By June 30, 2027, will NVIDIA and CoreWeave terminate or renegotiate their existing capacity-purchase agreement to reduce NVIDIA’s maximum purchase obligation?
Latest forecast
Generated Nov 13, 2026, 11:23 AM · High effort
Write-up
Executive Summary
The probability that the question resolves YES is 9.2%. The most likely outcome is continuity: no publicly established contractual reduction meeting the question’s threshold within the specified window.
YES requires an affirmative reduction of at least 10% in NVIDIA’s maximum remaining legal purchase obligation, effective from November 13, 2026 through June 30, 2027, and substantiated publicly by December 31, 2027. Lower actual purchases, greater third-party usage and ordinary fulfillment do not count.
The strongest YES pathway is a negotiated exchange: NVIDIA reduces its capacity-purchase obligation while providing CoreWeave with equity, lending or other non-capacity support. That could preserve the partnership while satisfying the question. No contract-specific negotiating mandate or release proposal was verified in the research.
Key drivers
1. Signed obligations are harder to reduce than new commitments
The covered arrangement is the September 9, 2025 order under the April 10, 2023 master services agreement, or MSA. It requires NVIDIA to purchase residual unsold capacity through April 13, 2032, subject to contractual conditions.
The public MSA permits termination for uncured material breach and qualifying insolvency proceedings; no general right to cancel for convenience is visible. However, the applicable order takes precedence over the MSA, and that order’s complete terms were not publicly established. The evidence therefore supports meaningful exit friction—not an absolute inability to exit.
The original $6.3 billion initial value is not a verified remaining-obligation baseline and cannot automatically determine the reduction threshold.
2. NVIDIA’s leadership transition creates scrutiny, but the official position favors continuity
Jensen Huang’s resignation became effective on November 11, 2026, before the qualifying window. Colette Kress became interim CEO, and NVIDIA’s board explicitly expected customer commitments to continue without interruption.
Reuters reported that Kress and her team had challenged large financing transactions involving other companies. That supports tighter scrutiny, but does not establish that this CoreWeave order is targeted. Kress had also publicly defended NVIDIA’s cloud-capacity financing model. Restricting new commitments is consequently better supported than unwinding this signed agreement.
3. The partnership and existing resale mechanics favor leaving the contract intact
NVIDIA’s completed $2 billion equity investment and expanded January 2026 collaboration demonstrate a relationship extending beyond one executive. CoreWeave management also described strong demand for capacity, although its demand claims were not independently verified.
Strong demand does not necessarily create a reason to amend the agreement. NVIDIA’s broader cloud disclosures describe commitments declining as capacity is used by NVIDIA or third parties. Those aggregate disclosures are not a reconciliation of this order, but they illustrate how economic exposure can decline under unchanged terms—without producing YES.
4. CoreWeave’s financial pressure cuts both ways
CoreWeave’s September 30, 2026 filing reported $47.505 billion of scheduled debt principal and $9.837 billion of liquidity. Only $4.184 billion was cash and marketable securities; the remainder was availability under financing facilities. The company reported compliance with its loan conditions.
Financial pressure creates a possible restructuring pathway. It also makes NVIDIA’s promise to buy unsold capacity more valuable to CoreWeave. Surrendering that protection would likely require acceptable compensation, rather than a simple request to reduce exposure.
5. Public proof is a separate hurdle
A private reduction is insufficient. Public evidence must establish its legal effect, effective date and threshold, after accounting for replacement capacity obligations. Material amendments have disclosure routes, but publication of every economic term needed for this question is not guaranteed. The later evidence cutoff helps, while confidential schedules remain a meaningful obstacle.
Scenarios
The forecast weighs four principal paths. Their relative importance reflects judgment about incentives, contractual constraints and disclosure—not measured historical cancellation rates.
| Scenario | Relative weight and rationale | How it leads to an outcome |
|---|---|---|
| Continuity; tighter controls mainly affect new deals | Dominant. Supported by the board’s continuity statement, the wider partnership and existing mechanisms that reduce exposure as capacity is used. | The agreement remains intact. Actual purchases may fall, but that alone does not qualify. |
| Active review of existing support | The main minority scenario and strongest YES route. Leadership change creates an opening, but CoreWeave must receive adequate value for relinquishing protection. | A binding release, lower ceiling or shorter coverage could qualify, potentially exchanged for equity or lending. Equivalent replacement capacity obligations would prevent a qualifying net reduction. |
| Acute financial or delivery stress | Smaller. CoreWeave’s debt burden supports a stress pathway, but covenant compliance and expected customer funding argue against assuming imminent insolvency. | A negotiated settlement or valid termination could reduce remaining obligations. Automatic operation of unchanged availability conditions would not count. Bankruptcy also does not automatically permit cancellation: federal law generally restricts termination solely because proceedings begin, subject to unresolved exceptions. |
| Regulatory intervention or corporate restructuring | Smallest. Political scrutiny exists, but no verified directive specifically required reducing this order. | A binding remedy or restructuring could qualify if it produces a timely, sufficiently documented net reduction. Disclosure reforms or restrictions on future deals would not. |
Across these paths, an actual qualifying reduction and sufficient public substantiation are assessed separately.
Implications
- Continuity is the appropriate planning assumption, not a guarantee of cash receipts. The forecast concerns the legal ceiling, not NVIDIA’s expected purchases or CoreWeave’s overall financial health.
- A YES outcome need not mean NVIDIA abandons CoreWeave. Replacing capacity support with equity or lending could preserve the commercial relationship while changing who bears capacity risk.
- A NO outcome would not prove that nothing changed privately. It includes situations where terms or amounts remain insufficiently disclosed. Aggregate commitment balances and cancellation headlines should therefore not be treated as decisive evidence.
Signs to watch
| Event or development | What would matter and forecast direction |
|---|---|
| November 17, 2026: NVIDIA’s scheduled fiscal third-quarter results and earnings call | An explicit review of already-signed capacity support would make YES more likely. Reaffirming existing commitments while limiting only new deals would make it less likely. |
| Customer funding over the three months following CoreWeave’s November 10 results | Management projected approximately $9 billion of prepayments and deposits. Receipt would weaken the acute-stress pathway; a material shortfall would strengthen it, although financial stress alone would not establish a qualifying reduction. |
| An unannounced support-swap proposal or binding amendment | A reduction exchanged for equity, lending or other non-capacity support would make YES more likely. An equivalent or larger replacement capacity obligation would not qualify. A binding agreement would warrant rerunning the forecast. |
| Formal performance disputes, termination notices or financing deterioration | A credible breach followed by an affirmative cancellation or settlement would make YES more likely. Improved funding conditions would weaken the stress pathway. A bankruptcy filing alone would not settle the legal question. |
| Publication of the operative order, amendments or lender-consent terms | A workable exit mechanism would make YES more likely; costly release terms or proven consent barriers would make it less likely. These documents could settle major uncertainties and would warrant rerunning the forecast. |
| A contract-specific regulatory directive | Binding action reaching this existing order would make YES more likely. General demands for transparency are weaker signals and do not themselves qualify. |
| June 30, 2027 effective-date deadline; December 31, 2027 evidence cutoff | A timely reduction may be proved later, but an agreement effective after June 30 cannot qualify. Subsequent filings establishing a timely threshold-meeting change would favor YES; insufficient public proof by the evidence cutoff means NO. Both deadlines use America/Los_Angeles time. |
Research and methodology
The research centered on primary contractual and financial documents, supplemented by official statements and attributed reporting:
- Contractual baseline: CoreWeave’s original September 2025 disclosure established the announced arrangement. The executed MSA supplied termination, payment-release and order-precedence provisions. Later filings were reviewed for identified amendments.
- Leadership and intent: NVIDIA’s November 2026 filing and official transition announcement established the leadership change, continuity statement and scheduled earnings date. Reuters reporting on Kress informed the possibility of greater financial scrutiny.
- Commercial incentives and financing: The January collaboration announcement, CoreWeave’s September 2026 quarterly filing and November earnings release informed partnership value and financial pressure. NVIDIA’s July 2026 quarterly filing clarified aggregate capacity-commitment mechanics.
- Legal and disclosure constraints: 11 U.S.C. §365, the August 2026 credit agreement and SEC Form 8-K requirements informed exit feasibility, possible financing restrictions and public-proof requirements.
- Historical comparisons: WhiteFiber’s annual report showed that capacity allocations can be amended, but also that early service termination can preserve payment of remaining contract value. These selected examples illustrate mechanisms, not a representative cancellation rate.
The quantitative framework used mutually exclusive dominant-driver scenarios, separating the chance of a timely threshold-meeting legal reduction from the chance of sufficient public evidence. The assumptions are judgmental rather than statistically estimated frequencies.
Limitations
- The operative order, complete amendment history, current maximum remaining obligation and full buyout or wind-down economics were not publicly established. These omissions affect both exit feasibility and the correct percentage denominator.
- Some financing agreements restrict adverse changes to material project contracts, but the research could not establish that this NVIDIA order is covered. Lender consent is possible friction, not a proven veto.
- No verified contract-specific negotiating mandate or release proposal was found. That limits visibility into private intent; it does not establish that private discussions are absent.
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, 11:23 AM with ed25519 key preseen-prod-ed25519-20260523 and externally timestamped Nov 13, 2026, 11:23 AM.
sha256:cdf5884a30fe54...154bafd489Want forecasts like this?
Preseen researches your questions and answers them with calibrated, source-backed forecasts.