David Brooks
My verdict is clear: the biggest deal in the history of artificial intelligence just fell apart, and the warning signs were there from the start.
On January 30, 2026, the Wall Street Journal revealed that Nvidia has frozen its plan to invest $100 billion in OpenAI. What was presented in September 2025 as a historic alliance between the world's largest AI chip maker and the company behind ChatGPT is now a phantom deal that was never actually signed.
But the story gets complicated. Just 24 hours later, Jensen Huang came out to partially deny the report, stating that Nvidia plans to make a "huge" investment in OpenAI, "probably the largest ever made." Who's telling the truth? Nvidia's own regulatory filings tell a very different story from its CEO's public statements.
What the original deal was and why it was never signed
In September 2025, Nvidia and OpenAI announced a letter of intent (LOI) for up to $100 billion. The plan: Nvidia would invest equity in OpenAI in exchange for OpenAI deploying infrastructure exclusively with Nvidia chips. The goal was to deploy 10 gigawatts of compute capacity, equivalent to the peak electricity demand of the entire city of New York.
Jensen Huang called it "the next leap forward." Sam Altman declared that "compute infrastructure will be the basis for the economy of the future." The first $10 billion tranche was tied to signing a definitive purchase agreement for the first gigawatt, using the Nvidia Vera Rubin platform.
The problem: it was never binding. It was a letter of intent, not a contract. And Nvidia knew it better than anyone.
The warning signs the market ignored
| Date | Signal | Source |
|---|---|---|
| September 2025 | LOI announced for $100B | Nvidia Newsroom |
| November 2025 | Nvidia declares to the SEC: "There is no assurance that we will enter into definitive agreements" | SEC Filing |
| December 2025 | CFO Colette Kress confirms: "We still haven't completed a definitive agreement" | UBS Conference |
| January 2026 | WSJ reports the deal has been frozen | Wall Street Journal |
Nvidia's CFO went further in December: she confirmed that the $500 billion demand guidance for Blackwell/Vera Rubin in 2025-2026 "does not include anything from the work we're doing with OpenAI." In other words, Nvidia never counted on that money.
Why Jensen Huang hit the brakes: three reasons that change everything
According to WSJ sources, Nvidia's reasons for freezing the deal are multiple:
1. "Lack of discipline" at OpenAI
Huang has privately expressed concern about OpenAI's business approach. And the numbers back him up:
| OpenAI Metric | Figure |
|---|---|
| 2025 Revenue | $20B (+233% year-over-year) |
| Projected 2026 Losses | $17B |
| Total funding raised | $57.9B (9 rounds, 62 investors) |
| Target valuation | $830B |
| Planned IPO | Q4 2026 (~$1 trillion) |
OpenAI generates $20 billion a year but expects to lose $17 billion in 2026. For an engineer like Huang, obsessed with efficiency, these numbers are hard to swallow.
2. The competition threatening OpenAI
Huang has emphasized to industry associates his concern about the competition facing OpenAI:
- Google: Develops its own TPUs, reducing Nvidia dependence, and competes directly with Gemini
- Anthropic: Valued at $183 billion, backed by Amazon ($8B) and Google
- DeepSeek: Trains comparable models at a fraction of the cost
Investing $100 billion in a single company when the market is fragmenting is a risky bet, even for Nvidia.
3. The ghost of circular financing
This is the point that concerns regulators the most. The scheme works like this:
Nvidia invests $100B in OpenAI → OpenAI uses that money to buy Nvidia chips → Nvidia reports record revenue → Nvidia stock rises → Nvidia can invest more
Economist Noah Smith compared it to Cisco's "vendor financing" during the 90s dot-com bubble, a model that ended in disaster. The FTC has already launched an investigation into generative AI investments and partnerships, and the DOJ supervises Nvidia for its 90%+ market share in data center GPUs.
CoreWeave: the bet Nvidia actually signed
While the OpenAI deal cooled, Nvidia made concrete moves. On January 26, 2026 (4 days before the WSJ report), it invested an additional $2 billion in CoreWeave, a cloud provider specializing in AI.
| Nvidia in CoreWeave | Figure |
|---|---|
| New investment (January 2026) | $2B |
| Previous cumulative investment | $3.3B |
| Total invested | ~$5.3B |
| Position | 2nd largest shareholder |
| Goal | 5+ GW of AI capacity by 2030 |
The logic is clear: CoreWeave is a "neocloud" that buys Nvidia chips massively and resells them as cloud services. Its contracts include $22.4 billion with OpenAI and $14.2 billion with Meta. For Nvidia, investing in CoreWeave means investing in its own ecosystem without the regulatory risks of directly financing a customer.
Amazon negotiates to fill Nvidia's seat
The vacuum left by Nvidia has suitors. Amazon is negotiating to invest up to $50 billion in OpenAI, according to CNBC. Sam Altman and Andy Jassy are directly negotiating a "chips-for-equity" model: Amazon contributes Trainium, Inferentia chips and AWS services in exchange for OpenAI equity.
| Investors in OpenAI's $100B round | Amount |
|---|---|
| Amazon | Up to $50B (negotiating) |
| SoftBank | ~$30B additional |
| Nvidia | "Tens of billions" (reduced) |
| Microsoft | Maintains existing 27% |
Amazon already has $8 billion invested in Anthropic (OpenAI's direct competitor). Investing simultaneously on both sides of the AI war is a strategy that echoes Microsoft's massive spending that cost it $357 billion in a single day.
AMD: the signed deal Nvidia wouldn't close
While Nvidia was negotiating letters of intent, AMD signed a real deal with OpenAI on October 6, 2025:
- 6 gigawatts of AMD Instinct GPUs for OpenAI
- OpenAI received warrants for 160 million AMD shares (potentially 10% of the company) at $0.01 each
- First MI450 chip deployment scheduled for the second half of 2026
- Broadcom also signed a $10 billion deal to develop custom AI chips for OpenAI
The difference is stark: AMD and Broadcom have signed agreements. Nvidia has a letter of intent that never materialized.
The 24-hour reversal: Huang denies... sort of
On January 31, just one day after the WSJ article, Jensen Huang publicly declared that Nvidia plans to make a "huge" investment in OpenAI, "probably the largest ever made," and denied being dissatisfied.
But Huang's words contradict his own company's regulatory filings:
- The November SEC filing says "there is no assurance"
- The CFO said in December "there is no definitive agreement"
- Nvidia's financial guidance excludes any OpenAI revenue
Who's right? My verdict: regulatory filings carry more weight than press statements. Nvidia is renegotiating, not freezing entirely, but the original $100 billion deal is dead.
What this means for the future of AI
For OpenAI
It needs to close the $100 billion round urgently to fund Project Stargate and prepare its Q4 2026 IPO. Amazon could replace Nvidia as the primary strategic investor, which would shift the balance of power in the industry.
For Nvidia
It maintains its dominant position as a supplier (>90% of the data center GPU market). It prefers investing in "neoclouds" like CoreWeave rather than a single customer. And it avoids antitrust issues by not tying itself exclusively to OpenAI.
For the ecosystem
Circular mega-deals are under growing regulatory scrutiny. The AI valuation bubble ($830B for OpenAI, $183B for Anthropic) depends on these deals materializing. If they don't, market corrections could be brutal.
Frequently asked questions
Has Nvidia completely canceled its investment in OpenAI?
Not entirely. The original $100 billion deal is frozen, but Nvidia is renegotiating for "tens of billions." Jensen Huang declared on January 31 that he plans a "huge" investment. However, Nvidia's regulatory filings indicate no definitive agreement has been signed.
Why is Jensen Huang criticizing OpenAI?
According to the WSJ, Huang has expressed concern about OpenAI's "lack of discipline" in business. With projected losses of $17 billion for 2026 despite $20 billion in revenue, and growing competition from Google and Anthropic, Huang questions whether OpenAI is the best bet for an investment of this magnitude.
Will Amazon replace Nvidia as OpenAI's main investor?
Amazon is negotiating to invest up to $50 billion using a "chips-for-equity" model (Trainium + AWS in exchange for stake). If finalized, it would be the single largest investment in OpenAI's $100 billion round, effectively displacing Nvidia.
What is circular financing and why does it matter?
It's a scheme where Nvidia invests in OpenAI, OpenAI buys Nvidia chips with that money, and Nvidia reports inflated revenue. The FTC is investigating these deals for their similarity to Cisco's "vendor financing" during the dot-com bubble.
How does this affect Nvidia's stock?
Surprisingly, very little. NVDA trades around $190 with a market cap of $4.65 trillion. The market understands that Nvidia sells chips to every player in the AI market, not just OpenAI.
