2026-08-04 — Opened: Oracle’s own arc, separated from six threads that only carry it as supporting cast ⟨crawl 2026-08-04⟩

• Oracle is a kingdom-rank board entry (~$64B commandable capital) that until now only appeared as an entities: tag on six other threads — ai-circular-financing-risk, stargate-buildout, ai-power-buildout, ai-datacenter-sites, nuclear-for-ai, datacenters-as-targets — each of which is really about a different actor’s story (Nvidia’s financing web, OpenAI’s Stargate buildout, the hyperscaler capex picture generally). This thread tracks Oracle’s own concentrated, debt-funded bet specifically — the same shape as softbank-all-in (SoftBank’s own concentration play) and closer in substance to coreweave-backlog-bet (a leveraged financing wager) than to arm-royalty-regime (a disputed-guidance story) — because the RPO figure itself is now well-sourced; the doubt is about the payer, not the receipt.
• Q4/FY2026 results (fiscal year ended 2026-05-31, reported 2026-06-10): total revenue $67.4B FY26 (+17%); cloud revenue $34.0B (+39%); OCI/cloud-infrastructure revenue $18.1B (+77%); RPO ended at $638B, up 363% YoY, up $85B sequentially from Q3 — confirmed directly from Oracle’s own SEC 8-K, which supersedes the board’s prior “$450B+ RPO, conf LOW, unverified” note with a larger, primary-sourced figure. Operating cash flow $32.0B (+54%); free cash flow -$23.7B; capex $55.7B; FY26 debt raised $43B, equity raised $5B; FY27 guided to raise **$40B more debt**. (Oracle Q4 FY26 press release, SEC EDGAR) ⟨crawl 2026-08-04⟩
• S&P downgraded Oracle one notch, BBB → BBB-, outlook stable, on 2026-07-09 — one step above junk — citing FY27 capex projected to jump to $90-95B, the FCF deficit widening toward -$42B, leverage rising to the mid-4x area, and explicitly that OpenAI accounts for roughly half of Oracle’s RPO. Moody’s separately carries a negative outlook (exact rating letter/date not independently confirmed this crawl); Moody’s had already named Oracle and CoreWeave jointly as “AI’s weakest credit link” (2026-07-25) and Oracle’s CDS hit a cycle high around 2026-07-20/21, spiking further into early August per later coverage. (S&P downgrade coverage, Yahoo Finance) ⚠ total-debt figure not reconciled: $156B (board) vs $167B vs a $130-156B range across sources — flagged, not resolved. ⟨crawl 2026-08-04⟩
• The concentration, not the RPO figure, is the live credibility question. Roughly half of Oracle’s RPO traces to OpenAI; one analyst-level breakdown puts it at ~$300B / 54.25% tied to the Oracle-OpenAI bilateral cloud contract specifically (a ~$300B, 5-year, 4.5GW deal announced July 2025 — distinct from Oracle’s smaller co-investment stake in the shared Stargate JV, where SoftBank holds financial responsibility and OpenAI holds operational responsibility). OpenAI reportedly missed its own internal user and sales targets for 2026 (reported ~2026-04-29) — flagged explicitly as thin/unconfirmed, since no clearly identified primary source was found for that claim, but it is the direct mechanism by which “can OpenAI pay what it owes Oracle” doubt entered the credit story. A separate, thinly-sourced “round-trip revenue” critique (2026-05-25) argues Oracle’s equity stakes in unprofitable labs (OpenAI, Anthropic) cycle back as its own cloud revenue — the same pattern ai-circular-financing-risk already tracks, noted here for completeness rather than duplicated. (RPO skepticism roundup, OpenAI-shortfall/concentration analysis) ⟨crawl 2026-08-04⟩
• Confirmed Stargate build sites (via OpenAI’s own announcements): Abilene TX (flagship, 1.2GW), Shackelford County TX, Doña Ana County NM, Lordstown OH, and a Wisconsin site — cumulatively approaching ~7GW planned capacity and $400B+ committed over three years, tracking ahead of pace toward the full $500B/10GW four-year target. Ellison, on the earnings call: “Oracle will build more cloud infrastructure data centers than all of our infrastructure competitors combined.” Most recently (2026-08-03): FY27 datacenter capex reported at ~$70B (vs S&P’s independent $90-95B projection), with Ellison framing Oracle’s own code increasingly as AI-model-written — a stance that sits awkwardly next to Oracle’s own ban on AI-generated code contributions to OpenJDK on safety/security grounds. (OpenAI’s Stargate site announcement, The Register, 2026-08-03) ⟨crawl 2026-08-04⟩
• Cerner/healthcare (MH-lens surface) stays a minor, separate story this quarter: Oracle Health launched an “AI-first” EHR built fresh on OCI (not atop legacy Cerner infrastructure), for ambulatory providers pending regulatory approval, with acute-care functionality and broader clinical AI agents planned for later in 2026. No material financial/credit event tied to Cerner found this window — it remains product-roadmap news, not a driver of the leverage story this thread tracks. ⟨crawl 2026-08-04⟩
• Next print: Q1 FY2027 earnings, 2026-09-14, after market close — the first test of whether the FY27 capex step-up and the reported OpenAI shortfall show up in updated RPO/FCF/capex numbers. ⟨crawl 2026-08-04⟩