Overview
Oracle is the enterprise-software company turned AI-infrastructure builder that became a central financier and operator of the Stargate AI data-center program with OpenAI and SoftBank. In the wiki it is the “Stargate financing partner” and “Project Jupiter anchor customer” — the cloud/data-center node downstream of Nvidia chip demand and upstream of the AI Buildout Grid Constraint.
Key Facts
- Stargate scale: the $500B, 10-GW-by-2029 AI-infrastructure program (OpenAI + SoftBank + Oracle); as of February 2026 ~7 GW planned and $400B+ committed, with ~3 GW added in the prior 90 days. (OpenAI, Data Center Frontier)
- Abilene, TX campus: Oracle deploys 450,000+ Nvidia GB200 GPUs under a 15-year lease; the full campus draws 1.2 GW (roughly one million homes), with the remaining buildings due mid-2026. (Data Center Frontier)
- OCI growth: CEO Safra Catz projects Oracle Cloud Infrastructure revenue +77% to $18B in FY2026, reaching $144B by 2030 — much of it the OpenAI/Stargate compute backlog (a reported ~$300B, 5-year deal).
Newsletter Relevance
Oracle is the wiki’s clearest case of demand becoming a grid problem: a single company’s 15-year GPU lease commits 1.2 GW at one Texas campus, and the Stargate program’s 10-GW target is a direct input to the AI Buildout Grid Constraint and the interconnection-queue / transformer-lead-time bottlenecks the newsletter covers (the 12-gigawatts buildout thesis, PJM Interconnection). It also embodies the financing risk the wiki flags — hundreds of billions committed against a multi-year buildout whose power and timeline may not be priced in.
Connections
- OpenAI — the Stargate anchor customer / compute counterparty
- Nvidia — the GB200 GPUs Oracle deploys at scale
- Stargate — the $500B / 10-GW program Oracle co-builds (page not yet created — flagged for follow-up)
- AI Buildout Grid Constraint / PJM Interconnection — the power/grid bottleneck its campuses load
- AI DRAM Crisis — the memory demand its GPU deployments imply
Source Appearances
- Referenced in the wiki’s AI-buildout cluster as the Stargate financing partner and Project Jupiter anchor customer
Open Questions
- Does Oracle’s 1.2-GW Abilene draw (and the 10-GW Stargate target) actually clear interconnection and transformer lead times, or does it hit the AI Buildout Grid Constraint wall?
- Is the ~$300B OpenAI compute backlog a binding obligation or a soft commitment — the same LOI-vs-purchase-order ambiguity the AI DRAM Crisis turns on?
Web Sources (researched 2026-05-31)
- OpenAI (five new Stargate sites); Data Center Frontier (Oracle/OpenAI $300B deal, Abilene campus); IntuitionLabs (Stargate analysis)