How much will NVIDIA commit to new equity investments in independent AI developers and GPU-cloud providers from November 13, 2026 through June 30, 2027?
Latest forecast
Generated Nov 13, 2026, 12:31 PM · High effort
Write-up
Executive Summary
The forecast is $3.75 billion at the 50th percentile (median), with a 10th percentile of $0.75 billion and a 90th percentile of $27.75 billion, all in nominal US dollars.
This favors selective continued investment rather than an extrapolation of NVIDIA’s recent spending headlines. The distribution is strongly skewed upward: one large new equity agreement could dominate the subtotal, and the upper percentile is not a cap.
The forecast covers agreements first becoming legally binding from November 13, 2026 through June 30, 2027, using public evidence available through December 31, 2027. It is not a forecast of cash payments, financing-round totals, or undisclosed investment activity.
Key drivers
Much of the apparent near-term pipeline is already committed
The largest visible forthcoming transactions do not automatically add to this measurement window:
- Anthropic: its November 10 prospectus says NVIDIA had already entered an agreement to purchase approximately $10 billion of newly issued shares. A later initial public offering (IPO) closing does not create a new commitment.
- Nscale: a further $1 billion arrangement arises from a September 15, 2026 agreement, despite an expected closing around November 16. Its alternative equity/debt structure adds another classification issue, but the earlier agreement date is already decisive.
- Thinking Machines Lab: reporting says NVIDIA’s approximately $2.5 billion investment became legally binding months earlier. That exclusion has weaker documentation because the executed agreement was not obtained.
These distinctions substantially reduce the pipeline that can confidently be treated as fresh financing.
Leadership change favors tighter approval, not a demonstrated shutdown
NVIDIA’s November 12 filing records Jensen Huang’s departure and Colette Kress’s appointment as interim CEO while remaining CFO. It announces neither an investment moratorium nor a numerical allocation cap.
Reuters reports that Kress and her staff resisted large financing exposures involving OpenAI and Lambda. Conversely, her August 26 earnings-call remarks defended equity investments in leading AI developers and their expected returns. The evidence therefore supports greater selectivity and lower tolerance for exceptional exposure—not an assumption that all new equity investment stops.
The remaining opportunities are real but conditional
Reported discussions include Figure, Firmus and Perplexity. Their current binding status, exact NVIDIA allocation or qualifying structure remains unverified. Reflection could result in a minority investment, but acquisition, hiring, licensing or computing-capacity alternatives would not automatically qualify.
OpenAI’s continuing fundraising and planned 2027 IPO provide potential large-investment opportunities, but the research found no confirmed new NVIDIA allocation. The earlier “up to $100 billion” letter of intent is not an unused contractual funding pool.
A few transactions matter more than a monthly spending rate
The historical record is concentrated in large transactions: OpenAI’s February 2026 announcement identified a $30 billion NVIDIA investment. That supports retaining a substantial upper tail even if ordinary dealmaking slows.
The background $42.4 billion of equity purchases and $25 billion of outstanding commitments measure different things: cash purchases and an existing stock of obligations. Neither establishes new qualifying agreements in this window. Public disclosure is also a separate constraint: a real investment can remain outside the subtotal if its binding date or NVIDIA-attributable amount cannot be verified.
Scenarios
The greatest weight rests on limited verified commitments and selective continuation. Complete cessation is possible, while renewed exceptionally large commitments receive less weight but account for much of the upside.
| Scenario | Why it carries weight | How it affects the outcome |
|---|---|---|
| No publicly quantified qualifying commitments | Approvals could pause, negotiations could slip, or public evidence could remain insufficient. Continued strategic incentives argue against making this the main case. | No qualifying amount is established; this would not prove that NVIDIA made no investments. |
| Limited verified commitments | Existing large obligations are excluded, the visible fresh pipeline is thin, and tighter approval could favor smaller transactions. This is the most emphasized individual case. | A small subtotal emerges from a limited number of disclosed subscriptions. |
| Selective continuation | Kress has defended equity investing, and NVIDIA retains incentives to support customers and its developer ecosystem. | Several qualifying subscriptions produce a larger subtotal without restoring the previous pace of exceptional commitments. |
| One large new subscription | Major developers continue to need financing, and historical transactions demonstrate the possible scale. No such new NVIDIA allocation is confirmed. | One agreement dominates the subtotal and pushes the outcome toward the upper part of the distribution. |
| Several large or exceptional strategic agreements | This requires both renewed approval appetite and multiple financing opportunities within the window. It receives the least weight. | Mandatory multiyear tranches count when committed, potentially producing an outcome beyond the upper percentile. |
Implications
- Do not treat NVIDIA as an automatic source of fresh financing. Selective investment is more strongly supported than either an assured backstop or a complete withdrawal.
- Separate existing funding from new approval appetite. Payments against old commitments can continue even while new investment slows sharply.
- Plan for concentrated exposure. The wide, upward-skewed distribution means the middle outcome alone is inadequate for assessing potential new obligations. One multiyear subscription could create a large measured commitment without an equally large immediate cash payment.
- Interpret a low subtotal carefully. It would not establish that NVIDIA stopped supporting the ecosystem: old obligations, debt, guarantees and commercial arrangements are outside this measure, as are unquantified amounts.
Signs to watch
| Timing or development | What would matter | Direction of change |
|---|---|---|
| Around November 16, 2026: expected Nscale closing; November 19, 2026: expected Anthropic listing/closing | Whether disclosures merely complete old obligations or establish genuinely new incremental subscriptions. | Closing alone should not move the forecast higher. New qualifying increments would move it higher; clearer agreement terms would narrow uncertainty. |
| November 17, 2026: scheduled quarterly results; November 18, 2026: company meeting | Statements distinguishing continued equity investment from a pause, exposure caps or reserved approvals during the CEO search. | An explicit pause would move the forecast lower and narrow the upper tail. Confirmed fresh approvals would move it higher. |
| Unscheduled: permanent CEO appointment or board investment-policy announcement | Who can approve subscriptions, and whether large commitments require different approval or are deferred. | A restrictive mandate would move it lower; demonstrated continuity would move it higher. A clear policy would narrow uncertainty. |
| Before June 30, 2027: live negotiations become signed agreements | Figure, Firmus, Perplexity, Reflection or another recipient could disclose attributable primary-equity terms. Reflection’s transaction structure is especially important. | Qualifying subscriptions would move it higher. Deferral beyond the window, acquisition or non-equity structures would move it lower. |
| During 2027: OpenAI financing or IPO developments | A new NVIDIA subscription—not simply a round total, listing plan or payment under an old agreement. No firm IPO date within the window was established. | Confirmed new participation would move it higher; delay beyond the signing window would reduce that upside. |
| Through December 31, 2027: retrospective filings | Later filings may reveal previously unknown amounts and execution dates. | Verified in-window commitments would move the assessed subtotal higher; evidence of pre-window signing would move it lower. Either would narrow uncertainty. |
The forecast is worth rerunning after an explicit investment-policy change, a permanent successor with a clear mandate, or a confirmed large new subscription. These would resolve major uncertainties rather than merely add another headline.
Research and methodology
The research assembled a targeted historical deal record and current financing pipeline, separating agreement dates from announcements, closings and payments. It assessed recipient independence, instrument type, NVIDIA’s attributable amount and the likelihood of eventual public verification.
A judgmental scenario distribution was used because transactions are sparse and highly concentrated. Historical deals informed possible scale, not a mechanically extrapolated monthly rate. The scenarios incorporate approval, timing, eligibility and disclosure risk; they are not statistically fitted frequencies.
The sources with the greatest bearing were:
- NVIDIA regulatory filings: the July 26, 2026 quarterly report distinguished purchases, outstanding commitments and funding capacity. The November 12 leadership-transition filing established the change in authority without announcing an investment cap.
- Recipient registration statements: Anthropic’s prospectus and Nscale’s prospectus supplied the most consequential evidence that upcoming closings concerned earlier obligations.
- Capital-allocation evidence: Kress’s August 26 earnings-call remarks, the Reuters profile and reported employee communications informed the balance between tighter approval and continuity.
- Large-transaction precedents: OpenAI’s February 2026 announcement informed transaction scale. The September 2025 letter-of-intent announcement illustrated why intended investment is not equivalent to a binding subscription.
- Pipeline reporting: the Thinking Machines/Figure update, Firmus reporting, Perplexity reporting and Reflection reporting identified opportunities while exposing unresolved signing and transaction-structure questions.
Limitations
The most important unavailable information was the board’s actual budget, approval thresholds and policy for new subscriptions during the successor search.
Executed agreements were not obtained for Thinking Machines, Figure, Firmus or Reflection, and no new NVIDIA allocation in OpenAI’s current financing or prospective IPO was established. Thinking Machines’ reported pre-window status relies on a secondary account of reporting whose original text was not fully accessible. That conclusion is less secure than the filing-based Anthropic and Nscale exclusions.
The historical record was targeted rather than a complete census. It supports the conclusion that commitments are lumpy, but not precise estimates of future signing frequency.
Forecast history
Hover a data point to trace its series, or click to view the forecast generated at that time.
Signed forecast receipt
Signed Nov 13, 2026, 12:31 PM with ed25519 key preseen-prod-ed25519-20260523 and externally timestamped Nov 13, 2026, 12:31 PM.
sha256:e7047fb7d90336...3a5a0d8316Want forecasts like this?
Preseen researches your questions and answers them with calibrated, source-backed forecasts.