Among the first to feel uneasy after Jensen Huang stepped down as Nvidia CEO were the AI startups that built their businesses on Nvidia's checks. TechCrunch reported on November 13, citing PitchBook data, that Nvidia has already taken part in 105 funding rounds this year, and in many of them it was the largest single investor.
“It's like a bomb went off,” one neocloud founder said, describing how the news landed.

Most of the checks were signed off by Huang himself
At Nvidia, which companies to back and how much to put in has largely been Huang's personal call. Umesh Padval, managing partner at Seligman Ventures, told TechCrunch that most of Nvidia's significant startup investment decisions came from Huang himself. Padval has known Huang for years and served on the board of Mellanox, which Nvidia later acquired for about $7 billion.
The person taking over has a different style. Nvidia's interim CEO is Colette Kress, previously its chief financial officer. One investor's assessment: “She's much more conservative,” and that she would not put up the same amount of money to hold up the early-stage ecosystem. These views come from anonymous investors. Nvidia has made no public statement on the pace of its investing, and whether Kress will actually tighten up cannot be verified at this point.
Nvidia's heaviest bets have gone to three kinds of companies: neocloud providers, inference providers and open-source model teams. Publicly known examples include an investment in CoreWeave before its IPO; participation in two Crusoe rounds, one of them a $1.4 billion Series E; the roughly $13 billion acquisition of Hugging Face this year, which is expected to close in 2027; and stakes in Mistral AI, Lambda and Black Forest Labs. The Wall Street Journal previously reported that Nvidia put $800 million into Reflection AI's $2 billion Series B last year, when the open-source model company had no meaningful revenue yet.
One venture investor described a common playbook: an infrastructure company that needs $100 million to $500 million of GPUs would often get half of the required amount from Nvidia first, at a valuation starting around $1 billion, and other investors only dared to follow once they saw Nvidia on the cap table. “When Nvidia was on the cap table, everyone felt safe,” Padval said.
Investors expect Nvidia to become pickier from here: fewer new deals, smaller follow-on checks for existing portfolio companies, and valuations coming down with them.
Can Google and Amazon fill the gap?
Some are hoping the big platforms will step in. Dharmesh Thakker, general partner at Battery Ventures, expects large tech companies to compete for the “investment crown” Nvidia leaves behind, with the contest decided by who has the better product, an ecosystem that can rival CUDA, and the financial resources to sustain it.
Thakker is also skeptical. Google, Amazon, Microsoft and AMD may back some more mature startups to promote their own chips and clouds, but he doubts they would pour hundreds of millions of dollars into pre-revenue infrastructure companies the way Nvidia has.
That is where the two investing logics differ. When Nvidia backs a neocloud, the money is used to buy Nvidia's cards and flows straight back. When a cloud provider backs the same kind of company, it is propping up its own competitor. TechCrunch also pointed to the flip side: Nvidia's generous checks have locked a group of companies onto its chips and strengthened its market position, so if it steps back, rivals gain room. Coincidentally, Google unveiled two new chips this very week.
On the open-source model side
In a post on November 14, AI2 researcher Nathan Lambert judged Huang's departure a major setback for the US open-source model ecosystem. Nvidia's top priority now, he wrote, is to steady the ship and deliver to customers on time over the next 24 months, and the money for Huang's kind of “founder-style” open-source bets will most likely dry up.
Lambert said Nvidia's own Nemotron team is ambitious and has good people, but is not yet at the point where it can declare success. The best-performing US open-weight model recently is Thinking Machines' Inkling 2 Small, which still trails Moonshot AI's Kimi K3.1 Pro. He cited SemiAnalysis CEO Dylan Patel on a podcast: “It's not money, it's talent, man.”
For open-source model teams in China, if Lambert's reading holds, the most free-spending player in the US open-source camp would be the first to let go, and the pace at which the US catches up with Kimi, DeepSeek and GLM could slow. Nvidia's acquisition of Hugging Face is still going through the process, and there is no public information on whether the platform's strategy will change after closing.
Sources: TechCrunch, Interconnects, CocoLoop, The Wall Street Journal. Round counts follow PitchBook's methodology; individual investment amounts are as publicly disclosed.