Why it matters
Nvidia sees a ~$600B compute opportunity through 2030 in OpenAI's existing and planned capacity, far above the $30B equity investment OpenAI announced. Nvidia counts ~12 of its systems in that capacity, or ~16 GW if Nvidia extends its PORTS-Pike rent guarantee. Nvidia has never stated the investment figure in a filing or earnings call.
Tap a figure for the line in its source.
Nvidia gives OpenAI
- in OpenAI's Feb 2026 funding round, by OpenAI's count; the 2025 letter of intent had said up to $100B, gigawatt by gigawatt, and no Nvidia filing states the $30B or shows it paid
OpenAI gives Nvidia
- OpenAI shares, priced at a $730B pre-money valuation; Nvidia's stake is undisclosed
- of Nvidia computing in OpenAI's existing and planned commitments through 2030, by Nvidia's count, or ~16 if Nvidia extends its PORTS-Pike guarantee beyond the first 4.25 GW; at those levels, roughly $600B of Nvidia compute, an opportunity rather than a purchase contract
What happened
In September 2025, Nvidia and OpenAI signed a letter of intent. OpenAI would deploy at least 10 gigawatts of Nvidia systems. Nvidia would invest up to $100 billion in OpenAI as each gigawatt came online. Nvidia's filings called it an opportunity to invest and gave no assurance of a deal. In February 2026, OpenAI announced $30 billion from Nvidia in its $110 billion funding round. Nvidia's filings and earnings calls never state that figure. Nvidia's filings show that it invested $18.6 billion across all private companies and funds combined in the quarter the round closed. That total is smaller than the amount OpenAI announced for Nvidia alone. In August 2026, Nvidia agreed to stand behind part of the rent on OpenAI's leased data centers in Ohio. Nvidia would pay only if OpenAI defaults. That guarantee has a separate deal record. Nvidia counts about 12 gigawatts of its hardware in OpenAI's existing and planned computing. Nvidia puts the figure at about 16 if it expands its Ohio guarantee. Nvidia estimates an opportunity worth roughly $600 billion in computing on its systems through 2030.
Sources 18 documents — 5 company releases · 5 SEC filings · 5 press reports · 1 transcript · 1 other document · 1 research note
Research trail Found 1,691 documents and read 1,578 of them while checking this deal and related ones.
- States 3 figures: up to $100B LOI investment · at least 10 GW LOI capacity · LOI first gigawatt
- States 2 figures: $30B round investment · 3 GW vera rubin capacity
- States 2 figures: 12 GW nvidia compute commitments · $600B compute opportunity
- States LOI status
- States agreement status
- States $18.6B funding evidence
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
- Cited by the record; no single figure is quoted from it.
The record's own summary of its sourcing: Nvidia + OpenAI LOI releases (Sep 22 2025); Nvidia 10-Q (Nov 2025), FY2026 10-K (Feb 2026), Q1 + Q2 FY2027 10-Qs and Q2 CFO commentary; OpenAI round posts (Feb 27 + Mar 31 2026); Nvidia blog (Aug 17 2026) + Q2 FY2027 call; CNBC, TechCrunch and press relays for the contested cells · as of Aug 2026
The evidence 10 sourced figures · 2 disputes
The record's headline Nvidia–OpenAI: up-to-$100B letter of intent (Sep 2025) → $30B of OpenAI's 2026 round, by OpenAI's count — no Nvidia filing states either figure
- Size
- $30B in OpenAI's round, by OpenAI's count · the LOI said up to $100B (neither in any Nvidia filing)
- When
- Sep 2025 LOI → Feb–Mar 2026 round
- Status
- $30B per OpenAI, unconfirmed by Nvidia — the LOI last appears in Nvidia's Nov-2025 10-Q; the round closed Mar 31 2026 on committed capital; no Nvidia filing shows the $30B paid
- Structure
- letter of intent (up to $100B, to be invested as each GW of Nvidia systems is deployed — 'an opportunity to invest' in Nvidia's 10-Q) → a share of OpenAI's priced Feb-2026 round (amount from OpenAI; form and funding undisclosed by Nvidia); in Aug 2026 a third form, contingent lease credit support at PORTS-Pike (nv-sbe-ports-pike)
- Programme
- Nvidia circular capital
- Parties
- Nvidia investor + gpu-supplier · OpenAI investee + chip-buyer
- Capital
- the seller's own money or backing comes back as its sales
- Record
- disclosed / primary source · as of Aug 2026
Flows 2 as the record states them
- Nvidia → OpenAIequity$30B in OpenAI's Feb-2026 round, by OpenAI's count — no Nvidia filing states it, its form or its payment (Nvidia's Q1 FY2027 investments in all private companies and infrastructure funds: $18.6B); the Sep-2025 LOI's up-to-$100B, per GW deployed, was never filed as a commitment
- OpenAI → Nvidiachip-spendLOI: at least 10 GW of Nvidia systems, the first GW in H2 2026 on Vera Rubin · Feb 2026: the use of 3 GW of inference + 2 GW of training on Vera Rubin (OpenAI) · Aug 2026: ~12 GW of Nvidia compute in OpenAI's existing and planned commitments, ~16 GW if Nvidia extends PORTS-Pike beyond its initial 4.25 GW; at those levels ~$600B through 2030 (Nvidia — an opportunity, not a contract)
Resolved figures 10 each with the verbatim line from its document
- loi-investmentup to $100Bup to $100B, 'progressively as each gigawatt is deployed' — the Sep 22 2025 letter of intent. The money was keyed to deployment ('as the new NVIDIA systems are deployed'), not to dates, and the release left definitive terms to 'the coming weeks'. No Nvidia filing ever states the $100B: the 10-Q for that quarter records only an 'opportunity to invest' (see loi-status), and Nvidia filed no 8-K for the letter or for any later investment in OpenAI (EDGAR filing index, Sep 2025 – Aug 2026: Items 2.02, 5.02, 5.07 and one 8.01 notes offering, until the Aug-17 PORTS-Pike 8-K)
“To support the partnership, NVIDIA intends to invest up to $100 billion in OpenAI progressively as each gigawatt is deployed.”
Nvidia newsroom release, 'OpenAI and NVIDIA Announce Strategic Partnership to Deploy 10 Gigawatts of NVIDIA Systems' (Sep 22 2025) · nvidianews.nvidia.com - loi-capacityat least 10 GWat least 10 GW of Nvidia systems 'representing millions of GPUs', for OpenAI to build and deploy — a floor in a letter of intent, with Nvidia 'a preferred strategic compute and networking partner', not an exclusive one. No price, purchase obligation or schedule beyond the first gigawatt appears in any document, and the basis (IT or facility power) is unstated. The later gigawatt figures are other scopes, not this one restated: 3 GW of inference + 2 GW of training on Vera Rubin (OpenAI, Feb 2026 — see vera-rubin-capacity) and ~12 GW of existing and planned commitments (Nvidia, Aug 2026 — see nvidia-compute-commitments); no document reconciles them
“OpenAI and NVIDIA today announced a letter of intent for a landmark strategic partnership to deploy at least 10 gigawatts of NVIDIA systems for OpenAI’s next-generation AI infrastructure to train and run its next generation of models on the path to deploying superintelligence.”
Nvidia newsroom release (Sep 22 2025) · nvidianews.nvidia.com - loi-first-gigawatt2026the first gigawatt in the second half of 2026, on Vera Rubin — the letter's only date. Nvidia's Q2 FY2027 commentary prepares 'for the introduction of Vera Rubin in the third quarter' (its fiscal quarter ending in late October 2026); no Nvidia filing since says whether, or where, OpenAI's first gigawatt is deployed
“The first gigawatt of NVIDIA systems will be deployed in the second half of 2026 on the NVIDIA Vera Rubin platform.”
Nvidia newsroom release (Sep 22 2025) · nvidianews.nvidia.com - loi-statusthe filed version of the letter, in the 10-Q for the quarter it was signed: 'a letter of intent with an opportunity to invest in OpenAI', with 'no assurance that we will enter into definitive agreements with respect to the OpenAI opportunity'. It was not booked as a $100B commitment: the same note puts all of Nvidia's investment commitments at $6.5B on Oct 26 2025, $5B of it Intel. This is the last Nvidia filing to mention the letter. The FY2026 10-K speaks instead of an 'investment and partnership agreement' being finalized (see agreement-status), and the Q1 and Q2 FY2027 10-Qs mention neither — the first never names OpenAI, the second names it only as the PORTS-Pike tenant
“There is no assurance that we will enter into definitive agreements with respect to the OpenAI opportunity or other potential investments, or that any investment will be completed on expected terms, if at all.”
Nvidia 10-Q, quarter ended 2025-10-26 (filed Nov 19 2025) — Liquidity and Risk Factors · sec.gov - agreement-statusFeb 25 2026, two days before OpenAI's announcement: Nvidia was 'finalizing an investment and partnership agreement with OpenAI', with 'no assurance' of a signing or a completed transaction — the sentence appears in both the risk factors and the liquidity section of the FY2026 10-K. Investment commitments at Jan 25 2026: $11.4B, 'subject to certain contingencies'. No later Nvidia filing names the agreement, its amount, its form or its closing
“We are finalizing an investment and partnership agreement with OpenAI. There is no assurance that we will enter into an investment and partnership agreement with OpenAI or that a transaction will be completed.”
Nvidia FY2026 10-K (filed Feb 25 2026) — Risk Factors; Liquidity · sec.gov - round-investment$30B$30B — OpenAI's figure for Nvidia's part of its $110B round at a $730B pre-money valuation (SoftBank $30B, Nvidia $30B, Amazon $50B), announced Feb 27 2026 beside 'next generation inference compute' secured from Nvidia. The round closed Mar 31 'with $122 billion in committed capital' at $852B post-money, 'anchored by our strategic partners Amazon, NVIDIA, and SoftBank' (OpenAI's closing post), which says neither how much was funded at the close nor what each investor put in. OpenAI calls it 'new investment' and never says cash. No Nvidia filing, earnings release, CFO commentary or earnings call on file states the amount. The Q2 FY2027 call gives only an aggregate — CFO Kress's prepared remarks say Nvidia has 'invested nearly $50 billion in the frontier AI labs', naming none and splitting nothing — and Huang on the same call speaks of 'investing in these two companies' (OpenAI and Anthropic, named in the question) and wishes he had invested 'more and sooner', with no figure. Amazon's 10-Qs track its $50B arriving in three instalments (az-openai-2026); nothing comparable exists for Nvidia's $30B (see funding-evidence)
“Today we’re announcing $110B in new investment at a $730B pre-money valuation. This includes $30B from SoftBank, $30B from NVIDIA, and $50B from Amazon.”
OpenAI, 'Scaling AI for everyone' (Feb 27 2026; Wayback snapshot) · web.archive.org - funding-evidence$18.6Bwhat Nvidia's filings show about the $30B: no amount, only a bound. In its first quarter of fiscal 2027 (Jan 26 – Apr 26 2026), which holds both the announcement and the Mar 31 close, Nvidia's investments in all private companies and infrastructure funds together came to $18.6B — so the $30B was not all invested in the quarter the round closed (the monitor's arithmetic on a filing that names no investee). Investment commitments, 'subject to certain contingencies', rose from $11.4B (Jan 25) to $27B (Apr 26) and stood at $25B on Jul 26 2026 (equity investment commitments, in the Q2 wording), $18B of it due in the rest of fiscal 2027 — in 'AI model makers, infrastructure financiers, and other private companies', none named. By Jul 26 half-year net additions to non-marketable equity had reached $31.0B (Q2 10-Q, Note 6: $17.9B + $13.1B), again unnamed, so from the Q2 filing on the filings can no longer show any of the $30B unpaid. When, and in what form, the rest was paid is not disclosed (see contested consideration-form)
“In the first quarter of fiscal year 2027, we made the following investments: •$18.6 billion in private companies and infrastructure funds. Some of these investments include AI model makers that may indirectly purchase or use our products in the cloud.”
Nvidia 10-Q, quarter ended 2026-04-26 (filed May 20 2026) — MD&A, Recent Developments, Future Objectives and Challenges · sec.gov - vera-rubin-capacity3 GW3 GW of dedicated inference plus 2 GW of training on Vera Rubin — the capacity side of the Feb-2026 announcement, in OpenAI's words 'the use of' that capacity, not a purchase: who builds, owns or hosts it is not said, and the same post places OpenAI's running Nvidia systems 'across Microsoft, OCI, and CoreWeave'. No document relates these 3 + 2 GW to the letter's 10 GW or to Nvidia's ~12 GW
“We are also expanding our long standing collaboration with NVIDIA, including the use of 3 GW of dedicated inference capacity and 2 GW of training on Vera Rubin systems.”
OpenAI, 'Scaling AI for everyone' (Feb 27 2026; Wayback snapshot) · web.archive.org - nvidia-compute-commitments12 GW~12 GW — Nvidia's count of Nvidia compute in 'OpenAI's existing and planned commitments', through 2030; ~16 GW 'if NVIDIA extends the PORTS-Pike arrangement beyond the initial 4.25 gigawatts'. The blog's paragraph before sizes that extension as 'the remaining capacity of 3.75 gigawatts' (the Q2 10-Q says 'approximately 3.8 additional gigawatts'; nv-sbe-ports-pike carries the range), and 12 plus either is ~16, so the initial 4.25 GW at PORTS-Pike is already inside the 12 — the monitor's reading of Nvidia's two paragraphs; never add nv-sbe-ports-pike's capacity on top. Nvidia does not define the commitments or say whom they are with (OpenAI's running Nvidia systems are 'across Microsoft, OCI, and CoreWeave', per its Feb-2026 post), so this is Nvidia's count of OpenAI demand, not a purchase contract between the two, and the basis (IT or facility power) is unstated. Written by Huang on Nvidia's blog the day of the PORTS-Pike 8-K (Aug 17 2026) and repeated in Kress's prepared remarks on the Aug 26 call; no filing states it
“More broadly, OpenAI has committed to substantial deployments of NVIDIA AI infrastructure through 2030. OpenAI’s existing and planned commitments represent approximately 12 gigawatts of NVIDIA compute, with an opportunity to expand to approximately 16 gigawatts if NVIDIA extends the PORTS-Pike arrangement beyond the initial 4.25 gigawatts.”
Nvidia blog, Jensen Huang, 'Securing the Infrastructure of Intelligence' (Aug 17 2026) · blogs.nvidia.com - compute-opportunity$600B~$600B — Nvidia's sizing of the relationship at the 12–16 GW levels: 'roughly $600 billion of NVIDIA compute through 2030'. An opportunity statement in a blog post, not a backlog, a contract or revenue guidance, and whether it means Nvidia revenue or the value of the systems is not said. Not OpenAI's own ~$600B target for TOTAL compute spend by 2030 across every supplier (CNBC, Feb 20 2026, unnamed sources): the same round number for a different quantity, and no document relates them. Huang's answer on the same page to 'Is this circular financing?' is about the PORTS-Pike leases: 'No. OpenAI will pay the lease.'
“At these levels, the opportunity represents roughly $600 billion of NVIDIA compute through 2030.”
Nvidia blog, Jensen Huang, 'Securing the Infrastructure of Intelligence' (Aug 17 2026) · blogs.nvidia.com
Contested figures 4 not reconciled — every position side by side, with its source and tier
Scope note what this record leaves to other records
NOT this record: the residual-value guaranties Nvidia signed on Aug 17 2026 behind OpenAI's PORTS-Pike leases — the third form Nvidia's support for OpenAI took, recorded in full (cap, phasing, exclusivity, OpenAI's reimbursement) at nv-sbe-ports-pike, which owns those figures; this record only points at it. Nvidia's third-party compute-financing platforms (MOUs with Apollo/BlackRock/Blackstone/Brookfield/Goldman/KKR, Aug 2026, 'designed to mobilize more than $500 billion of third-party capital over time' per the Q2 10-Q) — Nvidia-wide, not an OpenAI deal (not yet a record of its own). OpenAI's round beyond Nvidia's leg — Amazon's $50B (az-openai-2026), SoftBank's $30B, the $122B close — an OpenAI funding round not yet recorded separately. Nvidia's Anthropic investment (nv-msft-anthropic) and its portfolio-wide equity ($99B invested and $25B committed at Jul 26 2026), used here only to bound this deal. Not adopted: vesperosint's 10-K 'filed after the round was announced' (it preceded it by two days); The Register's ~$30B per GW and 'more than $300 billion' for 5 GW of Vera Rubin (its own arithmetic); OpenAI's ~$600B compute-spend target read as Nvidia's '$600 billion of NVIDIA compute' (different quantities).