~$250B bank loans2026Press-reported[55]
United States
Aggregate spending
S&P 500 Q3 EPS growth concentration (Goldman Sachs, Oct 2)New
Ben Snider (Goldman Sachs, Oct 2): consensus sees S&P 500 Q3 EPS +27% YoY — the strongest setup into an earnings season since 2021, though down from Q2’s 33% (ex-distortions). AI infrastructure companies alone are expected to drive more than half of the index’s EPS growth; IT + energy ~80%; the 10 biggest contributors 68%, with Micron and Nvidia over a third (Micron already reported: +1,003% YoY). The median S&P 500 company sees only +9% EPS growth — and market breadth is at its lowest since the dot-com bubble. The AI trade is carrying the index while the typical company slows: earnings concentration is the market’s version of the revenue-hurdle debate.
+27% YoY expected; AI infra cos >50% of index EPS growthQ3 2026Press-reported[56]
United States
Venture capital
Instinct $1B Series C at $10B valuation (Crunchbase, Oct 2026)New
Sequoia, Benchmark, and Coatue backed the year-old personal-AI-assistant startup Instinct at a $10B valuation; same week: EliseAI raised $350M at $4B (a16z, Bessemer) and Armadin raised $255.5M at >$2.5B (a16z, Accel). 9 of the week's 10 biggest US startup rounds were AI. The private-market funding engine is still running at full speed — the 'bits' side of the buildout keeps absorbing capital even as public-market breadth narrows to AI names.
$1B at $10B; EliseAI $350M at $4B; Armadin $255.5M at >$2.5BOct 2026Press-reported[57]
GMI Cloud
Hyperscaler capex
GMI Cloud $445M debt + $223M equity raiseNew
Neocloud financing keeps tilting toward debt alongside equity — another data point in the off-balance-sheet financing wave for AI infrastructure, alongside Amazon's $8B chip leaseback SPV (source 53) and JPMorgan AM's ~$250B+$250B global-capital count (source 55). The most expensive input in AI is increasingly being financed like real estate: borrowed against, leased back, and collateralized.
~$668M total ($445M debt, $223M equity)Oct 2026Press-reported[58]
Global (banks lead)
Adoption & productivity
Corporate AI spending and value capture (BCG Applied AI Index 2026)New
BCG surveyed 1,330 CxOs (published Sep 30, covered Oct 4–5): corporate AI spending doubled in a year from 1.7% to 3.3% of revenue — and more than 80% now sits outside the enterprise IT budget. The ‘future-built’ cohort grew from 5% to 7.5% and, with the ‘scaling’ group (41%), nearly half the market captures real value (2.3× shareholder return, 2.4× revenue growth, 2.8× EBITDA vs laggards). Banks put 5.7% of revenue into AI (tech 4.1%, telcos 3.7%) for a 1.2% revenue uplift and 1.6% cost cut. The paradox: telcos have the clearest AI strategy (53/100) yet the joint-lowest returns (0.6% uplift) — knowing the playbook isn’t scoring. By 2030, 42% of companies expect to give AI agents real decision authority, but only 5% have the controls to do it safely.
Banks: 5.7% of revenue on AI (highest of 10 sectors); ~half of companies now capture meaningful AI valueOct 2026Press-reported[59] [60]
Healthcare
Adoption & productivity
Healthcare AI ROI (NVIDIA State of AI in Healthcare and Life Sciences 2026)New
Healthcare is ahead of most industries (~50% AI adoption): 63% of healthcare and life-science organizations actively use AI, another 31% are piloting. Among health systems that can quantify returns, more than half report ≥2× ROI on deployed AI — averaging $3.20 per dollar invested in about 14 months. The second wave (ambient documentation, revenue-cycle optimization, predictive clinical risk, AI-accelerated drug discovery) is less flashy than the first wave’s diagnostic-imaging tools but far more deployable. Healthcare AI just flipped from capex bet to cost-reduction line item — this is the moment its economic impact starts landing in health-system margins and medical costs.
63% actively using AI (31% piloting); $3.20 return per $1 invested within ~14 monthsOct 2026Press-reported[61]
India
Sovereign AI
India AI Mission GPU pool: government-controlled expansionNew
The India AI Mission can currently access only ~30,000 GPUs against 45,000 committed — private providers can’t deliver pledged capacity amid rising equipment costs. The Centre is responding by building its own pool: 3,000 GPUs with C-DAC, plus an 8-exaflop AI supercomputer (~10,000 GPUs of compute) from G42 + Cerebras + C-DAC + MBZUAI, to be owned by the Indian government as a ‘UAE government gift’ (termsheet signed in the PM’s presence; final agreements being finalized). Governments are now directly buying the compute stack because the market can’t supply it — sovereign AI is a fiscal line item, not just a strategy document.
~30,000 of 45,000 committed GPUs accessible; +3,000 via C-DAC, ~10,000 via UAE ‘gift’ supercomputerOct 2026Press-reported[62]
Maritime & Logistics
Venture capital
G&P Strategic Innovation Fund (Granite Asia + PSA International)New
Announced Sep 30 (covered Oct 4): Granite Asia ($11B AUM) and port operator PSA International launched a $50M fund to co-invest with Granite Asia’s existing funds in tech companies for ports, logistics, and supply chains — explicitly targeting AI, robotics, and automation moving ‘beyond pilots into practical, scalable adoption.’ PSA brings a global operating footprint (45 countries) to test systems in working ports. Automation capital is following the same path as compute capital, just into cranes instead of data centers.
$50M for AI, robotics, and automation in ports, logistics, and global supply chainsOct 2026Press-reported[63]
Asia-Pacific / Global
Venture capital
OneByZero $20M Series A (Jungle Ventures, first outside funding)New
Singapore enterprise-AI-agents firm OneByZero (founded 2023) raised its first external funding: $20M led by Jungle Ventures (Oct 5). Its forward-deployed model puts its own engineers inside client companies across 9 markets (APAC + US; Japan next) to build and govern AI agents inside enterprise systems — mainly banks, telecom operators, and retailers. Revenue has more than doubled each year for three years; deployments automated >90% of some customer-facing interactions and sped data-modernization projects by 50%. The jobs displacement is showing up in revenue lines, not forecasts — and enterprises pay for governed agents with defined roles, human managers, and KPIs.
$20M; revenue >2×/yr for 3 years; >90% of some customer-facing interactions automatedOct 2026Press-reported[64]
Global FinTech
Sector / company signals
AI payment-fraud surge (Treasury Dragons/nsKnox 2026 index; Fideuram case)New
The 2026 Treasury Dragons/nsKnox Payment Fraud Index (104 corporate treasury professionals, released Oct 1): 76% experienced at least one payment fraud incident in the past 12 months (up from 73%), 44% were hit more than once, and 53% faced a confirmed or suspected deepfake-related attempt — yet only 44% have formal AI-fraud policies and just 18% continuously revalidate supplier bank details. The case study: Fideuram (Intesa Sanpaolo’s private bank) lost ~€95M ($108M) in February to fraudsters using a WhatsApp impersonation plus an AI-cloned voice of a lawyer; ~€53M was recovered via China/Portugal freezes, ~€36M remains missing and converted to crypto; Milan prosecutors are investigating. AI is cutting verification costs and raising attack sophistication at the same time — fraud defense is now a corporate-finance line item.
76% of treasury teams hit by fraud; 53% faced deepfake attempts; €95M Fideuram AI-voice theftOct 2026Press-reported[65] [66]
Anduril / Space-Tech & Defense
Sector / company signals
$4.7B defense-contract weekNew
Anduril won the Army's Next-Generation Command and Control contract ($162.8M base, up to $1.8B over five years, Lattice software as the data layer) and a $2.9B Navy submarine-production deal, with a planned $3.7B Anduril shipyard at Sparrows Point, Baltimore, operating from 2030 — right after DoD tapped Palmer Luckey and Elon Musk to pursue 'absolute technological dominance.' Defense AI is the fastest-converting demand segment in the buildout: these are contracted dollars for deployed software, not roadmap capex.
$1.8B Army NGC2 + $2.9B Navy submarine-production dealOct 2026Press-reported[67]
US Department of Defense
Sector / company signals
Seven-vendor deals to deploy frontier AI on classified networksNew
The Pentagon signed agreements (week of Sep 14–18) with seven firms to deploy frontier AI into classified operational networks — multi-vendor architecture for planning, target identification, and logistics. The widest official deployment of commercial AI into classified military domains to date. Pairs with the Anduril wins: the sovereign-demand floor under AI revenue is going up at the same time private monetization is being questioned.
Google, OpenAI, AWS, Microsoft, Nvidia, SpaceX, ReflectionSep 2026Press-reported[68]
SpaceX / SpaceXAI
Sector / company signals
$1.11B/month AI compute hosting dealNew
SpaceX CFO Bret Johnsen at Goldman Sachs Communacopia (Sep 10): a new AI hosting deal worth $1.11B/month (~$13B annualized) starts December 1, 2026, toward an implied $100B ARR ambition; terrestrial compute at ~2 GW by year-end, 5–10 GW target next year, orbital compute satellites from 2027. Follows the xAI–SpaceX $1.25T merger (Feb 2026). One more datapoint that AI revenue is increasingly sold as contracted compute, not software seats.
~$13B annualized from Dec 1, 2026; ~2 GW terrestrial compute by year-endSep 2026Press-reported[69]
Enveda Biosciences
Sector / company signals
Series E for clinical-stage AI natural-molecule drugsNew
Enveda's AI platform discovers drugs from natural molecules; valuation doubled in 12 months to $2B, and its candidates are now in clinical trials (skin, weight-loss maintenance, IBD). The largest of three AI-drug-discovery raises totaling $526M in two days (with Basecamp's $140M and BigHat's $75M). AI drug discovery has reached the clinical-readout phase as an investment category — capital is now priced against trial results, not model benchmarks.
$311M at $2B valuation (led by Catalio Capital)Sep 2026Press-reported[70]
BigHat Biosciences
Sector / company signals
Series C for AI protein designNew
BigHat's AI protein-design platform raised $75M; its antibody–drug conjugate BHB810 is already in the clinic. Same two-day wave as Enveda ($311M) and Basecamp ($140M): $526M of capital priced against clinical progress rather than AI hype.
$75M (DFJ Growth, Premji Invest)Sep 2026Press-reported[70]
Anew Labs (ByteDance spin-off)
Sector / company signals
Inaugural external round for AI drug discoveryNew
ByteDance spun off its AI drug-discovery unit as Shanghai-based Anew Labs and raised $290M (HSG, IDG Capital, Hillhouse, 5Y Capital co-leading) at a $1.5B valuation; ByteDance keeps 56%. Funds biomolecular structure prediction and molecular design. The fourth large AI-biology raise in two weeks — China is capitalizing the same clinical-AI-biology trade the US venture market just repriced.
$290M at $1.5B valuation; ByteDance retains 56%Sep 2026Press-reported[71]
Agriculture / High-Tech Agriculture
Sector / company signals
2026 Crop Robotics LandscapeNew
The Mixing Bowl's 2026 landscape maps 400+ crop-robotics companies (vs 330 in 2024, 240 in 2022); Europe holds half, California 40%+ of US companies. Companion datapoints: Orchard Robotics' $22M Series A (Quiet Capital, Shine Capital) for AI crop management; Greenfield Robotics opened a Reg A+ public offering for chemical-free weed-cutting robots. High-Tech Agriculture — one of the briefing's 22 sectors — is scaling as a real deployment market, not a pilot economy.
400+ companies worldwide (+25% since 2024; 240 in 2022)Sep 2026Press-reported[72]

Sources

  1. Worldwide AI spending to grow 47% in 2026 (Gartner)
  2. Why AI companies may invest more than $500B in 2026 (Goldman Sachs)
  3. AI Capex 2026: The $690B Infrastructure Sprint (Futurum Group)
  4. Hyperscaler capex >$600B in 2026, +36% YoY (IEEE ComSoc TechBlog)
  5. Financing the AI Supercycle (MUFG)
  6. Generative AI worldwide (Statista)
  7. Generative AI market report (Grand View Research)
  8. Generative AI market (Coherent Market Insights)
  9. Sovereign AI Index (CNAS)
  10. UK Sovereign AI Fund (Wikipedia)
  11. Projected impact of generative AI on future productivity growth (Penn Wharton Budget Model)
  12. AI and the global economy (WEF)
  13. AI Landscape, 7 stack layers (ValueAdd VC)
  14. AI in manufacturing to 2035 (Precedence Research)
  15. Can AI investment drive S&P 500 earnings even higher? (Goldman Sachs)
  16. Data Center Capex Grew 92 Percent in 2Q 2026, Driven by Surging AI Demand and Memory Costs (Dell'Oro Group (via PR Newswire))
  17. The AI Build-Out Is Becoming the Biggest Economic Bet in U.S. History (Brookings $10.3T, crowding-out, smelter) (Dow Jones Newswires / WSJ (Putzier & Lahart))
  18. Global AI Investment Is Forecast to Exceed $1 Trillion in 2026 (Goldman Sachs)
  19. Goldman Sachs Research Breakdown: Exceptional Earnings, But Not a Bubble; AI Tailwinds to Weaken Next Year (TechFlow Post)
  20. AI's $1.1 Trillion Capex Boom Now Has to Earn Its Keep (Goldman monetization math) (The Dark Side of the Boom)
  21. US AI infrastructure spend could hit $10tn by 2032 (FactSet $4.2T, debt financing) (Construction Briefing)
  22. Gartner Identifies 4 Shifts Shaping the Future of Work (30% AI-layoff rehire by 2029) (Gartner)
  23. SoftBank issues $11.1 billion in bonds in OpenAI financing push (Reuters (via WIXX))
  24. Venture capital startup funding roundup September 24, 2026 (Basecamp Research $140M Series C) (Tech Startups)
  25. Google to launch Suncatcher AI data center into space on October 1 (Tech Startups)
  26. It's Not All Doom and Gloom: AI Added 1 Million US Jobs, 5 for Every 1 Cut (The Economist estimate) (Cybernistas)
  27. US hyperscaler capex to top $700bn in 2026, investors fear overbuild and weak returns (Moody's) (Data Center Dynamics)
  28. Tech Layoffs Outpace 2025 As Big Companies Shift Spending To AI (Crunchbase)
  29. Goldman Sachs calculates the revenue hyperscalers need to justify $1.7T in AI capex (Ranzware)
  30. The AI industry has spent about $1.4 trillion and earned back roughly half (Startup Fortune)
  31. Tech Layoffs Top 225,000 in 2026 as Companies Blame the Same AI They're Funding (Startup Fortune)
  32. NVDA Stock In Focus: China Reportedly Weighs Letting ByteDance, Alibaba Buy Some Nvidia AI Chips (Stocktwits (via TradingView))
  33. U.S. Hyperscaler Capital Spending Seen Reaching $1.2 Trillion in 2027 (TokenPost)
  34. New York Fed survey points to retraining, not mass layoffs (HotTea (citing Federal Reserve Bank of New York))
  35. AMD will acquire Fei-Fei Li's World Labs for $8.2 billion (TechCrunch)
  36. McKinsey Says AI Could Force 11 Million U.S. Workers to Switch Jobs (Seoul Economic Daily (via Bloomberg))
  37. Job Displacement Prompted by AI Could Force 11M Workers Into New Roles (The HR Digest)
  38. Workday (WDAY) Stock Slips After Announcing 500 Job Cuts in Latest Restructuring Move (Parameter (citing Form 8-K))
  39. NVIDIA (NASDAQ:NVDA) to Buyback $150.00 billion in Outstanding Shares (American Banking News)
  40. Nvidia (NVDA) talks to insurers about chip-backed AI loans (AI Stock Wire (citing Financial Times))
  41. Samsung affiliates commit $1B to Nvidia/KKR-backed Helix Digital Infrastructure (Channel Insider)
  42. AI spending growth seen slowing sharply by 2028 (Morgan Stanley: $286B projects delayed, 57 GW shortfall) (The Hindu BusinessLine (ANI, citing Morgan Stanley))
  43. Oracle has reportedly signed a $7 billion deal with Tencent (100K chips, 5-year lease) (Morningstar (MarketWatch, citing Financial Times))
  44. Japan AI Data Center Push Could Reach $140B as Dell, JERA Start $15B Build (Channel Insider (citing Reuters / Financial Times))
  45. Financing the AI buildout (BPEA conference draft: $10.3T, $8.2B/200MW, off-balance-sheet risk) (Brookings Institution)
  46. Anthropic warns government attitudes may hurt customer ties, IPO prospectus shows (Inside Telecom (Reuters))
  47. Job Cuts Fall in September; Hiring Plans Up 3% Over 2025 On Weak Early Seasonal Hiring (Challenger, Gray & Christmas) (Challenger, Gray & Christmas)
  48. Tech layoffs: 165,925 jobs slashed in 2026 as September cuts rise (Fast Company)
  49. Bank of England warns of AI debt (FPC September 2026 record; $450B issuance) (Computer Weekly)
  50. The Macroeconomic Effect of AI: Sizing the Software Engineering Channel (NBER WP 35793) (NBER)
  51. AI could boost software engineer productivity by 32.6% (CIO)
  52. AI faces $6trn test to justify data centres, Bain says ($6T by 2031, $4.2T gap) (Moneyweb (Bloomberg))
  53. Amazon reportedly explores moving $8 billion of Nvidia chips into an investor vehicle and leasing them back (Ground Truth (FT via Reuters))
  54. Amazon is hiking chip-rental prices ~15% (EC2 Capacity Blocks); Jassy: $220B capex, demand outstrips supply through 2028 (Morningstar (MarketWatch))
  55. Global investments drive US AI expansion despite worries about hidden debt (JPMorgan AM’s Charles Wu, SuperReturn Singapore) (The Paper (Bloomberg))
  56. AI spending to drive another strong S&P 500 earnings season, Goldman Sachs says (Snider: +27% Q3 EPS, AI infra >50% of growth) (Seeking Alpha)
  57. Instinct raises $1B Series C at $10B valuation (Sequoia, Benchmark, Coatue; 9 of week’s 10 biggest US rounds AI) (Crunchbase)
  58. GMI Cloud raises $445M debt + $223M equity (neocloud debt-financing wave) (Traders Union)
  59. BCG Applied AI Index 2026: nearly half of companies now capture meaningful AI value (corporate AI spending doubled to 3.3% of revenue) (PR Newswire (Boston Consulting Group))
  60. Banks top AI spending as telcos struggle to show returns (BCG Applied AI Index 2026: banks 5.7% of revenue) (TechCentral)
  61. NVIDIA State of AI in Healthcare 2026: 63% actively using AI, $3.20 per $1 in ~14 months (BBN Times (NVIDIA survey))
  62. Centre plans 10,000-GPU AI compute expansion (C-DAC 3,000 GPUs; UAE 8-exaflop 'gift' supercomputer; ~30,000 of 45,000 GPUs accessible) (Communications Today)
  63. Granite Asia ($11B AUM) & PSA International launch $50M G&P Strategic Innovation Fund for ports, logistics, supply chains (CaproAsia)
  64. OneByZero raises $20M Series A (Jungle Ventures; revenue >2x/yr; >90% of some customer interactions automated) (TNGlobal)
  65. Treasury Dragons & nsKnox 2026 Payment Fraud Index: 76% hit by fraud, 53% faced deepfake attempts (GlobeNewswire (nsKnox))
  66. Fideuram AI-voice scam: ~€95M stolen via WhatsApp + cloned voice, ~€36M missing in crypto (Milan prosecutors investigating) (Cybernews (Reuters))
  67. Anduril wins $4.7B in defense contracts (Army NGC2 $1.8B + Navy submarine-production $2.9B; Oct 5–6) (Investor's Business Daily)
  68. Pentagon inks landmark deals with 7 tech firms (Google, OpenAI, AWS, Microsoft, Nvidia, SpaceX, Reflection) to deploy AI onto classified networks (week of Sep 14–18) (Defense Tech Digest (LinkedIn))
  69. SpaceX locks $1.11B/month AI hosting deal toward $100B ARR (Goldman Communacopia, Sep 10) (TeslaNorth)
  70. Three AI drug-discovery startups raise $526M in two days (Enveda $311M at $2B, Basecamp $140M, BigHat $75M; Sep 23–24) (BiopharmaTrend)
  71. ByteDance completes $290M fundraising for AI drug unit Anew Labs after spin-off ($1.5B valuation; Sep 16) (Reuters)
  72. 2026 Crop Robotics Landscape maps more than 400 companies worldwide (+25% since 2024; Sep 16) (Wine Industry Advisor (The Mixing Bowl))
  73. Legal AI startup Harvey reaches $15.5B valuation in $550M round (Sep 9; 80% of top-100 US firms) (Reuters)

How to cite this page

SundayPyjamas Impact Foundation. "AI Economy Metrics." Impact Suite. Accessed October 7, 2026. https://sundaypyjamas.org/tracker/ai-metrics.