When $100 Billion Isn’t Enough, Nvidia Rethinks OpenAI

Last fall, the optics were perfect. Cameras, applause, big numbers, bigger adjectives. At Nvidia headquarters in Santa Clara, Jensen Huang stood with Sam Altman and Greg Brockman to unveil what was framed as a historic AI infrastructure partnership, a plan to build at least 10 gigawatts of Nvidia-powered compute for OpenAI, plus a headline-grabbing “up to $100 billion” investment path tied to deployment milestones.

Four months later, the story has changed from “largest computing project in history” to “still talking.” Multiple reports say the deal has not advanced beyond early-stage paperwork and internal hesitation inside Nvidia has effectively frozen the mega-commitment.

In the bluntest translation possible, this is what happens when a $100 billion plan meets a CFO’s calculator.

What Actually Froze?

According to reporting that traces back to The Wall Street Journal, Nvidia’s internal doubts have kept the partnership from moving past its initial, nonbinding framework. The same reporting says newer discussions have shifted toward something smaller, think “tens of billions” rather than “$100 billion.”

That detail matters because the original announcement was often treated like a done deal. Nvidia, for its part, had already been careful in public filings to describe the arrangement as nonbinding and uncertain, even while the market celebrated.

So yes, the party happened. The contract did not.

Why Nvidia Is Nervous

You do not stall a partnership like this unless the risk math starts feeling… personal.

Reporting indicates Huang has privately stressed that the original agreement was not finalized and has questioned OpenAI’s “commercial discipline,” while also pointing to rising competitive pressure in the AI model market.

And even if you ignore the spicy phrasing, Nvidia’s concerns are easy to understand when you look at the broader strategic landscape:

• If OpenAI keeps winning, Nvidia keeps shipping GPUs at scale.
• If OpenAI loses ground to competitors who train on non-Nvidia silicon, Nvidia loses not just a customer, but momentum.

That second scenario is no longer theoretical. Competitors have credible alternative compute stacks, including Google’s TPU ecosystem and cloud-native accelerators used by rivals.

In other words, Nvidia is weighing whether it should be a supplier with leverage or a sponsor with exposure.

The Market Pressure OpenAI Cannot Ignore

While Nvidia is hesitating, the competitive story around OpenAI has gotten louder.

Similar web tracking referenced in multiple recent analyses points to a meaningful shift in traffic share, with ChatGPT’s dominance narrowing as Google’s Gemini gains. Several reports also cite average daily visits decline for ChatGPT over recent weeks in the low double digits, alongside Gemini growth.

Traffic is not revenue, but it is sentiment with a pulse. And nothing spooks infrastructure partners like the feeling that the world’s most expensive machine might be powering someone else’s product lead next year.

A Quick Pointer Paragraph, What This Means Right Now

• Nvidia still wants OpenAI as a customer, but seems less eager to sign up as the world’s largest financial backer without tighter terms.
• OpenAI still needs enormous compute and capital, which is why it is simultaneously courting multiple mega-investors even while the Nvidia headline cools.

The Side Plot, Everyone Else Is Circling

The moment a “$100 billion” deal wobbles, the rest of Big Tech hears the dinner bell.

Amazon has been reported to be in discussions to invest up to $50 billion in OpenAI, with Andy Jassy involved in talks with Altman, according to reporting picked up by Reuters and others.

Yes, that is the same Amazon that has also backed OpenAI rival Anthropic. If you are thinking, “That’s hedging,” congratulations, you understand modern strategy.

SoftBank Group has also been repeatedly mentioned in coverage of OpenAI’s broader fundraising ambitions, part of a wider effort that some outlets describe as aiming for up to $100 billion at a valuation around $830 billion.

It is not that investors suddenly love writing giant checks. It is that training frontier models is now a sport where the entry fee looks like a national budget.

The Joke Nobody Wants to Tell Out Loud

The AI boom has produced a weird new genre of headline: “Company raises more money than some countries, still unprofitable, planning IPO anyway.”

And OpenAI is not alone in that genre. What makes this moment notable is that the hardware king, Nvidia, is effectively saying: we will keep selling you the shovels, but we are not sure we want to buy the whole gold mine with you.

That is not a breakup. It is a prenup conversation that started late.

Where This Likely Goes Next

Based on the reporting so far, there are three plausible paths:

1) The mega-deal shrinks into a structured investment.
Instead of “up to $100 billion,” Nvidia commits a smaller equity amount tied to performance milestones, governance rights, or purchase commitments. Recent reporting already points to conversations in the “tens of billions” range.

2) Nvidia stays close, but mostly as a supplier.
This is the quiet outcome. Nvidia sells an ocean of GPUs, keeps the relationship, but avoids writing a check large enough to become part of the story.

3) OpenAI raises the money elsewhere and Nvidia loses influence.
If Amazon and others step in with massive funding, Nvidia’s leverage could weaken, unless OpenAI still commits its core training stack to Nvidia long-term.

None of these outcomes mean OpenAI is “finished.” They do mean the market is maturing into something less myth-driven and more spreadsheet-driven.

Bottom Line

The Nvidia OpenAI mega-plan was supposed to symbolize inevitability, unlimited compute, unlimited ambition, and a near-straight line to the next era of AI.

What it symbolizes now is something more realistic: even in AI, even in 2026, even with trillion-dollar dreams floating around, a $100 billion commitment still has to survive internal scrutiny, competitive risk, and the basic question every serious partner eventually asks:

“Show me how these turns into a business, not just a moment.”

And if you listen closely, you can hear the entire industry learning the same lesson, one frozen megadeal at a time.