Nvidia’s written supply and capacity commitments rose from $28.9 billion in October 2024 to $279 billion in July 2026, according to its own quarterly filings: about 9.7x in seven quarters, by our arithmetic. Elon Musk wrote on X on 14 November that many of Nvidia’s supply deals “were a handshake with Jensen, not a written contract”.
This explainer sets out what Nvidia’s filings say is written down, what they say can still change, and what a memory supplier says about its own contracts. Our piece on Musk’s post covers the claim itself; this one covers the mechanics.
Business Pill · Vertical integration
An evergreen Business Engineer explainer on vertical integration. It teaches the general idea only and does not discuss Nvidia’s filings or Musk’s post.
The key insight: As we read the filings, the line Musk’s post points at has grown fastest in the most recent quarters: Nvidia’s written supply and capacity commitments were roughly flat near $30 billion through April 2025, then reached $95.2 billion in January 2026 and $279 billion in July 2026, by its own figures.
Nvidia’s 10-Q for the quarter ended 26 July 2026 lists $279 billion of supply and capacity commitments. The filing says Nvidia has “partnered with our extensive network to secure the necessary supply and critical components needed to meet demand for the next several years”, and that the commitments are “primarily memory and manufacturing facilities”.
The same table lists four other kinds of commitment: $29 billion of cloud service agreements, $25 billion of data center leases not yet started, $25 billion of equity investments and $8 billion of capital expenditures. Separately, it lists $36 billion of AI cloud agreements and $20 billion of data center leases it expects to reassign to third parties.
Nvidia supply and capacity commitments at each quarter-end, as stated in its 10-Q and 10-K filings from October 2024 to July 2026; labels changed across filings, July 2025 is left out because only an all-commitments total was given, and 9.7x is our arithmetic.
How the Number Grew, Quarter by Quarter
Nvidia has described the line in slightly different words over time, so the series below uses each filing’s own figure and label. In October 2024 the 10-Q gave “outstanding inventory purchase and long-term supply and capacity obligations totaling $28.9 billion”. The annual report for January 2025 gave $30.8 billion and the April 2025 10-Q $29.8 billion.
From October 2025 the filings give a commitments figure for manufacturing, supply and capacity: $50.3 billion in the October 2025 10-Q, $95.2 billion in the January 2026 annual report and $119 billion in the April 2026 10-Q. The July 2025 10-Q gave only a $45.8 billion total for all commitment types, so it is left out of the series.
From January to July 2026, the figure rose about 2.9x in two quarters, by our arithmetic. At $279 billion, it is about 8.8x the $31.6 billion of inventory Nvidia held on 26 July 2026.
What Can Still Change
Written does not mean fixed. The July 2026 10-Q says Nvidia enters into agreements that let suppliers procure inventory to its criteria, and “in certain instances, these agreements may be cancelable, rescheduled, or adjustable for our businessneeds prior to placing firm orders”. It adds: “Changes to these agreements may result in additional costs”.
The annual report for fiscal 2026 describes the other direction: “In periods of growth, we may place non-cancellable inventory orders for certain product components in advance of our historical lead times, pay premiums, or provide deposits to secure future supply and capacity and may need to continue to do so”.
The prepaid part has been small next to the totals. The January 2025 annual report says “Prepaid supply and capacity agreements of $3.3 billion” were in current assets, against $30.8 billion of obligations that quarter.
When Demand Moves, the Written Part Costs Money
The April 2025 10-Q shows what happens when written commitments outrun demand. After the US government required a licence for H20 exports to China, Nvidia booked a $4.5 billion charge for H20 excess inventory and purchase obligations in the quarter, including a $2.6 billion excess inventory purchase obligation charge for H20 orders, the filing says.
On 26 July 2026, Nvidia carried $2.1 billion of excess inventory purchase obligations in accrued liabilities, down from $2.7 billion in January, according to the 10-Q.
The Written Deals That Can Run the Other Way
Some written agreements give the other side the option. On the $36 billion of AI cloud agreements, the 10-Q says AI clouds “procure our data center infrastructure products and we commit to cloud service agreements, which the AI clouds can unilaterally stop providing to us and sell to third-party customers at more advantageous rates”. Our piece on Nvidia’s and Broadcom’s backstops covers those arrangements.
What a Memory Supplier Says About Its Contracts
Nvidia says most of the $279 billion is for memory and manufacturing facilities. SK hynix, one of the main memory makers, said in its third-quarter release of 28 October that it “has currently signed LTAs with 14 major customers, including leading global cloud service providers (CSPs)”, with contract terms that are “typically five years”.
SK hynix said “The LTA volume secured to date corresponds to more than 50% of total production capacity for 2027 and 2028”, priced through “a mechanism that reflects market conditions within a price range agreed upon with customers”. The release does not name the 14 customers. Our July piece on Nvidia and SK hynix covers the earlier memory deal reports.
Where Musk’s Account Fits
Musk’s post makes a claim about how agreements were reached: “Without Jensen, at least some of those critical deals will be lost”. Nvidia’s filings give totals and terms, not how each agreement was negotiated or who negotiated it, and they do not split the $279 billion by supplier.
Musk did not name a supplier or a deal. The filings show a commitments figure that kept rising through the July quarter; they cannot show allocation decisions made outside written agreements.
Frequently Asked Questions
How much has Nvidia committed to suppliers?
$279 billion of supply and capacity commitments as of 26 July 2026, per its 10-Q, out of $366 billion of total future commitments in that table.
Are those commitments binding?
Partly. The 10-Q says some agreements may be cancelable, rescheduled or adjustable before firm orders, and that changes may add costs. The annual report says Nvidia may also place non-cancellable orders, pay premiums or give deposits.
What are the commitments for?
Data center infrastructure systems, “primarily memory and manufacturing facilities”, the 10-Q says.
Has a written commitment ever cost Nvidia money?
Yes. The April 2025 10-Q reports a $4.5 billion H20 charge for excess inventory and purchase obligations after the US licence requirement.
What do suppliers say?
SK hynix says it has long-term agreements with 14 major customers covering more than half of its 2027 and 2028 capacity, typically for five years. It does not name the customers.
The Structural Read
As we read it, the filings describe two kinds of written terms: agreements that may be cancelable, rescheduled or adjustable before firm orders, and non-cancellable orders, premiums or deposits placed in periods of growth.
As we read the April 2025 10-Q, the H20 charge shows the cost side of written commitments: a $4.5 billion charge for excess inventory and purchase obligations after the US licence requirement.
Nvidia 10-Q, quarter ended 26 July 2026
“Changes to these agreements may result in additional costs.”
Three Implications
FOR NVIDIA’S SUPPLIERS Nvidia’s 10-Q says the $279 billion is primarily for memory and manufacturing facilities, and that some agreements may be adjusted before firm orders.
FOR MEMORY MAKERS SK hynix says it has long-term agreements with 14 major customers covering more than 50% of its 2027 and 2028 capacity, with terms typically five years.
FOR INVESTORS The filings give totals and terms; they do not split the $279 billion by supplier or by how each agreement was negotiated.
The framework says: “HBM growth is outpacing accelerator demand, making memory the rate-limiting factor in the AI scaling equation.”
As we read it, Nvidia’s 10-Q describing its commitments as primarily memory and manufacturing facilities fits that frame; the framework predates the filing and does not discuss it.
What Is Not Established
The filings do not name suppliers, give the share of the $279 billion that is cancelable, or say how much was added on the strength of any individual’s agreement. Musk’s claim about unwritten deals is Musk’s own and is not confirmed by Nvidia’s filings or by any supplier we read.
The labels for the series changed across filings, as noted above. Nvidia and SK hynix were not contacted for this explainer.
This Week In The AI Supercycle
AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source
This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.
Nvidia’s filings show $279 billion of written supply and capacity commitments on 26 July 2026, up from $28.9 billion in October 2024, mostly for memory and manufacturing, with some agreements adjustable before firm orders. Musk’s point is about what sits outside those documents; the filings cannot measure that. Nvidia reports third-quarter results on 17 November, and our first piece on the exit began this coverage.
A note on sourcing. The commitments figures, terms and charges come from Nvidia’s 10-Q and 10-K filings for October 2024 through July 2026, read on SEC EDGAR on 14 November 2026; the supplier contract terms come from SK hynix’s third-quarter release of 28 October 2026; the Musk quote comes from Musk’s post on X of 14 November. Ratios are our arithmetic. Nvidia and SK hynix were not contacted for this explainer.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
