2026-08-26 — Nvidia’s Q2 print, the “first read” this thread flagged on 08-22, comes in: capex guidance goes up, not down

• Nvidia reported Q2 FY2027 results after market close on 08-26 (quarter ended 07-26): revenue $96.2B (+18% q/q, +106% y/y), Data Center revenue $89.0B (+117% y/y). CFO Colette Kress told analysts Nvidia’s own order backlog is now “more than $2 trillion,” and put fresh numbers on the aggregate this thread tracks: top-five hyperscaler capex ~$800B in 2026, rising to ~$1.3 trillion in 2027 — a 62%+ step-up, not the plateau some had priced in. CEO Jensen Huang separately forecast 70% revenue growth for Nvidia’s own fiscal 2028. This directly answers the 08-22 entry’s open question — whether Nvidia’s 15%+ price hikes to server makers force hyperscalers to buy less compute for the same money or spend more — and the market’s answer, via a raised aggregate forecast, is “spend more,” at least by Nvidia’s own account of its customers’ plans. (NVIDIA — primary, CNBC)

2026-08-25 — OpenAI’s data-center chief departs mid-buildout

• Chris Malone, who ran the data-center buildout underpinning OpenAI’s roughly $600bn compute-spend target through 2030, left the company. His responsibilities were split among three named leaders reporting to VP Sachin Katti; reporting counts him as the thirteenth senior departure of 2026, ahead of a planned 2027 listing. The buildout is the company’s largest financial commitment and its coordinator changed mid-flight. (CNBC)

2026-08-22 — Every 2026 capex plan was sized against a cost base that just moved

• Nvidia told Microsoft’s, Google’s and Oracle’s server manufacturers that prices rise more than 15% on early-2027 systems. The combined 2026 plan this thread tracks — roughly $610-650bn across Google, Microsoft, Amazon and Meta — was budgeted at one cost per unit of compute. A 15%-plus rise in the dominant line item forces one of two answers: the same money buys materially less compute, or the plans go up again. Both are consequential and they point in opposite directions for the “capex is peaking” thesis. Nvidia’s Q2 print on 08-26 is the first read on which way it breaks. (Bloomberg, CNBC)

2026-08-18 — Dell’Oro projects worldwide data-center capex past $3 trillion by 2030

• Dell’Oro Group’s own 2030 worldwide data-center capex outlook “has nearly doubled since the January 2026 forecast,” driven by higher hyperscale capex guidance and a newly modeled “AI-specialized cloud” segment (AI model builders + neoclouds) growing at ~60% CAGR. ⚠️ This is a market-research firm’s forward-looking estimate, not a hyperscaler’s own disclosed guidance — treat as an analyst forecast, not a company commitment. (Dell’Oro Group via PR Newswire)

2026-08-03 — BofA puts a number on the next leg up: $1.2T over the next 12 months (caught 08-05)

• Bank of America now sees hyperscaler capex reaching $1.2 trillion over the next 12 months — a step up from the ~$700B-for-calendar-2026 framing this thread has tracked, and a forward 12-month window rather than a single fiscal year, so the two figures aren’t directly comparable but both are climbing. BofA analyst Vivek Arya named nine semiconductor beneficiaries with 30%+ upside (Marvell, Micron, Intel, Nvidia, Credo, KLA, Applied Materials, Broadcom, AMD) across compute, memory, semicaps, power semis and optics. (Benzinga)

2026-08-03 — Oracle’s stock jumps 9.2% while its credit still prices the same capex as distress

• Oracle rose 9.2% as equity and credit split on the same capex — ORCL closed $141.85 (+9.22%) on the risk-on tape plus an “expanded” Google Cloud AI partnership, but this does not reverse its credit story: S&P downgraded it to BBB-/A-3 on 07-09 over a projected $42B FY2027 FCF deficit, and its CDS hit a record 198bp on 07-17. (Motley Fool / stockanalysis.com; S&P/Forbes for credit, 08-03)

2026-08-01 — Cumulative 2026 capex guidance reaches $720-745B as the chip complex sells off

• The four hyperscalers’ cumulative 2026 capex guidance now totals $720-745B, which post-earnings coverage frames as reassuring to investors who had feared a slowdown. (CNBC)
• Amazon raised its own FY2026 cash capex guidance to ~$220B from ~$200B on the 07-30 call, citing higher memory costs. (CNBC)
• That guidance sits against a semiconductor index that fell 22.1% in the same month — demand guidance and chip equities pointing in opposite directions, which is this thread’s central tension stated numerically.
• TSMC’s CEO reiterated robust AI demand through 2029-2030, consistent with the guidance rather than the equity move.

2026-07-30 (later) — Three of four now in: Amazon’s the cleanest beat of the week

• Amazon reported: AWS +37% YoY (fastest in 18 quarters), net sales +20%, property/equipment purchases up $66.1B — completes the three-of-four-hyperscalers-report picture for this week (Meta, Microsoft, Amazon; Alphabet’s own capex context still tracked via Google Capex). Full detail: AWS Capex.

2026-07-30 — Two of four report: capex still climbing, the market draws the monetization line

• Meta and Microsoft both reported 07-29: neither shows a capex peak — Meta raised its 2026 floor to $130-145B (second raise this year); Microsoft’s FY27 guide stayed “grow year-over-year” plus $50B+/quarter next quarter. Amazon reports later today (07-30) — no result in hand as of this run. (Microsoft IR)
• The market drew a monetization line, not hyperscaler-vs-chipmaker — Microsoft (Azure +43%) bought tolerance for its capex; Meta (EPS miss, near-zero FCF) did not, despite raising its own guide. Full cross-name read in The Great Rotation.

2026-07-28 — Eve of Meta/Microsoft earnings: financing keeps migrating off-balance-sheet, Nvidia becomes financier-of-last-resort

• Two same-day instances of the off-balance-sheet pattern: Meta formalized a second 80/20 asset-sale JV (BlackRock, $14B, El Paso — see Meta Capex), and Nvidia emerged as a prospective ~$250-500B financing backstop for OpenAI’s 10GW Ohio site (see Stargate) — the chip vendor increasingly underwriting its own customers’ buildouts.
• Meta reports tomorrow (07-29), Microsoft the same day, Amazon 07-30 — consensus has coalesced around ~$136-138B FY26 Meta capex (mid-guidance, up ~$67B YoY) with 2027 seen rising further; this week’s commentary (CNBC, Seeking Alpha, Yahoo/Motley Fool) frames Meta/Microsoft/Amazon as the next test after Alphabet’s guidance raise, with combined 2026 AI spend across Alphabet/Amazon/Meta alone put over $500B before Microsoft is even added.

2026-07-27 — THE CAPEX PICTURE: four actors, one table

Four parallel destination crawls (findings + 167 sources in artifacts/bundles/-capex-2026-07-27/). The comparable read:*

The four cross-cutting findings:

1. The silicon split is the industry’s common non-disclosure — 4 of 4 refuse the own-chip vs Nvidia breakdown. The labs leak what the hyperscalers won’t (Anthropic: 1M TPUs and 1M Trainium).
2. The payoff layer is the weakest — backlogs and run-rates are press-corroborated; primary filings were unreachable or the mappings (e.g. $250B OpenAI → capex) simply don’t exist publicly.
3. Power is the binding constraint being solved three ways — nuclear restarts (‘27) → Gen-IV/SMR options (‘30–‘39) → and now Meta’s open gas pivot, the first break in the clean-power consensus.
4. Capex is migrating off balance sheets — Meta→Blue Owl 80%, BlackRock’s $12.5B bonds, Microsoft’s lease book: the buildout is becoming other people’s balance sheets. (Feeds Circular Financing / Nvidia as Lender.)

2026-07-23 — Meta-thread opened

• Opened as a meta-thread grouping four sub-threads — google-capex, meta-capex, aws-capex, microsoft-capex — after Ben asked for the individual hyperscalers as their own threads plus “the big picture” as an overlapping parent (/steer, “figure out how to do meta-threads”). A meta-thread’s card shows its children as linked rows, not a flat re-listing of their items; its ranking score aggregates their activity.

← Backstory

2026-07-22 — The first capex verdict of the cycle

• Alphabet raised 2026 capex guidance to $195-205B (from $180-190B), the first of the four hyperscalers to report this cycle — GOOGL fell ~5% after-hours despite a revenue/Cloud beat, reading as “spend still accelerating, not yet proof it pays off.” Full detail in Google Capex. (CNBC)

2026-07-20 — The question gets a date

• Alphabet/Tesla/Intel earnings framed as the first real test of whether AI revenue justifies Big Tech’s collective $725B 2026 capex plan. (Tech Times)