Nvidia to invest up to $100 billion in OpenAI with 10-gigawatt chip pact, drawing antitrust scrutiny

Nvidia commits up to $100B to OpenAI: implications for capital flows, competition, and your models

Nvidia will invest up to $100 billion in OpenAI and supply it with data center systems. The structure: Nvidia takes non-voting equity; OpenAI uses the proceeds to buy Nvidia hardware. Two transactions, one looped cash flow.

The companies signed a letter of intent to deploy at least 10 gigawatts of Nvidia systems for OpenAI. That's power on the scale of more than 8 million U.S. households. Deliveries could start in late 2026, with the first gigawatt in the second half on Nvidia's upcoming Vera Rubin platform.

Deal snapshot

• Investment size: up to $100B (non-voting shares).
• Initial tranche: $10B after a definitive purchase agreement for Nvidia systems.
• Compute: at least 10 GW of Nvidia systems earmarked for OpenAI.
• Timing: first deployments 2H 2026; platform named "Vera Rubin."
• OpenAI valuation context: ~$500B.

"Everything starts with compute," said OpenAI CEO Sam Altman. The message is clear: access to advanced chips is the constraint, and capital follows the bottleneck.

Market reaction

• Nvidia shares rose as much as 4.4% to a record intraday high after the announcement.
• Oracle gained about 6% amid its role building data centers and its link to the $500B "Stargate" initiative with OpenAI, SoftBank, and Microsoft.
• Broadcom slipped 0.8% on the day; OpenAI's custom silicon effort with Broadcom and TSMC continues unchanged.

Strategic context

This deal layers on top of existing ties: Microsoft has invested billions into OpenAI since 2019; Nvidia recently announced a collaboration with Intel on AI chips; Nvidia also committed $5B to Intel and participated in OpenAI's $6.6B round in October 2024. The network is consolidating around capital, compute, and distribution.

Analysts flagged "circular" concerns: some investment dollars may loop back to Nvidia via chip purchases. As Bernstein's Stacy Rasgon put it, the agreement supports ambitious infrastructure buildouts, but fuels questions about circular financing.

Antitrust and policy risk

The Justice Department and FTC have a framework to probe Microsoft, OpenAI, and Nvidia's roles in AI after a 2024 jurisdictional agreement. The current U.S. administration has taken a lighter stance on competition than the prior one, but scrutiny could increase as dollar amounts and market share rise.

Antitrust lawyer Andre Barlow warned the deal could entrench Nvidia's chip lead with OpenAI's software lead, making it harder for rivals like AMD and competing model providers to scale.

• Reference points: Nvidia investor relations; DOJ Antitrust Division

OpenAI's parallel chip track stays intact

OpenAI's custom chip plans (with Broadcom and TSMC) reportedly remain in place. Its partnership with Microsoft also continues. Expect a dual-track approach: secure near-term capacity via Nvidia while pursuing cost/control with in-house silicon over time.

Why this matters for finance

• Capex visibility: A 10 GW build implies multi-year AI data center spend with knock-on effects for builders, power, networking, and memory.
• Gross margin durability: Watch how Nvidia balances premium pricing against long-term supply commitments and potential regulatory constraints.
• Customer concentration: OpenAI becomes an even larger anchor tenant for Nvidia. That concentration cuts both ways for forecasting risk.
• Policy overhang: Any probe that slows procurement or alters deal terms can shift deployment timelines and revenue recognition.
• Second-order winners: Data center landlords, grid upgrades, and integrators could see sustained demand; vendors without exclusive positions may face pricing pressure.

Timeline and catalysts

• Near term: Finalize definitive agreements; clarity on exact investment tranches and purchase orders.
• 2026: First deliveries on Vera Rubin; track capacity on-ramps by quarter.
• Regulatory: Signals from DOJ/FTC; any conditions around exclusivity, supply allocation, or data center buildouts.
• Competition: Updates from AMD on high-end GPU wins; progress on OpenAI's custom chips (tape-outs, yields, and performance targets).

Portfolio implications

• Nvidia: Positive demand signal and deeper integration with a top buyer; monitor "circular" optics and any constraints from antitrust review.
• Microsoft: Indirect beneficiary via OpenAI alignment and Azure demand; watch for capital efficiency and margin mix in AI workloads.
• Oracle: Data center build momentum; confirm contract scope and utilization tied to Stargate.
• Broadcom/TSMC: Custom silicon optionality remains; near-term narrative may skew to Nvidia until custom chips ship at scale.
• AMD/Intel: Path depends on securing flagship AI deployments; pricing flexibility could be the wedge.
• Utilities/power: 10 GW commitments highlight grid and power procurement as a gating factor; assess regional exposure and PPAs.

Practical next steps for finance teams

• Update AI capex models: bake in a staged 2026-2028 compute ramp and sensitivity to delivery slippage.
• Stress-test margins: scenario-plan for GPU pricing normalization vs. scarcity premiums.
• Track procurement signals: watch for component lead times (HBM, networking) and power availability that can delay revenue.
• Monitor policy: any consent decrees, exclusivity limits, or divestiture talk would alter risk/reward.

If you're budgeting AI adoption inside finance teams, here's a vetted overview of tools and costs to benchmark against: AI tools for Finance.