2026-07-23 — Thread opened; first signal is a headcount cut, not capex
- Opened alongside its three siblings (google-capex, meta-capex,
microsoft-capex) so the term list catches AWS capex news going forward
— no dedicated item swept yet, stated plainly rather than backfilled
with a stretch. ⟨steer 2026-07-23⟩
- Amazon cut AGI research roles (model-customization/post-training)
while the $200B 2026 capex plan holds — a pivot toward enterprise
deployment over frontier research, not a capex-figure change; first
real test is the ~07-31 earnings call.
(Techtimes) ⟨daily 2026-07-23⟩
← Backstory
(Finding: artifacts/findings/aws-capex-2026-07-27.md · Bundle:
artifacts/bundles/aws-capex-2026-07-27/provenance.yaml)
- The capex buys landlord capacity, not a frontier-research bid —
the AGI-team cuts and the $200B plan read as coherent, not
contradictory: Anthropic committed over $100B to AWS over the next
decade (2026-04-20) and OpenAI signed a separate $100B/8-year AWS
compute deal (2026-02-27, alongside a $50B Amazon-into-OpenAI
investment). AWS’s Q1 2026 backlog hit $364B, up from $244B, excluding
the Anthropic deal.
(Anthropic)
⟨crawl 2026-07-27⟩
- Project Rainier (New Carlisle, Indiana) came online 2025-10-29 —
~500,000 Trainium2 chips, built for Anthropic, one of the world’s
largest AI compute clusters, initial $11B investment; a further $15B
followed for more Indiana capacity (2025-11-24). Mississippi’s
commitment was raised to $25B (2026-04-09); Pennsylvania ($20B,
2025-06-09) and a new ~$10B North Carolina site (2026-07-20) round out
the confirmed US build-out.
(Google News RSS convergence — About Amazon/DCD/DCK/Semafor)
⟨crawl 2026-07-27⟩
- AWS-Anthropic relationship grew, not shrank, despite Anthropic’s
Google TPU deal — ten days after Anthropic expanded with Google and
Broadcom for TPU capacity (2026-04-06/07), Anthropic and Amazon
announced an expanded deal: Anthropic already on 1M+ Trainium2 chips,
nearly 1GW combined Trainium2/3 capacity by end of 2026 (headroom to
5GW), Amazon investing up to $25B in Anthropic. Reads as multi-cloud
diversification, not AWS displacement — no numeric compute split
across AWS/Google/Microsoft exists in any source found.
(TechCrunch)
⟨crawl 2026-07-27⟩
- Power: 1.92GW under firm nuclear PPA now, 5GW+ SMR pipeline by
2039 — the AWS-Talen 17-year PPA (1,920 MW from Susquehanna, through
2042, ~$18B) was restructured in 2025-06 specifically to avoid the
FERC approval that killed the original 2024 co-located deal. X-Energy
(Amazon-backed, IPO’d 2026-04-23 raising ~$1.02B) joined the DOE’s
Project Prometheus alongside NVIDIA and AWS on 2026-07-22, reaffirming
the 5GW-by-2039 nuclear option.
(X-Energy IR) ⟨crawl 2026-07-27⟩
- The OpenAI ~$50B deal is CONFIRMED, not a rumor — officially
announced 2026-02-27: $50B total Amazon-into-OpenAI investment plus an
expanded $100B/8-year AWS compute deal, ~2GW Trainium capacity, AWS as
exclusive third-party cloud distributor for OpenAI’s “Frontier”
platform. Triggered a Microsoft exclusivity dispute that resolved
2026-04-27 (Microsoft’s exclusivity ended; OpenAI now multi-clouds
across AWS and Azure; hard 2032 Microsoft-OpenAI partnership end date).
(TechCrunch) ⟨crawl 2026-07-27⟩
- Earnings-date correction: Amazon’s Q2 2026 call is confirmed for
2026-07-30, not “~07-31” as this thread’s watch line has said since
opening — worth fixing on next
/daily//steer touch. Watch for: a
raised $200B capex guide (Alphabet just raised its own 2026 guide
07-22/23), backlog growth past $364B, FCF commentary (Amazon reportedly
borrowing an extra $25B as FCF turns negative under the capex load,
per TIKR/Motley Fool 07-08/09), and any analyst question on the AGI
cuts vs. capex tension. ⟨crawl 2026-07-27⟩ - Open gap: the brief’s “$151B TTM” capex figure was not
independently confirmed this crawl — every direct-fetch attempt
against Amazon’s own IR/blog domains 404’d. Verify at 07-30 earnings.
⟨crawl 2026-07-27⟩
