Nvidia Will Survive Jensen Huang. Your Cap Table Might Not.
Nvidia's order book and CUDA will outlast Jensen Huang. The AI startup valuations set by his price-insensitive checks won't, and your exposure runs well beyond the cap table.
OPC Community
OPC Community

The stock lost $546 billion in a morning. The bigger loss has no ticker. It's buried in private rounds that quietly assumed one man would keep saying yes.
Nvidia shed roughly $546 billion of market value at Thursday's open, a 14% drop on the news that Jensen Huang was out after 33 years. That was the cheap part of the week.
The expensive part is every private AI round priced since 2023 on the assumption that Nvidia would show up in it.
The coverage has gone exactly where you'd expect. Who succeeds him? Can the board, which voted to remove him amid a personal scandal still under investigation, hold its nerve? Does the "Stand with Jensen" petition, now near 180,000 signatures with Elon Musk's among them, change anything? Dan Ives says there is "no chance" he returns. Treasury Secretary Scott Bessent wants an outside hire.
Every one of those is a story about Nvidia. And Nvidia is the one company in this mess that's going to be fine.
Nvidia Has an Order Book. You Have a Logo Slide.
Steve Blank published the sober version on Sunday. His read is that the order book and CUDA carry Nvidia for two or three years no matter who sits in the chair. The chips are sold. The software lock-in took fifteen years to build, and it won't dissolve in a quarter.
He's right. That's exactly why founders should feel worse.
Nvidia's strength sits on things that outlast a person: contracts, fabs, a developer base. The startups orbiting Nvidia were sitting on something with no shelf life beyond the man himself. They were sitting on Jensen's conviction.
Blank's other detail explains how that happened. Huang ran the company with about 60 direct reports, discouraged one-on-ones, and anointed no successor. Founders told Business Insider last year that he personally approved every startup investment. PitchBook counted 67 Nvidia venture deals in 2025, about one every five days, and that count leaves out NVentures, its formal venture arm.
Most of that pace ran on one person's judgment.
You can inherit an investment committee. You can inherit a written thesis. What you can't inherit is a founder-CEO's gut call, made somewhere between a keynote and a supply meeting, that this robotics company or that neocloud deserves $200 million and priority allocation. Nobody wrote it down, so there's nothing to hand the next CEO.
The Only Investor Who Didn't Care About Price
Here's the thing I think most people half-sensed and never said out loud. Nvidia was never really a venture investor. It was running a customer-acquisition program that happened to pay in equity.
A VC needs the equity itself to return the money. Nvidia needed the startup to buy GPUs. When it put money into a model lab or a neocloud, a large share of that money came straight back as chip revenue, at gross margins north of 70%. The equity was a bonus. The stake could go to zero and the deal could still have paid for itself.
So Nvidia was the one large investor in AI for whom valuation barely mattered. And a buyer who doesn't care about price is the buyer who sets it.
Look at how the rounds actually worked. Nvidia comes in. The valuation lands at a number a pure financial investor would have fought. The financial investors follow anyway, because Nvidia's presence implied three things at once: the company would get chips, it would get more money later, and the smartest person in the industry had looked at it and nodded.
One TechCrunch source put it bluntly last week: Nvidia has been carrying almost every AI-native startup "from application to neocloud."
Now the interim CEO is Colette Kress, Nvidia's CFO since 2013, and she's keeping that job while she runs the company. Investors quoted by TechCrunch describe her as more conservative and expect the venture spree to shrink.
Ask yourself what a CFO does with a strategic investment portfolio in the middle of a leadership crisis, a board investigation, and a Senate summons for the former CEO. She prices it.
The marginal buyer in AI private markets just went from price-insensitive to price-aware. Founders and VCs are already telling reporters they expect smaller checks and lower valuations. The headlines call that a hole in funding. I'd put it more narrowly. The buyer who set the price has left, and the price is leaving with them.
The Loop Was the Product
The check is the shallow exposure. The deep one is the loop the check sat inside.
Take CoreWeave. Nvidia held equity in it, sold it GPUs, and in September 2025 agreed to buy up to $6.3 billion of its unsold capacity through 2032. CoreWeave borrowed billions with those GPUs as collateral. Its customers included labs Nvidia had also funded. That same month, Nvidia signed a letter of intent to put up to $100 billion into OpenAI, which would spend most of it on Nvidia systems.
Critics called this circular financing. Huang called the investments a "once-in-a-generation" opportunity and said the risk was limited. For three years he was right, because the loop needs only one thing to hold: whoever sits at the center has to keep choosing to extend it.
Follow that loop outward and you run into companies that have never taken a dollar from Nvidia.
Say you sell developer tooling to AI labs. Your customers' budgets came from rounds Nvidia anchored. Your compute comes from a neocloud whose debt is collateralized by Nvidia hardware and whose demand was backstopped by Nvidia. Your Series A investors marked you against comps that had Nvidia in them. You're three steps removed, and all three steps run through one balance sheet and one man's appetite.
That's why neocloud stocks tumbled on Thursday. Public markets can reprice a dependency in an afternoon. Private markets take six months and do it in board meetings.
South Korea's AI strategy council held an emergency meeting on Friday. Taiwan's economy minister said the departure won't change the industry's growth much. Both reactions make sense, and put side by side they give you the map. The countries and companies that sell into AI demand are calm. The ones that were allocated their place in it are nervous.
Demand for AI didn't change on Thursday. What changed is who decides which companies get carried while they wait for that demand to arrive.
“Nvidia was the lender of first resort, and nobody priced in that it was one person.”
Run the Audit Before Your Investors Do
For four days I've watched founders say some version of "we're not exposed, Nvidia isn't on our cap table." Wrong test. The cap table is the least important place Nvidia shows up.
Four questions, in order of how fast they can hurt you.
Who sets the price of your next round? Open your fundraising plan. If the lead you were counting on is Nvidia, NVentures, or a fund whose pitch to you included "and we can bring Nvidia in," rebuild the plan without them. Then reprice it. If your last valuation only made sense with a strategic who didn't need a return on the equity, your real valuation is whatever a financial investor would pay. Go find that number now, privately, before a term sheet finds it for you.
Whose balance sheet is under your compute? Skip the question of who your cloud provider is. Ask who financed them. If your GPUs come from a neocloud, read your contract for what happens on a change of control, a default, or a price reset. Ask them point-blank how much of their capacity is covered by Nvidia commitments and how their debt is secured. A good provider will answer. One who won't has told you something too.
Then run one production workload on something that isn't Nvidia. AMD, a TPU, Trainium, whatever your stack can reach. You're not switching. You're finding out how many weeks a switch would take, so it becomes a number you know and stops being a fear you carry around.
How much of your revenue is the same dollar going around? List your top ten customers. Mark every one that raised from Nvidia or runs mostly on Nvidia-financed compute. If the marked ones add up to more than a third of revenue, you have a concentration risk that doesn't look like concentration, because the names are all different. Their renewals now hang on their next rounds, and those rounds hang on the same repricing you just modeled for yourself.
What is the logo doing in your deck? Inception program member. Nvidia partner. "Backed by." Be honest about which enterprise deals and which investor meetings that badge opened. It told people that someone with perfect information about compute demand had vouched for you. The badge is still there. The person who gave it meaning is gone. Whatever it was proving, you need a second proof, and a customer's signed contract is the only kind that doesn't lean on anyone else's reputation.
None of this takes a week. Most of it takes an afternoon and some honesty you won't enjoy.
I'll admit where I'm unsure. Kress could surprise everyone. The board could hire a risk-taker, and Blank frames that as its central choice: does it still want an innovative company, or an execution machine? A new CEO could decide the venture program was the best customer-acquisition spend in corporate history and double it.
But think about what waiting for that would mean. You'd be staking your company's survival on the judgment of a person who hasn't been hired yet. Same mistake, second time around, and now you don't even know whose judgment you're betting on.
Read the Petition Again
A hundred and eighty thousand signatures is a lot of affection for a CEO. Some of it is real. Huang built the thing, wore the jacket, signed the servers, and people loved him for it.
Now read the list of who's loudest. Founders. Investors. Executives whose companies buy his chips or hold his money or both. A structurally sound market doesn't pass around a petition when one executive leaves. It names a successor and carries on.
A petition is what a single point of failure looks like after it has failed.
Nvidia reports earnings on Tuesday. The numbers will be enormous, the order book will be full, and the stock will find a floor. That company was built to outlast its founder, even if its founder never planned for it.
The question worth your week is whether yours was built to outlast him too.
Frequently Asked Questions
How does Jensen Huang leaving Nvidia affect AI startup valuations?
Nvidia invested to sell GPUs, not to earn a return on the equity, so it could accept valuations a financial investor would have fought. Other investors followed because its presence signaled chips, follow-on money and Huang's personal approval. With CFO Colette Kress as interim CEO and described as more conservative, founders and VCs already expect smaller checks and lower valuations.
My startup never raised from Nvidia. Am I still exposed?
Possibly, because the cap table is the least important place Nvidia shows up. Your customers' budgets may come from rounds Nvidia anchored, your compute may come from a neocloud whose debt is secured by Nvidia hardware, and your valuation may be marked against comps with Nvidia in them. If customers that raised from Nvidia or run on Nvidia-financed compute make up more than a third of your revenue, you have a hidden concentration risk.
What should founders do now to reduce their dependence on Nvidia?
Run a four-part audit: - Next round: rebuild your fundraising plan without Nvidia or NVentures, and reprice it at what a financial investor would pay. - Compute: ask your provider how its capacity and debt are backed, and run one production workload on non-Nvidia hardware to learn how many weeks a switch would take. - Revenue: mark which of your top ten customers depend on Nvidia money or Nvidia-financed compute. - Credibility: replace the Nvidia badge in your deck with signed customer contracts as proof.
Join the OPC Community
Connect with solo founders worldwide. Get daily insights, curated opportunities, and peer support.
Join free