Executive summary
- NVIDIA partnership: OpenAI and NVIDIA announced a plan to deploy at least 10 gigawatts (GW) of NVIDIA systems for OpenAI’s next‑gen AI infrastructure. NVIDIA “intends to invest up to $100 billion” in OpenAI, released progressively as each gigawatt comes online; first 1 GW is slated for 2H 2026 on NVIDIA’s Vera Rubin platform. OpenAI+1
- AMD deal (and equity warrant): OpenAI signed a multi‑year chip‑supply agreement with AMD and received a warrant to buy up to 160 million AMD shares (≈10%) at $0.01/share, vesting in tranches tied to milestones (including next‑gen MI‑series deliveries). The buildout aims at ~6 GW beginning in late 2026. Reuters+1
- Oracle collaboration: OpenAI selected Oracle Cloud Infrastructure (OCI) in June 2024 to extend Microsoft Azure capacity. In 2025, Oracle deepened its role as a Stargate infrastructure co‑developer on multiple U.S. data‑center sites. Oracle+1
- “Stargate” $500B program: OpenAI introduced The Stargate Project (Jan 21, 2025), a new company/joint venture that intends to invest $500 billion over four years to build ~10 GW of U.S. AI infrastructure; OpenAI, SoftBank, Oracle and MGX are involved. OpenAI then announced five new Stargate sites (Sept 23, 2025), saying the program is ahead of schedule toward the $500 billion/10 GW target by end‑2025. OpenAI+1
- Valuation: Following a secondary share sale in early October, OpenAI’s valuation is ~$500 billion (not a primary raise). Multiple outlets (Reuters, Bloomberg) reported employees sold ~$6.6 billion in shares to a buyer consortium. Reuters+1
1) NVIDIA × OpenAI: 10 GW roadmap and a staged $100 billion commitment
On Sept 22, 2025, OpenAI and NVIDIA unveiled a strategic partnership to build and deploy at least 10 GW of AI data‑center capacity using NVIDIA systems — described as “millions of GPUs” for OpenAI’s next‑generation models. To support the deployment, NVIDIA intends to invest up to $100 billion in OpenAI, drawn down progressively per‑gigawatt. The partners target the first 1 GW in 2H 2026 on the Vera Rubin platform. This is framed as infrastructure financing and technology enablement rather than a single lump‑sum equity check. OpenAI+1
Why it matters: If executed as announced, 10 GW would be unprecedented in AI compute, helping OpenAI bypass a key bottleneck (GPU access) and locking in next‑gen NVIDIA systems at scale. It also complements OpenAI’s multi‑vendor strategy (see AMD, Oracle below). Financial Times
2) AMD × OpenAI: supply pact plus up to 10% equity via a warrant
On Oct 6, 2025, AMD disclosed a multi‑year AI chip‑supply deal with OpenAI. The agreement provides OpenAI access to hundreds of thousands of accelerators, with deployment plans around ~6 GW beginning late 2026 (including a 1 GW facility). As part of the deal, OpenAI received a warrant to purchase up to 160 million AMD shares (≈10% of AMD) at $0.01/share, vesting in tranches tied to delivery/technical milestones and other conditions. The news sent AMD shares sharply higher. Reuters+1
What the warrant means (in plain English): OpenAI doesn’t instantly own 10% of AMD; it has the option to obtain that stake if it achieves the agreed milestones. This aligns incentives: AMD wins a marquee customer and volume; OpenAI gains a potential financial upside while diversifying beyond NVIDIA. Reuters
3) Oracle × OpenAI: cloud capacity extension and Stargate build‑out
- Cloud capacity (2024): On June 11, 2024, Oracle announced OpenAI selected OCI to extend Azure’s AI platform, giving OpenAI additional capacity across OCI AI superclusters. Microsoft remains OpenAI’s primary cloud partner; OCI operates as an integrated extension. Oracle
- Stargate infrastructure (2025): Oracle’s role expanded in 2025 as a build partner and investor across new U.S. sites under Stargate, a national‑scale AI infrastructure effort backing OpenAI’s compute needs. Trade and infrastructure reporting detail projects and multi‑site commitments over the next several years. Data Center Frontier
Takeaway: Oracle has become a critical capacity and construction partner — not just a cloud vendor — in OpenAI’s push to secure massive, geographically distributed compute. Data Center Frontier
4) “Stargate”: the $500 billion, ~10 GW AI infrastructure program
- Launch: On Jan 21, 2025, OpenAI announced The Stargate Project — a new company that intends to invest $500 billion over four years to build ~10 GW of U.S. AI infrastructure dedicated to OpenAI workloads; it said $100 billion would begin deploying immediately. Partners/investors include SoftBank, Oracle, MGX, with OpenAI handling operations. OpenAI
- Progress: On Sept 23, 2025, OpenAI said five new sites push Stargate ahead of schedule toward the full $500 billion/10 GW goal by end‑2025. The company highlighted a pipeline that, if realized, would secure the entire commitment this year. OpenAI
- Context: “Stargate” originally surfaced in 2024 reporting as a Microsoft–OpenAI supercomputer concept priced near $100 billion (separate from the 2025 joint‑venture structure). The 2025 program is substantially larger and spans multiple partners. Reuters
Bottom line: Stargate formalizes an industrial‑scale buildout of AI compute — with U.S. siting, multi‑partner financing, and a multi‑GW roadmap designed to underwrite OpenAI’s next model generations. OpenAI
5) Valuation: ~$500 billion after a secondary sale (Oct 2025)
In early October 2025, OpenAI completed a secondary share sale (existing shareholders transacting), implying a ~$500 billion valuation — up from ~$300 billion earlier in 2025. Reuters and Bloomberg report ~$6.6 billion in employee shares sold to a buyer group that included large global investors. Reuters+1
Note: Because it’s a secondary, the proceeds largely went to sellers (not OpenAI’s balance sheet). Still, the implied valuation places OpenAI among the world’s most valuable private companies. Reuters
Strategic implications (for readers)
- Supply diversification: Pairing the NVIDIA roadmap with a large AMD commitment and Oracle capacity reduces single‑vendor risk and could improve cost/performance over time. Reuters+1
- Financing at unprecedented scale: The $100 billion staged NVIDIA commitment and $500 billion Stargate plan reflect a shift from classic SaaS scaling to capital‑intensive infrastructure, with complex vendor‑financing, warrants, and multiyear offtake structures. NVIDIA Newsroom+1
- Execution risks: Multi‑GW power procurement, grid interconnects, supply chains (HBM, packaging), and permitting are gating factors. Nonetheless, OpenAI’s September update claims pipeline sufficiency to meet the 2025 Stargate objective. OpenAI
What’s confirmed vs. contingent (quick guide)
- CONFIRMED by official releases:
- CONTINGENT/CONDITIONAL:
- MARKET‑IMPLIED:
- ~$500 billion valuation came via secondary share pricing, not a primary capital raise; valuation can fluctuate with future transactions. Reuters
Timeline highlights
- Mar 29, 2024: Reporting surfaces “Stargate” as a Microsoft–OpenAI ~$100 billion supercomputer concept (2028 target). Reuters
- Jun 11, 2024: Oracle says OpenAI selected OCI to extend Azure capacity. Oracle
- Jan 21, 2025: OpenAI announces “Stargate Project” — $500 billion/four years, ~10 GW plan. OpenAI
- Sept 22–23, 2025: NVIDIA–OpenAI 10 GW partnership (with “up to $100 billion” staged NVIDIA investment); OpenAI adds five Stargate sites (ahead of schedule). OpenAI+1
- Oct 6, 2025: AMD–OpenAI deal announced, including warrant up to ~10% of AMD. Reuters
- Oct 2–6, 2025: Reports of ~$500 billion valuation following secondary share sale. Bloomberg+1
Sources & further reading (selected)
- OpenAI Partnerships and Major Investments Overview
- OpenAI ↔ NVIDIA joint announcement; NVIDIA newsroom (10 GW plan; staged $100 B). OpenAI+1
- AMD ↔ OpenAI deal, 10% warrant (Reuters; The Guardian). Reuters+1
- Oracle ↔ OpenAI cloud extension (Oracle press release); Stargate coverage (Data Center Frontier). Oracle+1
- Stargate official posts (launch; site expansion). OpenAI+1
- Valuation reports (Reuters, Bloomberg). Reuters+1
