Why did Jensen Huang hit the "investment brake": The underlying risks behind the subtle changes in the relationship between NVIDIA and OpenAI

The market once cheered for a trillion-dollar partnership, but Jensen Huang’s recent comments have shattered that dream. The NVIDIA CEO clarified in Taipei that the so-called $100 billion investment commitment is just an “invitation,” not a legally binding formal promise. This statement has prompted the industry to reevaluate NVIDIA’s seemingly close business relationship with OpenAI. What underlying business considerations might be driving Huang’s change in tone?

From “Golden Boy” to “Gradual Evaluation”: Shifting Investment Attitudes

According to The Wall Street Journal, Huang’s remarks directly countered market speculation that the deal had already been approved. The previously planned 10-gigawatt computing infrastructure has now become a “gradual evaluation” approach. The clear signal: NVIDIA is reassessing the value and risks of this investment.

Sources familiar with the matter reveal internal doubts within NVIDIA about the deal terms. Huang privately criticized OpenAI for “lack of discipline” in business. This isn’t just Huang’s personal view. Last year’s leaked information showed Microsoft CFO Amy Hood expressed similar concerns, believing OpenAI’s demand for computing resources and financial planning were unrealistic. These voices from investors indicate that OpenAI’s business model is beginning to unsettle its main supporters.

OpenAI’s Funding Dilemma: Why Equity Financing Could Be a Hidden Risk

On the surface, OpenAI’s choice of equity financing to swap for compute leasing seems routine, but from a risk management perspective, it introduces multiple hidden dangers.

In traditional debt financing, creditors hold the highest priority for repayment. Even if the company faces operational crises, creditors can protect their interests through senior claims. But OpenAI’s approach is entirely different—it has brought suppliers like NVIDIA and Oracle into the riskiest shareholder positions. This means that if OpenAI encounters liquidity issues or cash flow disruptions, these supplier-shareholders can’t simply walk away like debt holders; they may be forced to share risks passively to protect their initial investments.

More troubling is the ambiguity around ownership stakes. NVIDIA invests in OpenAI, which uses that money to buy NVIDIA chips, creating a closed financial loop. While accounting standards may allow such related-party transactions to go uneliminated within certain limits, in practice, this raises reasonable doubts among investors: Is market demand real? Or artificially inflated?

The data is startling: OpenAI currently bears commitments for up to $1.4 trillion in compute—more than 100 times its revenue forecast last year. Under such enormous financial leverage, any risk is magnified exponentially.

Huang’s Genuine Concerns: The Changing Competitive Landscape

Another key reason for Huang’s shift is the subtle change in the AI chip market landscape. Google’s Gemini and Anthropic’s Claude are steadily eroding ChatGPT’s market share, and competitors are no longer solely relying on NVIDIA. Anthropic heavily depends on Amazon-designed Trainium chips and Google’s TPU for training models, directly challenging NVIDIA’s long-standing GPU market dominance.

If OpenAI falls behind in this fierce competition, NVIDIA—its main creditor, investor, and supplier—will face direct sales impacts. This isn’t just about an investment; it’s about the entire business ecosystem’s risk assessment. Huang’s decision to slow down investments essentially aims to hedge against a potential “problem asset.”

NVIDIA’s Real Situation: Supply Constraints Are the Key

However, it’s worth noting that although Huang announced a reduction in OpenAI investment commitments, it doesn’t mean NVIDIA will completely withdraw. Current reports indicate NVIDIA still participated in this funding round, roughly around $20 billion. This may be more about adjusting the framework rather than pulling out entirely.

Arguments that NVIDIA will suffer setbacks need a more measured analysis. While NVIDIA’s business isn’t highly diversified, its data center segment serves a broad client base—including Microsoft, Google, Oracle, Amazon, and rapidly emerging sovereign AI clients. A reduction in OpenAI orders might slow growth but is unlikely to reverse the already steep growth curve.

More critically, evidence from TSMC CEO Wei Zhejia’s latest research, NVIDIA’s pursuit of capacity allocation at TSMC, and mainstream investment banks’ consensus all point to the same fact: NVIDIA’s bottleneck is supply shortage, not demand weakness. In other words, regardless of whether the OpenAI order materializes, NVIDIA’s GPU capacity is likely to be quickly absorbed by the market.

The Fracture in Business Tacit Agreements: Chipmakers’ Strategic Moves

A detail worth noting: after OpenAI reached an investment intention with NVIDIA, it quickly established partnerships with AMD, Broadcom, and other competitors. While diversifying supply risks is wise, this may undermine some implicit business understanding. As OpenAI seeks alternative chip solutions, NVIDIA and Huang may feel marginalized—possibly a trigger for Huang’s attitude shift.

This market shift reflects a new phase in the tech industry’s investment and supply relationships. Huang’s comments are less about “cutting ties” and more about a rational adjustment in a more complex business environment—balancing support for key clients while avoiding excessive risk exposure.