CNBC’s read of NVIDIA’s latest 10-Q and CFO Colette Kress’s August 26 earnings call together disclose a structure the company itself has now quantified: vendor, investor, and underwriter to the same handful of AI counterparties — all at once.
CNBC’s September 4 reading of NVIDIA’s most recent 10-Q found that the company’s strategic investment holdings — when three separate balance-sheet lines are summed — total roughly $99 billion as of July 26, 2026. The three lines: approximately $51.2 billion in non-marketable and equity-method investments, roughly $42.8 billion in marketable equity securities, and about $5 billion in long-term publicly-held equity. That roll-up is CNBC’s aggregation, not a single figure the filing reports; the accurate framing is “roughly $99 billion across three lines,” not “the 10-Q reports $99 billion.” One year earlier, the same three-line sum came to about $7 billion, confirmed in the prior-year 10-Q — roughly a fourteen-fold increase in twelve months.
On NVIDIA’s August 26 earnings call, CFO Colette Kress provided the destination of that capital. She said NVIDIA has invested “nearly $50 billion in the frontier AI labs” — that figure is her call statement, not a disclosure in the filing itself. She also said those labs are expected to represent approximately a quarter of NVIDIA’s business next year. Two numbers, one sentence: the companies NVIDIA is betting on financially are also the companies projected to supply a quarter of its revenue.
The 10-Q names no equity investee. Individual stakes reported in the press — roughly $30 billion in OpenAI, approximately $2 billion each in CoreWeave and Nebius, smaller amounts in optical and networking suppliers — come from company announcements and external reporting, not from the filing. Other names circulating in coverage, including specific values tied to xAI or garbled Intel and SpaceX figures, are unconfirmed and are not asserted here. Separately, the filing discloses a guarantee capped at $105 billion, tied to leases for an OpenAI Group affiliate’s approximately 4.25 gigawatts of compute capacity, plus $25 billion in future equity commitments. That $105 billion guarantee is credit support, not equity — it should not be added to the portfolio total.
NVIDIA’s strategic investment portfolio totals roughly $7 billion across the same three balance-sheet lines.
H1 FY2026 — 10-Q Disclosure
NVIDIA purchases roughly $42 billion of equity securities in the first half of the fiscal year alone. Future equity commitments of $25 billion also disclosed. A $105B guarantee tied to an OpenAI affiliate’s compute leases is separately disclosed — credit support, not equity.
August 26, 2026 — Earnings Call
CFO Colette Kress states “nearly $50 billion in the frontier AI labs” and projects those labs at roughly a quarter of next year’s revenue. Jensen Huang’s stated position: regret is not investing more and sooner.
Alongside — Third-Party Financing Platform
NVIDIA announces a platform mobilizing more than $500 billion with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR — extending the financing infrastructure beyond NVIDIA’s own balance sheet.
July 26, 2026 — Latest 10-Q (as-of date)
Three-line portfolio total: ~$51.2B non-marketable/equity-method + ~$42.8B marketable equity + ~$5B long-term public ≈ $99B (CNBC’s aggregation). The filing names no individual investee.
The key insight: NVIDIA has now quantified the loop itself. The company that manufactures the chips is simultaneously the largest disclosed strategic investor in the companies that buy those chips — and projects those same companies to represent a quarter of its revenue next year. Revenue and balance sheet are no longer independent variables; NVIDIA’s own disclosures confirm the entanglement is measured in the hundreds of billions.
The Structural Read
This is circular financing at the largest scale in the industry — and what makes this moment distinct is that NVIDIA has quantified the loop itself rather than leaving it to analysts to infer. The structure is simple to state and hard to unwind: NVIDIA sells GPUs to frontier labs, invests billions in equity into those same labs, and backstops their infrastructure leases through guarantees, so its product demand, its investment portfolio, and its contingent liabilities all converge on the same handful of counterparties.
Earlier today, a smaller version of the same pattern surfaced when a GPU cloud took equity in a robotics customer it also supplies — covered in the Nscale-Figure compute deal analysis here. NVIDIA is the same structure with two more zeros. The difference is that NVIDIA is not just a participant in this pattern — it has become its architect at scale, extending the loop outward through a $500B-plus third-party financing platform with Apollo, BlackRock, Blackstone, Brookfield, Goldman, and KKR. When a vendor finances the customers who fund the vendor, and then mobilizes the broader financial system to do more of the same, demand and balance sheet stop being independent variables. Both the bull and bear readings of that are legitimate, and they need to sit side by side.
CFO Colette Kress — NVIDIA Earnings Call, August 26, 2026
“Nearly $50 billion in the frontier AI labs… [those labs] will be roughly a quarter of our business next year.”
The bull case — made explicitly by Kress and Jensen Huang — is that this is ecosystem-building only NVIDIA is positioned to execute. Outside lenders still underwrite each deal on its own merits; NVIDIA makes no direct loans; the underlying equipment is redeployable if a borrower fails. The company with the clearest view of AI demand is seeding the buildout that only its chips can enable. Huang’s stated regret is not having invested more and sooner. That is a coherent argument: if you have an information advantage on which bets will pay off, deploying capital alongside your product is rational compounding, not conflict of interest.
The bear case is the mirror image, and equally coherent. Circular financing can flatter demand. Revenue that is partly underwritten by the vendor’s own capital is not the same quality signal as revenue from independently funded buyers — the vendor’s investment is partly what makes the purchase possible. And the entanglement concentrates risk directionally: a genuine AI-spending slowdown would hit NVIDIA’s income statement, its investment portfolio, and its guarantee exposure simultaneously, in the same direction, from the same counterparties. There is no natural hedge inside the structure. For a fuller treatment of how these dynamics connect to the week’s supply-chain and unit-economics threads, see the Five Through-Lines synthesis.
Business Engineer Framework — Map of AI
Vendor-as-Investor-and-Creditor: The Enabler-Underwriter Tension
In the Map of AI stack, NVIDIA occupies the infrastructure layer — the picks-and-shovels position that is supposed to profit regardless of which application layer wins. What the $99B portfolio and the $105B guarantee disclose is that NVIDIA has stepped out of that neutral position and into the capital layer simultaneously. It is now both the enabler of the AI buildout and its underwriter. That is a structurally different company than the one the infrastructure-layer thesis describes — and the disclosures don’t resolve which version of the thesis applies. They establish that the stakes of the question are measured in the hundreds of billions.
Three Implications
IMPLICATION 1 — Revenue Quality Has a New Variable
When a supplier’s own capital partially underwrites its customers’ ability to buy, reported revenue carries a different signal than revenue from independently funded demand. That does not make the revenue fake — NVIDIA’s chips are real, the compute is real — but it means the demand signal is partly endogenous to NVIDIA’s own balance-sheet decisions. Analysts and operators modeling NVIDIA’s revenue trajectory now need to factor in how much of that demand would persist if NVIDIA’s own capital were withdrawn. Kress’s “quarter of next year’s business” figure is the number to watch for that reason above all others.
IMPLICATION 2 — The $500B Financing Platform Changes the Systemic Scope
NVIDIA’s own balance sheet — $99B in holdings, $25B in future commitments, $105B in guarantees — is large. But the separately announced third-party financing platform, mobilizing more than $500 billion with Apollo, BlackRock, Blackstone, Brookfield, Goldman, and KKR, takes the loop beyond NVIDIA’s own capital and into the broader financial system. If the circular-financing risk materializes, the question is no longer just whether it hurts NVIDIA — it is how widely the stress propagates through the institutions that co-signed the infrastructure buildout alongside it.
IMPLICATION 3 — The Infrastructure-Layer Neutrality Thesis Needs Updating
The canonical bull case for infrastructure players is neutrality — you sell to all sides and collect margin regardless
This is businessanalysis, not investment advice and no view on the stock. The ~$99B total is CNBC’s aggregation of three 10-Q balance-sheet lines (as of July 26, 2026), not a single filing figure; the ~$50B-in-labs figure is CFO Colette Kress’s Aug 26 call statement. The filing names no investee — individual stakes are from reporting. The circular-financing risk is analysis, which NVIDIA disputes.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
