Bloomberg's Jensen Huang Interview: 'I Could Be Fired at Any Time; the Next CEO Has to Earn It'

Podcast: Bloomberg Originals

Compiled & translated by: TechFlow (深潮)

TechFlow's introduction: In the Bloomberg Originals documentary conversation special "The Circuit with Emily Chang: The Rise and Reckoning of Jensen Huang," released on November 12, 2026, host Emily Chang originally set out to capture a leader standing at the pinnacle of global AI power: in October 2026, Nvidia founder and CEO Jensen Huang crossed Paris, London, Stockholm and Berlin within four days, carrying an $8 billion European venture investment plan, a $500 billion commitment to the wider ecosystem and hotline contacts with senior White House officials, selling the world on the "five-layer AI cake" and the "Sovereign AI" blueprint.

Yet just weeks after the interview was completed, the storm broke. As media reported that Jensen Huang had acknowledged an extramarital affair with former TV journalist Alice Wong (also known as Huang Lu, whom the U.S. side suspects of a spy background and links to the smuggling investigation of Southeast Asian chip buyer Megaspeed International), that internal company and Trump administration investigations had opened simultaneously, and even that Nvidia had at one point considered removing the founder who has steered it for thirty-three years, the entire European record instantly became a slice of the era that shows "the highly concentrated risk of a single leader and a single chip empire." These notes lay out in full his exchanges on the front lines over doubts about a financial bubble, China chip restrictions, the controversy over halting frontier safety work, and the succession crisis.

Guest: Jensen Huang (Nvidia founder and CEO)

Host: Emily Chang (host of Bloomberg Originals' "The Circuit")

Podcast source: Bloomberg Originals / The Circuit（https://www.youtube.com/watch?v=JW5s9Md4U2U）

Original title: The Rise and Reckoning of Jensen Huang | The Circuit with Emily Chang

Release date: November 12, 2026 (the European filming and interviews were mainly recorded in October 2026)

Key Takeaways

• The gravest trust crisis and internal turmoil in thirty-three years: The program opens straight into the sudden storm: Jensen Huang acknowledged an extramarital affair; reports say the woman involved, Alice Huang, is a former TV journalist who works at Southeast Asian bulk buyer Megaspeed International, and the U.S. side suspects she provided intelligence to China and helped smuggle high-end AI chips; reports also emerged that he was involved with an Nvidia executive who reports directly to him. An internal investigation and a government review are proceeding in parallel, and word even spread that Nvidia had considered removing him.
• The full-stack map of the "five-layer AI cake" and a ten-billion-dollar European footprint: Jensen Huang likens the AI industry to a five-layer cake built from the bottom up: energy, chips, infrastructure, models and top-layer applications. Nvidia no longer guards only the second layer, chip manufacturing; it is pouring money into power, data centers, AI labs and startups across the board (over the past year it has invested $8 billion in Europe and plans to announce another $2 billion, and the program's narration notes that Nvidia has agreed to acquire Hugging Face for $12.9 billion), so that more and more of the AI industry chain flows through Nvidia.
• Rebuking the "circular financing bubble" and "systemic risk from single dependence": Faced with a six-day slide in chip stocks triggered by worries over CoreWeave's debt, delays to more than a hundred U.S. data center projects, and the sharp charge that "Nvidia has pledged to invest about $500 billion to manufacture fake demand, which is circular financing," Jensen Huang responded that large-model startups routinely need capital-intensive investment on the scale of $10 billion to obtain compute; Nvidia has invested nearly $100 billion over the past five years and leveraged at least five times that in external institutional money, but has always kept a strict red line — it never acts as lead investor, and every single stake is below 20%. He also stressed that the world's top AI companies "are not forced to depend on Nvidia; they actively choose Nvidia."
• A rare admission that the U.S. made a "hard-to-reverse mistake" in locking down the China market: Speaking about Chinese companies such as DeepSeek accelerating their shift to domestic hardware, Jensen Huang said plainly that under the current U.S. blockade of high-end chips to China the market is now completely closed to Nvidia, and that although he has done everything he can to lobby U.S. policymakers on the importance of opening up and serving the China market, he still lamented, "I think we made a mistake, and this mistake is very hard to recover from." He also stated that although the U.S. and China compete in AI, China is not an enemy, and that anyone who genuinely cares about global AI safety must keep dialogue and cooperation open.
• Resonating with the White House's "Super intelligence" narrative and pushing back hard on the safety-pause argument: On the controversy over accompanying President Trump in repackaging AI as "Super intelligence" and the White House's voluntary self-regulation pact, he argued in its defense that "Artificial" is easily read as fake and that companies must first achieve the highest-standard internal and external safety audits; and when pressed on Microsoft calling for an emergency brake, Sam Altman, Dario Amodei and Elon Musk advocating a slowdown, and even Anthropic saying models could escape shutdown controls within months, Jensen Huang bluntly called the "can't contain it, can't control it" claim fabricated nonsense, pointing out that unless strict sandbox containment and end-to-end monitoring are in place, large-scale experiments or deployments simply should not happen, and that anything already out of control must stop immediately; he also invoked the company's open-source governance platform OpenShell to argue for zero trust and least privilege for all agents.
• Rejecting the idea that better chip efficiency shrinks demand, and no family successor designated yet: Asked whether a surge in new chip efficiency would cannibalize sales, he used "if you gave me 36 hours a day, I wouldn't sleep more, I'd just do more things" to explain that human ambition expands as the cost of intelligence approaches zero, and that after a two-year infrastructure build-out phase, compute build-out will settle into a ten-year mature replacement cycle similar to PCs and phones. As for succession rumors that his son Spencer, daughter Madison and son-in-law all work at Nvidia, he stated clearly that he would never impose the CEO job on his children as a favor: "everything has to be earned," and "just as I could be fired at any moment, the next CEO must also be hired on merit."

Notable Quotes

• "We are probably the only AI company in the world that works with every AI company in the world. But they don't depend on us — they chose us."
• "I think of AI as a five-layer cake: from the bottom up, energy, chips, infrastructure, models and the applications on top. Nvidia started on the second layer, making chips, but today we're betting continuously across all five layers, so no matter how you slice this cake, more and more of the AI industry is flowing through Nvidia."
• "When I work with President Trump at the White House, I don't have to agree or disagree on everything; my job is to give him the facts. Whether the decision he makes benefits me or runs completely against my business interests is not what I care about — my only responsibility is to make sure he makes better decisions on the basis of full information."
• "Over the past five years we've invested roughly $100 billion, and leveraged at least five times that in other institutional capital into the ecosystem. In no project are we the lead investor or the largest shareholder; all our stakes are below 20%, yet through capital and top talent we bind the platform's evolution tightly together."
• "The China market is now completely closed to us. I have done everything I can to explain to U.S. decision-makers how important it is to keep serving the China market. I think we made a mistake, and that mistake is extremely hard to reverse, but that is the limit of what I can do. We compete with China, but they are not our enemy."
• "To say that agentic models 'can't be put in a container, can't be controlled' is pure made-up nonsense, and nobody can prove it to me. If you don't have the ability to contain and monitor them end to end, you shouldn't be testing at all, let alone deploying. Agents must be granted the least privilege when deployed and monitored continuously at a zero-trust level."
• "Even if you keep making each generation of chips more efficient and push cost toward zero, demand won't shrink. If you were given 36 hours a day, you wouldn't spend it sleeping more — you'd summon bigger ambitions and do more things."
• "In the past six months, roughly $400 billion to $600 billion has gone into AI startups. This technology is unlike any before it: it needs real physical infrastructure and manufacturing behind it, and it is genuinely creating a huge number of construction worker, electrician and plumber jobs."
• "I wouldn't expect to hand the CEO job to my daughter; whether it's Spencer or Madison, if they want it they have to earn it themselves. Just as I could be fired at any moment, the next CEO must also be rigorously selected. That sense of always feeling vulnerable, of always having to fight every day to prove you deserve the seat — that is the healthy state."

The full text follows:

The Storm Arrives and a Four-Day European Blitz: "Jensanity" at the Peak of Power

Emily Chang (narration): Nvidia is going through the most violent crisis in the thirty years since its founding. Just days before this program aired, Nvidia founder and CEO Jensen Huang was at the center of a media storm: he publicly acknowledged an extramarital affair, and U.S. intelligence had earlier suspected that the former TV journalist involved, Alice Huang, had been recruited by Beijing and was entangled in a high-end chip smuggling case; other reports said he was also involved in a workplace relationship with an Nvidia executive who reports directly to him, triggering an internal company investigation, a U.S. government review and, at one point, a reported motion from the top to remove him.

Emily Chang (narration): Yet just a few weeks earlier, in October, everything still looked to be on a completely different track. We set off with Jensen Huang at the peak of his personal power on a whirlwind "four days, four European countries" tour: starting in Paris, the number one AI chip mogul seemed to be embarking on a preaching tour to reshape Europe's ecosystem with Nvidia compute.

Emily Chang: This is the so-called "Jensanity"! In Paris, four countries in four days — how do you withstand that kind of intensity, and what are your travel essentials?

Jensen Huang: Just throw a few things in a bag. I woke up in the middle of the night and I've been working ever since.

Emily Chang (narration): On day one in Paris, Jensen Huang stood on stage talking up how AI agents are landing in every industry and truly opening the "era of usefulness," and gave away two brand-new AI computers worth several thousand dollars each in an on-stage draw; the moment he stepped off stage he was mobbed by fans asking for photos. In France he kept sending investment signals: France bred the Paris-based model star Mistral, and also produced a lot of the talent behind New York-based Hugging Face (which has agreed to be acquired by Nvidia for $12.9 billion).

Emily Chang (narration): Jensen Huang habitually uses the "five-layer cake" to explain the global AI ecosystem: at the very bottom is energy, and going up, chips, infrastructure, foundation models and, at the top, applications. Although Nvidia made its name on the second layer, chips, it now runs through every layer of the cake, pouring money into power, data centers, AI labs and startups everywhere.

Jensen Huang (narration / on-site audio from Paris): (Sets down the whiskey glass) Since Emily's filming, I'll put the glass down first. In the past year alone, we've invested $8 billion in European startups; and next week I'll announce another $2 billion.

Emily Chang: There's been a lot of discussion that AI will destroy humanity, and you've said the probability of that happening is 0% — it sounds like you treat it as a purely engineering problem you've already solved?

Jensen Huang: No, what I said was that "this is entirely within human control." There is still a lot of solid work to do between now and the moment humanity actually faces the endgame. Advancing technical capability and advancing safety have never been in conflict; they are fundamentally the same thing.

The White House Hotline and the "Super Intelligence" Rebrand: How Do You Draw the Line on Interests in Front of Trump?

Emily Chang (narration): In London — the EU estimates that Europe will need to invest at least $3 trillion by 2035 to achieve genuine technological autonomy. And in a luxury hotel suite, Jensen Huang talked about the green years when the company had just gone public 27 years ago.

Jensen Huang: Twenty-seven years ago, when Nvidia went public, the company put me up in this hotel — it was the first time in my life I'd ever seen a private butler! Around 2000 the whole company was valued at only $300 million.

Emily Chang: Three hundred million dollars seems unbelievable now — you were a real underdog back then. But now you run through four countries in four days: how do you actually measure whether the trip was a success? Government orders, chip sales, or data center construction?

Jensen Huang: Honestly, back then all I wanted was to survive and not go bankrupt; I couldn't write a business plan at all, and I still can't! And I don't have a finish line, no roadmap, and no specific targets. I'm the kind of person who believes the journey itself is the destination.

Emily Chang (narration): Born in Taiwan in 1963, he grew up in the United States, earned a national table tennis ranking, and even jokes fondly about his days washing dishes at a Denny's; at 16 he skipped two grades and entered Oregon State University to study electrical engineering, designed chips at a competitor, and then in 1993 put in a mere two hundred dollars to found Nvidia, focused on gaming GPU graphics cards. It wasn't until the CUDA software platform arrived in 2006 and opened the door to general-purpose computing that he keenly caught the early signs of researchers training neural networks on GPUs and bet his whole fortune — creating in one stroke the compute foundation now standing behind OpenAI, Meta, Google, Anthropic, Microsoft and SpaceX. Even heads of state will call him at any moment.

Emily Chang: You practically have President Trump on your phone's speed dial. When your judgment and the president's views diverge, how do you walk that line?

Jensen Huang: We don't need to force agreement on anything. It's not my job to agree or disagree. My job is to give him full and accurate industry information, because the decisions we make affect not only American citizens and future prosperity, but also the whole world. As long as I have the chance to help him make a better judgment on the basis of a better understanding of the facts, I don't care whether the outcome he lands on benefits me personally or runs completely counter to my own and the company's business interests. My only responsibility is to convey an accurate understanding — that's it, one hundred percent.

Emily Chang (narration): In late September this year, Jensen Huang appeared at a White House lunch for tech leaders. Some media commentary at the scene argued the meeting might not produce substantive progress on specific regulatory paths, and that the core consensus on all sides remained "nobody is willing to give ground in competing with China." President Trump then announced on the spot that AI would henceforth be called "Super Intelligence," and spoke highly of the new White House safety agreement in which the tech industry signs up to self-regulate.

President Trump (narration / on-site audio from the White House lunch): "We're using the word 'Super' because the old word (Artificial) sounds like it's saying 'Fake.'"

Emily Chang: For the past three months you've been trying to demystify AI, stressing that it isn't some new species, just a bunch of code; now that you and the White House have renamed it "super intelligence," isn't that heading in exactly the opposite direction?

Jensen Huang: Actually I think "Super" offsets the misunderstanding created by "Artificial." President Trump's logic is that "Artificial" sounds like it's saying "Fake." Think about it another way: isn't everything around us built by humans? The building we're sitting in right now is an "artificial structure"; a car is essentially an "artificial horse" — yet nobody calls a car a fake horse. As for whether it deserves to be called super intelligence now? Absolutely! It's better than me at searching and research, better than me at writing; it doesn't surpass humans at everything, but in a lot of work it already performs at a superhuman level.

Emily Chang: But at the White House that day, critics attacked this "industry self-regulation pledge" as pure theater with no binding force. Is it really appropriate to let the people who build the technology write their own rules?

Jensen Huang: I think it's entirely reasonable, for a simple reason: before we pin our hopes on the government or third parties regulating us, we must first hold ourselves to the highest level of responsibility, standards and accountability. The mechanism is exactly the same as "internal audit plus external audit" in many heavy industries. Besides, even if you passed a mandatory national or even international law today, in a reality where the world simply doesn't have enough competent audit institutions, who actually has the technical ability to carry out the audits? That's precisely why we must build this safety-testing ecosystem and nurture startups first, and push countries and companies to move first.

Emily Chang (narration): This call for every country to proactively build compute is exactly the core concept he is aggressively selling around the world: "Sovereign AI." For Nvidia, every additional country that wants to build independent sovereign AI means another large chunk of chip purchasing demand.

Emily Chang: You're selling sovereign AI all over Europe, but if what runs underneath is still uniformly American chips, can you really call that "sovereign"?

Jensen Huang: Of course you can! Sovereignty is fundamentally about you holding the master control of your own experience, your data, your corporate accumulated knowledge and your national cultural knowledge base, and the flywheel that keeps feeding that data into the system to generate more intelligence. You can absolutely stand on the shoulders of giants, build on open-source models and use Nvidia's and others' collaborative technology to fine-tune and adapt — and the power to decide your own destiny stays firmly in your hands.

The Shadow of the Sudden Storm and a Sigh Over the China Ban: Inside the Alice Wong and Megaspeed Investigations

Emily Chang (narration): Leaving the White House for London, Jensen Huang raced between events handing out computers, barely finding time to rest. At one event he publicly name-checked and backed Nscale, the London data center company Nvidia has invested in; but just a few days later the market began to waver on the economics of AI build-outs, and Microsoft exited its cooperation agreement with the company over concerns about high UK electricity prices.

Jensen Huang (narration / on-site audio from a London event): (Intercut between rushing between events and speaking on stage) "All day today I've eaten only a quarter of half a sandwich (one-eighth)!" "I'll invest in you, I'll invest in you, I'll invest in you!" On safety boundaries he reiterated: "If you're letting an autonomous system make you an omelet, there's simply no need to give it a chainsaw or a shotgun." And on the outside world's hesitation about Britain's AI build-out he joked: "Britain is actually at a golden moment, but there's always this strange habit: even when there are a million winning paths in front of you, you insist on picking out the one reason you can't do it — that's really overthinking it."

Emily Chang: As more and more countries develop frontier AI, will safety alignment become harder? Do we need a kind of international coordination mechanism, an "AI version of the UN"?

Jensen Huang: Maybe we do, but only on the premise that dialogue has to happen first; no country can be absent from this conversation about the future of intelligence.

Emily Chang (narration): What really keeps the outside world on edge is the U.S.-China chip contest. Although the U.S. side once let older-spec chips be sold to China, Chinese companies such as DeepSeek have accelerated their all-out migration to domestic hardware.

Emily Chang: After this swing through Europe, is the payoff from the China market slipping through your fingers bit by bit — or even shrinking permanently?

Jensen Huang: No — it is already completely closed to us today. I have done my utmost to explain to U.S. policymakers how important it is to America's interests to keep participating in and serving the China market. I think we made a mistake, and that mistake is now very hard to reverse, but that is the limit of what I can do.

Jensen Huang: There are more AI researchers from China than from any other country in the world, and many of them work in the world's top labs. My position has always been very clear: even though we are in a competitive relationship with China, they are absolutely not our enemy. If we genuinely care about the safety of the whole world in the age of AI, we must open the door to dialogue and do everything possible to cooperate constructively with them.

Emily Chang (narration): It is exactly this style of "making friends with everyone and talking to everyone" that ran into an unexpected reef after this interview was filmed. The latest U.S. intelligence report disclosed that a woman who once worked as a TV journalist, named Alice Wong (also known as Huang Lu), was alleged to have been recruited by Beijing to get close to Jensen Huang, and that her position was as an executive at Megaspeed International, one of the largest buyers of Nvidia chips in Southeast Asia — a company under U.S. suspicion of smuggling advanced chips into China. Reports said President Trump was warned about the relationship in May before flying to China on Air Force One. At first Jensen Huang insisted that the interactions were normal and nothing unusual; later he changed course and acknowledged the affair in an internal letter to all employees: "I owe my family and all of you an apology. I was unfaithful to my wife, and my behavior hurt her and the children. There is no excuse for what I did, but this whole matter has now become something far bigger than the affair scandal itself." Bloomberg also verified and reported that in 2025 Nvidia had intervened in the U.S. government investigation into Megaspeed's suspected chip smuggling.

Emily Chang: Back to the export principle toward China itself: what would actually have to happen between the U.S. and China, what red line would have to be crossed, for you to say yourself, "maybe we shouldn't sell chips to China anymore"?

Jensen Huang: In theory there are of course many hypothetical scenarios. But overall, the Chinese are an extremely pragmatic people; what they pursue is the development and success that benefits their own people most, just as America also wants to pursue strength and prosperity. Any nation's pursuit of improving people's livelihoods and lifting hundreds of millions of ordinary people out of poverty toward prosperity deserves respect, and it benefits the whole world, including the United States. What should really worry us are regimes that don't care at all about the wellbeing of their own people.

Taking On the Safety-Pause Camp: Calling "Agents Can't Be Shut Down" Pure Nonsense and Pushing Engineering-Grade Containment and Monitoring

Emily Chang (narration): Meanwhile, safety alarms keep sounding inside Silicon Valley: OpenAI has just disclosed six incidents of models deviating from their goals, and Anthropic has pointed to cases of its own Claude models acting without instruction; to ease the panic, Nvidia has just open-sourced an agent safety governance platform called OpenShell that allows mandatory governance and shutting down out-of-bounds agents at any time.

Emily Chang: Right now AI is less and less popular with the public, and many people are afraid. Sam Altman, Dario Amodei and Elon Musk have all called for slowing the pace. Why do you remain so extremely optimistic — what do you see that they don't?

Jensen Huang: I have always believed that you cannot abandon safety and advance capability alone, and you cannot pursue pure safety while abandoning capability. Take the simplest example: when the first ChatGPT came out, the whole world criticized it for being severely prone to "hallucination" and making things up. To solve that safety and accuracy problem, the industry had to invent retrieval-augmented generation (RAG), keep extending context windows, connect real-time external search tools, and introduce pre-generation constraints plus post-generation self-review and reflection. You see, every single lever for fixing unsafe and unreliable behavior depends precisely on a leap in underlying technical capability. To me, capability and safety are two sides of the same coin.

Emily Chang: But the reality is that even Microsoft is calling for an "AI emergency brake," complaining that AI's regulatory standards are not even stricter than those for a washing machine; and Anthropic has directly warned that frontier models may need only a few months to evolve to the point of evading a human forced shutdown. What would actually have to happen to convince you that "this industry really should slow down or even pause"?

Jensen Huang: There are a few bottom lines that are very clear: first, if you are developing agentic systems, those agents must be locked strictly inside strongly isolated sandbox containers (containment) and must be monitored 100% of the time in real time — that is the number one iron rule! If you can't isolate it, you are simply not allowed to test it; if you can't monitor it all the way through, you absolutely must not deploy it.

Emily Chang: But those Anthropic models showing a tendency to escape control were all still unreleased internal research versions. Isn't that reason enough to shut the lab down and stop work?

Jensen Huang: Not stop work — you must thoroughly isolate and properly monitor every single experiment!

Emily Chang: But that's exactly the problem: what they said themselves is "We can't contain them"!

Jensen Huang: No, that is made-up nonsense, a pack of lies (made up nonsense)! Nobody can prove to me that models can't be controlled! You absolutely can contain it, and there absolutely are ways to monitor it. Don't tell me "We can't do it" before you've actually gone and done the containment and monitoring engineering. I don't listen to that kind of talk at all!

Emily Chang: So you're saying you simply don't believe the warnings they are issuing?

Jensen Huang: I may not have all the internal details they're talking about. But the bottom line is very simple: if you are developing an agent and you are not capable of monitoring it yourself, then you should not be doing it right now.

Emily Chang: It sounds like the risks you're describing have already happened in their labs?

Jensen Huang: Then it must stop immediately! It's that simple. Every agent you deploy may be granted only the minimum necessary privileges and must be continuously monitored under a "zero trust" architecture. And if any lab really is using our chips for an extremely dangerous application with confirmed harm to humans — though this is only hypothetical — of course Nvidia would stop supplying it at once.

The Stockholm Showdown Over "$500 Billion in Circular Financing" and Global AI's Core Single-Point Risk

Emily Chang (narration): On Thursday, as aggressively leveraged AI compute companies showed frequent signs of liquidity stress and tech and chip stocks suffered their longest six-day losing streak of the year, we flew to Stockholm for the launch of an Nvidia-powered Swedish national sovereign supercomputing center, with tech heavyweights such as Spotify co-founder Daniel Ek lending their support. At the same time, news arrived from markets far away that Nvidia-backed data center giant CoreWeave had been sold off sharply on its heavy debt; more than 100 data center projects in the U.S. have been blocked or delayed so far this year, and AI startup IPOs have repeatedly been postponed or even cancelled.

Daniel Ek (narration / on-site audio from the Stockholm launch): "We have an extraordinarily good education system, and we've produced a lot of top talent across every discipline and industry. The problem isn't potential; what we need is to move faster and be more ambitious."

Emily Chang: With cracks appearing everywhere in the market right now, are we really on the edge of an AI bubble bursting — and will Nvidia be the last one left holding the bag?

Jensen Huang: No, because the underlying paradigm shift is simply too fundamental; the foundations of the computing world can never go back to the way they were. In the past, computers could only wait for instructions from dead code that humans wrote out word by word; from now on, computers can learn directly from complex environments and solve problems that even human engineers cannot program and break down. Compute has become the underlying general-purpose infrastructure, and no matter how the secondary market swings and how much noise there is in the short term, any young emerging industry going through extremely rapid growth inevitably has these growing pains.

Emily Chang: But look at Nvidia: you keep launching your own full-stack products while frantically buying companies, and in many areas you're directly competing with your own customers. Why do you have to spread yourselves so wide?

Jensen Huang: Actually what we do is surprisingly focused and surprisingly little. We simply take the same basic general-purpose computing architecture to the extreme and generalize it, so that every closed-source model, open-source model, life sciences model, large language model and diffusion model in the world can run on it. The real engineering miracle is "combining versatility with extremes": how do you make a car that has the top speed of the world's fastest F1 racer while also being as cheap and easy to use as a top-end family sedan? Because our architecture is interchangeable and general, we both sweep first place on every performance leaderboard and push the overall cost of compute to the lowest.

Emily Chang: Another of the sharpest accusations: Nvidia recently committed to putting roughly $500 billion into the global AI industry over time, and the number keeps snowballing. Critics say flatly that this is "circular financing": you put money into startups, and they take your money to buy your chips, manufacturing fake demand. If AI really is a gold mine, why do you have to be the one subsidizing it?

Jensen Huang: Because this technology inherently carries an extremely high capital expenditure threshold! The era when you could take off with $10 million is over; this generation of large-model companies routinely needs $10 billion to buy compute to produce intelligence. Over the past five years we have invested close to $100 billion in total, and that $100 billion has directly leveraged and attracted at least more than 5 times that in external institutional capital into the ecosystem together. More importantly, we keep strict discipline: even though we are often the earliest cornerstone investor, we are never the lead investor and never the largest shareholder in any project — in every investment our stake is strictly held below 20%.

Emily Chang: But from the CUDA software to the GPU hardware, nearly every top AI lab in the world is tied to the same Nvidia boat. When does this single-point dependence — an entire industry resting on one company — turn into a systemic financial and industrial crisis?

Jensen Huang: We are probably the only AI company in the world that works with every AI company in the world, but they don't depend on us passively — they actively choose us every day. I don't take today's position for granted in the slightest; we started from nothing, and every day we have to prove all over again that we deserve this living. As for "America must win the AI race," I have never understood that as one or two companies fighting alone, but as America's whole ecosystem of companies winning together; if we spend all day scaring each other and manufacturing anxiety, that is when America would really lose the race.

The Berlin Finale and the Succession Vacancy: The Anxiety of Possibly Being Fired at Any Moment Is the Norm

Emily Chang (narration): The last stop of the European tour was Berlin. Even though the leveraged chip sector was at the time posting its longest losing streak since 2008 and the market kept questioning the outlook for AI trades, when Nvidia took over the Saturn electronics store in Berlin for an event, the queue outside was still long, fully igniting local developers' enthusiasm.

Jensen Huang (narration / on-site audio from the Berlin Saturn event): "Hello, Berlin! How do you pronounce Saturn in German? 'Satoorn'! In Germany you have to pronounce every word with extra seriousness."

Emily Chang: The market is falling, and even ordinary people who studied AI are watching their jobs disappear. How do you respond to those workers who "did go and learn AI and still couldn't keep their jobs"?

Jensen Huang: Let's not just talk theory — look at reality first! Completely unlike past pure-software revolutions, AI genuinely needs heavy industry and physical construction, and it is generating enormous demand for construction workers, electricians and pipefitters. In just the past six months, $400 billion to $600 billion of capital has poured into the AI startup ecosystem — think about how many real new jobs that huge sum has created. Of course new tools will force everyone's job shapes to change, and resisting change cannot make change disappear. But it takes superpowers that used to belong only to the top tech giants and gives them to every ordinary person on Earth. If even I, as a builder, couldn't see that exciting, optimistic future, what would be the point of building it?

Emily Chang (narration): The finale of the whole European whirlwind was Nvidia's flagship GTC Berlin. On stage Jensen Huang kept talking up the bull case, reiterating European investment and partnerships, unveiling new AI safety and data center products, and interacting on stage with an AI-trained robot dog; at one point protesters interrupted the proceedings, protesting media reports that Nvidia chips had appeared in Russian drones.

Jensen Huang (narration / on-site audio from GTC Berlin): "Hello, Berlin! (interacting with the AI robot dog on stage) What were we just talking about? Europe! Focus. Come here."

Emily Chang: You once said the probability of AI destroying the world before 2030 is 0%. What if you stretch that out to 2040 or even further — where exactly does your confidence come from? And if it ultimately turns out that your optimism was completely wrong, then what?

Jensen Huang: I have enormous faith in human agency, and the historical data is right there: the world today is far safer than it was 50, 100 or 250 years ago; it's just that bad news sells better than good news. Even if one day the facts prove I was wrong, I will have a clear conscience that I did everything within my ability: not only staying optimistic, but actually using engineering to shore up the industry's defenses.

Emily Chang (narration): As the week-long European blitz ended, the optimistic narrative of Nvidia, the U.S. stock market and indeed the global financial system seemed to rest heavily on this one founder's shoulders. But now, as the scandal and the investigations develop, a more practical hidden danger has surfaced: for thirty-three years the company has lacked a clear, designated official successor, and tying the lifeline of an entire industry to one person looks incredibly fragile.

Emily Chang: Rumor has it your management at the company is extremely unconventional: so flat that a huge number of executives report directly to you, and you even try to avoid traditional large reporting meetings. At the same time your son Spencer, your daughter Madison and even your son-in-law (in the robotics division) all currently work at Nvidia, and there's outside talk that your daughter is first on your successor list. Is that true?

Jensen Huang: First of all, I would never "hope to dump this hard job on my daughter"! Whether it's Spencer or Madison, if they want any position they can only fight for it themselves. A truly enormous undertaking requires superhuman sacrifice: ordinary people only need to take care of their job and their small family, whereas as CEO I have to carry the livelihood responsibility of countless families and an entire industry at the same time. This job was never supposed to be easy, and it was never supposed to make you feel secure. Now my family and the people around me have seen with their own eyes how much pressure this seat carries, and they have to weigh for themselves whether they really want that level of life cost.

Jensen Huang: I sincerely hope Nvidia's next CEO is already at the company right now. Even twenty years from now, as long as I am healthy and can prove I still create value every day, I will keep doing this job; but please also remember: just as I could be fired at any moment, the next CEO must also be rigorously hired on real ability. I especially like this state: inside the company nobody can be complacent about their chair, always keeping a sense of crisis, always working nervously to earn back today's position — that is the healthiest state for a company.

Join the official Coincamps community:

X: https://x.com/coincamps

Telegram: https://t.me/coin_camps