Nvidia Joined 105 Startup Rounds This Year. Then Its CEO Left.
Nvidia took part in 105 startup funding rounds in 2026, and investors say most of the big calls came from one person. He resigned on November 11. Signed deals are holding. Unsigned ones are in doubt. Here is what that teaches a first-time founder about counting on any single backer.

Nvidia took part in 105 startup funding rounds this year, according to PitchBook data cited by TechCrunch. It was often the largest check in the round.
On November 11, the person who approved most of those checks resigned as CEO.
This article isn't about why. Other outlets are covering that. It's about a quieter problem that hit hundreds of founders in the same hour: the investor they were counting on turned out to be one human being, and that human is no longer in the chair.
You probably aren't raising from Nvidia. The lesson still applies to you, maybe more than to them.
What changed at Nvidia this week?
Jensen Huang co-founded Nvidia and ran it for 33 years. DigitalToday, citing reporting from The Information, says he resigned on November 11. Chief financial officer Colette Kress is interim CEO and keeps her finance role.
For startups, the key detail is how Nvidia's investing worked. Umesh Padval, managing partner at Seligman Ventures, told TechCrunch that most major startup investment decisions came from Huang himself.
So this wasn't a fund with a committee and a ten-year mandate. It was closer to one person's conviction, backed by one of the largest balance sheets on earth.
Investors now expect a more careful approach. One unnamed investor said of Kress, "She's much more conservative." Per TechCrunch, the expectation is that Nvidia gets pickier about new companies and writes smaller follow-on checks to ones it already backs.
How much did startups lean on Nvidia?
A lot. Look at the size of the commitments.
DigitalToday reports Nvidia struck deals worth about $140 billion in roughly two months this year, counting large credit guarantees. TechCrunch notes it put $800 million into Reflection AI's $2 billion Series B last year, before that startup had meaningful revenue, citing the Wall Street Journal. It joined two rounds at Crusoe, including a $1.4 billion Series E.
The pattern was strongest in infrastructure. A company that needed $100 million to $500 million in chips would often raise half of that from Nvidia, at a valuation of $1 billion or more.
Shamillah Bankiya of Dawn Capital described the dependence this way: "Nvidia is carrying almost every AI-native startup from application to neocloud."
And the name did work beyond the cash. As Padval put it, "When Nvidia was on the cap table, everyone felt safe." Other investors followed. Lenders relaxed. Finance people even had a nickname for it, the "Nvidia put," meaning the belief that Nvidia would step in if things got rough. DigitalToday reports that executives now say that belief has largely disappeared.
Which deals are safe and which aren't?
This is the most useful part of the week's reporting, because the line is so clean.
Signed deals are holding. DigitalToday reports that Safe Superintelligence, with a $5 billion Nvidia investment, and Thinking Machines Lab, at about $2.5 billion, signed legally binding agreements months ago. People at both companies expect those to stand.
Unsigned deals are in doubt. The same report names three:
- Permus, an Australian cloud company seeking at least $2 billion after pulling its IPO. Nvidia discussed investing in late October.
- Enscale, a neocloud that talked with Nvidia about guaranteeing more than $5 billion in borrowing. Nothing was signed as of the end of October.
- Figure AI, the humanoid robot maker, which was in talks for another $1 billion. Status unconfirmed.
Same investor, same month, same enthusiasm in the room. The only difference is ink.
Every founder has heard "we're in" on a call and started spending in their head. Don't. A verbal yes is a mood. A signed document is an asset.
Your AI co-founder is ready when you are.
Foundra turns everything in this article into an actual plan. Validation, customers, pricing, launch. In one place, in your voice, in an afternoon.
Get started→$39/month. Cancel anytime.
Why does this matter if you're raising $500K?
Because the same risk shows up at every size. It just wears different clothes.
At the seed stage, your "Nvidia" might be:
- One angel who said they'd cover half the round and bring friends.
- One partner at a fund who loves you, while the other partners have never heard your name.
- One corporate champion at a big customer who promised a pilot and an investment.
- One accelerator director who waved you in.
People change jobs. They get sick, get fired, get overruled, or get busy with a crisis of their own. When your champion goes, the institution doesn't remember the promise. It remembers the paperwork.
There's a second trap. When a famous name commits, founders stop talking to everyone else. Why keep pitching when the round is "done"? That's how a single departure turns into a company-ending event. Shares of CoreWeave and Iren, both Nvidia-supported, fell about 7% on November 12 on this news alone, per DigitalToday. Public companies can absorb a bad day. A startup with eight weeks of cash can't.
How do you check your own exposure?
Run this once before your next raise and again whenever a big commitment lands.
- Count the decision-makers. For each investor, who else has to say yes? If the answer is "nobody," your deal depends on one person's calendar and health.
- Sort commitments into three piles. Signed and wired. Signed, not wired. Spoken only. Plan your spending on the first pile.
- Find the 30% line. If one backer supplies more than about 30% of your round, write down what happens if they vanish. Who fills the gap, and at what price?
- Ask what the name is doing for you. If other investors came in because of one logo, their conviction is borrowed. Talk to them directly about why they believe in the business.
- Keep two conversations warm. Even after a yes, send updates to the investors who passed or went quiet.
Then put numbers on it. Build one version of your financial plan where the big check arrives and one where it never does. A spreadsheet handles this, and so does a planning tool like Foundra, which lets first-time founders keep financial projections and scenarios in one place. The tool matters less than having the second version before you need it.
Is the money gone for good?
No, and it's worth keeping some perspective.
Few investors told TechCrunch that a slower Nvidia would stop the AI boom by itself. Capital is still being raised. One day after the resignation, Norwest announced a $4 billion fund, the largest in its 65-year history. Senior managing partner Jeff Crowe said the firm is especially interested in early-stage companies building AI-native applications and infrastructure.
Others may move into the space. Dharmesh Thakker of Battery Ventures expects big tech rivals to compete for Nvidia's role. He also doubts that Google, Amazon, Microsoft or AMD will put hundreds of millions into pre-revenue infrastructure startups. They're more likely to back mature companies.
So the likely result is narrower, not empty. Money for proven businesses. Less for bets that needed one believer with a very large checkbook.
What should you be careful about here?
Three things.
This story is days old. Nvidia hasn't announced a new investing policy, and much of the reporting rests on unnamed sources and expectations. Kress may surprise people. A permanent CEO may restart the dealmaking.
Strategic investors aren't bad. Nvidia's money built companies that wouldn't exist otherwise. Zain Javaid, a founder at hardware startup Ghost, credited those bets with letting an ecosystem form despite brutal costs. The mistake isn't taking a big backer's money. It's building a plan with no version that survives without them.
And diversifying has a cost. Ten small investors mean ten sets of updates and opinions. For a tiny round, two or three committed backers with signed documents can beat a crowd.
Key takeaways
- Nvidia joined 105 startup rounds in 2026, per PitchBook data cited by TechCrunch, and one person made most of the big calls.
- That person resigned on November 11. The interim CEO is expected to invest more cautiously.
- Signed agreements with Safe Superintelligence and Thinking Machines Lab are expected to hold. Unsigned talks with Permus, Enscale and Figure AI are in doubt.
- Key-person risk exists on the investor side too, at every round size.
- Plan spending on signed and wired money only.
- Keep a version of your financial plan that works without your largest backer.
Frequently asked questions
How many startups did Nvidia invest in this year? It took part in 105 funding rounds in 2026, according to PitchBook data cited by TechCrunch.
What is key-person risk? It's the danger that a deal or relationship depends on one individual. If that person leaves, the commitment can leave with them.
What is a strategic investor? A company that invests in startups partly for business reasons, such as selling them its products, and not only for financial return.
Are Nvidia's existing investments being cancelled? Reporting so far says legally binding agreements are expected to be honored. Deals that weren't signed are uncertain.
How much of a round should one investor supply? There's no fixed rule. Above roughly 30%, you should have a written backup plan in case that investor drops out.
Does a verbal commitment count? Not for planning. Treat money as real once documents are signed and the funds have arrived.
Sources
- Jensen Huang's departure blasts big hole in AI startup funding, VCs and founders say, TechCrunch, November 13, 2026
- Jensen Huang resignation throws Nvidia investment talks into doubt for Enscale and Permus, DigitalToday, November 13, 2026
- Jensen Huang's exit from NVIDIA could disrupt start-up ecosystem, NewsBytes, November 14, 2026
- Norwest pushes deeper into AI with new $4B fund, TechCrunch, November 12, 2026
You just read the theory. Ready to build the thing?
Foundra is your AI co-founder. It turns an idea into a validated business plan, a go-to-market, and your first 10 customers. In an afternoon, not a semester.
$39/month. Cancel anytime. Works in 20 languages.