Nvidia wanted OpenRouter, and it moved too slowly to get it. Company executives told the startup’s team that Nvidia intended to make a generous offer but needed more time to assess the deal. OpenRouter’s founder was not willing to wait. Nvidia never put a formal bid on the table, and Stripe won the company with an $8 billion offer, according to a person familiar with the talks.
The near-miss tells one part of the story. The rest is what Nvidia’s deal team did next. Over the following two months it locked in transactions worth more than $140 billion in total, including a $105 billion credit guarantee. By the end of July, Nvidia held nearly $100 billion in equity investments and another $25 billion in future commitments, with more deals still possible.
The pivot is the clearest sign yet of how Nvidia is spending the extraordinary cash its chips generate. The company has become a financier as well as a supplier, using its money and its stock to buy positions across the AI ecosystem, from models and software to robots and the startups that build on top of its hardware.
OpenRouter operates a marketplace that routes developers to different AI models, a piece of the software layer that sits above the chipmaker’s hardware. Owning it would have given Nvidia a direct view of which models developers actually use, data that could inform everything from chip design to pricing. That is presumably why the deal team wanted it, and why its loss stung enough to send the team hunting elsewhere.
One of the newer discussions involves Figure, the humanoid-robot maker. Nvidia has talked about adding $1 billion to its investment in the company, whose latest round priced it at about $38 billion pre-money, a person familiar with the matter said. The deal, which had not been previously reported, would deepen Nvidia’s stake in a category it has repeatedly called the next big computing platform.
Robots are only one front. Nvidia has also agreed to pay about $6 billion for a software license from Poolside and to absorb Poolside’s team, a transaction driven in part by a desire to advance Nemotron, Nvidia’s own open-source model. The deal shows the company buying its way into model development rather than building every capability in-house.
Nvidia’s response to losing OpenRouter was to speed up elsewhere. The $140 billion of activity over two months suggests a deal team that learned from being outbid and resolved not to let the next target slip away. The pace is remarkable even for a company of Nvidia’s size, and it has turned the chipmaker into one of the most active technology investors in the world.
OpenRouter had built a fast-growing marketplace at the moment AI developers were realizing they wanted to switch between models rather than commit to one. Stripe’s $8 billion purchase fit its own push beyond payments into the tools developers use to run their businesses. For Nvidia, the appeal was different: a view into which models developers actually choose, a signal the chipmaker could use to steer its own road map.
The Figure investment is part of a longer Nvidia push into robotics. The company has backed humanoid-robot makers as a bet that the next wave of computing will be physical, and its chips and simulation tools are already the default for many of them. Adding $1 billion to a company valued at $38 billion before the round is a way to stay close to a leader without buying it outright.
The $105 billion credit guarantee, the largest single item in the two-month total, turns Nvidia from an investor into a backstop. Guaranteeing that much credit signals the company is willing to put its own balance sheet behind the ecosystem’s buildout, a step few chipmakers have taken. The Poolside deal points the same direction: paying $6 billion for software and a team to advance Nemotron is a bet that Nvidia should own a model, not just the hardware that runs everyone else’s.
Analysts said the spending has a strategic logic that goes beyond financial returns. Every dollar Nvidia puts into a model developer, a robot maker or an infrastructure startup is a dollar that strengthens the ecosystem around its chips, which in turn supports the demand that drives its core business. The investments are, in effect, a way to buy demand for the products Nvidia already sells.
The risk is concentration. Nvidia’s equity book is now heavily exposed to the same AI boom that drives its chip sales, which means a downturn in the sector would hit the company twice, once in its operating results and once in its investment portfolio. Investors who once saw Nvidia as a pure play on chips now have to reckon with a sprawling balance sheet of AI bets.
For now, the company shows no sign of slowing. The $25 billion in future commitments and the talks still under way suggest the deal machine is still running. OpenRouter got away, but the larger question is whether Nvidia’s spending can buy the ecosystem it wants without becoming the ecosystem’s biggest single point of failure.
