# The Circuit: Last Month He Said a CEO Should Be Held to Account; This Month the Board Hired a Firm to Find His Replacement Source: Realpha Blog (blog.getrealpha.com) Original article and charts: https://blog.getrealpha.com/en/blog/circuit-2026-11-12-jensen-huang-resigns-nvidia-s-succession-crisis-an/ > On 12 November 2026 Bloomberg's The Pulse replayed a clip from last month's episode of The Circuit, where Emily Chang asked NVIDIA CEO Jensen Huang about accountability and he said the pressure of the job is high and supposed to be. The same morning, Bloomberg reported the board was working with an outside firm to identify successors; the host put the loss since Sunday's disclosures through post-market trading at $671 billion, with the stock down more than 30% in 30 days and market cap back under $4 trillion from $5.5 trillion two weeks earlier. The allegations remain at the reporting and investigation stage and China has denied them. These are educational listening notes on governance, succession and supply-chain concentration — no investment advice, no price targets, and the judgements here may be wrong. Published: 2026-11-13 Locale: en Tags: NVIDIA, podcast-notes, corporate governance, succession planning, supply chain concentration ![A large keynote stage after the audience has gone, an empty podium in the foreground under a single spotlight with tape marks still on the floor, a floor-to-ceiling portrait banner behind it lit on one side only, rows of empty seats receding into darkness](/covers/circuit-2026-11-12-jensen-huang-resigns-nvidia-s-succession-crisis-an-cover.png) > A state rises through one man and falls through one man. The worthy man grieves not for his own death but for the decline of his state, and so he must first find another worthy man before he may die. > > —— Su Xun, "On Guan Zhong" (Northern Song; my own translation) On the 12 November 2026 edition of Bloomberg's The Pulse, host Francine Lacqua played a short clip from last month's episode of The Circuit, in which Emily Chang interviewed NVIDIA CEO Jensen Huang. In it he says you have to hold yourself responsible, you have agency, you are the CEO of the company, your standards and expectations have to be high, and the pressure is high because it is supposed to be. On the morning that clip was replayed, Bloomberg reported the board was working with an outside firm to identify potential successors. Lacqua put the loss from Sunday's disclosures through overnight post-market trading at $671 billion; Bloomberg Tech Europe anchor Tom Mackenzie put the stock down more than 30% in 30 days, with market cap back below $4 trillion from $5.5 trillion two weeks earlier. The allegations on the show remain at the reporting and investigation stage, and China has denied that any spy was targeting him — so what follows sticks to governance, succession and supply-chain concentration. ![Two bars of clearly different height: the left bar, $5.5 trillion two weeks ago, stands well above the right bar, below $4 trillion in overnight after-hours trading, with a note on the right that the share price has fallen more than 30% in 30 days.](/figures/nvda-market-cap-step-down-en.svg) ## Last month he sat in front of a camera and talked about accountability That interview was recorded before the storm, and it plays like a man setting standards for his future self: the responsibility is mine, the agency is mine, this job should not be easy, and it should not be taken for granted. A month later the same words run next to news of a board meeting, and they land differently. Later in the show, a Bloomberg Opinion columnist drew a parallel. When Emily Chang interviewed Sam Altman, she asked whether people should trust someone who wields that much power, and he said no — not long before he was removed. Her observation is that these CEOs build themselves up while conceding they probably should not hold so much power, and then push comes to shove and here we are. I paused on that. The same humble line reads as a virtue beforehand and as a trailer afterwards. ## Sunday's admission, and the forty-eight hours after it The crisis starts on Sunday. Huang admitted an affair with a woman whom US officials reportedly believe to be a Chinese spy. Overnight, The New York Times reported a second affair, with a senior executive who reported to him. The board was meeting while the show was on air, and The Wall Street Journal's account is that it is set on pushing him out of the CEO role. Mackenzie supplied the number that matters structurally: Huang owns only 3.6% of the shares, so the board can oust him, and there is no question about that. He also said that, looking at the evidence in hand and the question marks over the company, the scenario where he hangs on looks untenable to many. The other gap everyone kept returning to: this is a company with no succession plan, nothing it can pull out of its back pocket. ![In one long horizontal bar, only a small sliver at the far left is marked as the founder's stake, while the entire remaining stretch belongs to other shareholders, showing how tiny the proportion is.](/figures/nvda-founder-stake-sliver-en.svg) ## What a reporter saw on the floor at GTC Mackenzie had been with him at GTC in Berlin two weeks earlier, and his details are more concrete than any governance analysis: floor-to-ceiling posters printed with his quotes, key rings with his face on them, his own avatars. His verdict was that the cult of Jensen is real, and that this is a company moulded around him. More important than the posters is the next layer. Mackenzie said the supply chain was built aggressively on his relationships, leaned on at government level and corporate level globally. The columnist added the figures: 60 direct reports, 33 years running the company, and several of the circular financing deals propping up the AI trade traced to his personal relationships. The host reached for JP Morgan losing Jamie Dimon as a comparison, and Mackenzie said this would be considerably worse, because it is a different organism, a corporate structure built around one person. One aside: hours before the show, Elon Musk said it would be an honour to hire him. ![One central circle is surrounded by eight boxes, and every connecting line runs only to the center, with no lateral links between the boxes at all.](/figures/nvda-single-hub-structure-en.svg) ## Three names on the list, and one counterexample Three successors came up. CFO Colette Kress, who has done a good job by most accounts but lacks Huang's engineering and technical experience. Ian Buck, who invented CUDA — which locks in much of the company's customer base — and now oversees the data centre business, but lacks the operational experience. And a co-founder. Prediction markets favour Buck; the Bloomberg Opinion columnist leans Kress, on the grounds that what is moving right now is investor sentiment, that sentiment carries enormous weight in the AI trade, and that the board wants a safe pair of hands. She also offered the counterexample. When Steve Jobs stopped being CEO of Apple, people said the golden age was over, and Tim Cook went on to steer the company to higher profitability and growth; her read is that Huang may have laid the groundwork for a successor to continue. What I wrote down was a different line: the pool of people globally with that skill set is very small, and that is the risk investors are holding today. After the show aired, NVIDIA said Kress would serve as interim CEO while keeping the CFO role, that the board will consider internal and external candidates, and that there is no timetable. ![A four-quadrant chart on the axes of technical background and operating experience, with the two candidates sitting in the upper left and lower right, and the upper-right corner marked empty by a dashed box.](/figures/nvda-successor-two-axes-en.svg) ## Why this landed on next week's diplomatic calendar Two guests, from Chatham House and the Atlantic Council, pushed the story outward. The US and Chinese presidents meet next week, and there is a truce on the table: China holds off on rare-earth export controls, the US applies fewer export controls to some top-tier chips, and next week's conversation is whether it extends past January. This story lands in the middle of that. Senate Democrats say they will investigate the national security ramifications of the relationship and what it means for chips going to China. One detail worth copying down: export controls were meant to be about national security, and the definition has been stretched to something where you can be blocked for national security reasons, or pay a fee to bypass the block. Useful for the US tax take, surprising to old-school national security practitioners, and the guest's point was that the US loses something here too. The other thread is voters. The midterms kept coming back: a data centre backlash, worries about AI job losses, candidates in both parties running on those platforms, and a group of politicians arriving in January with a different approach to AI regulation. Americans opening their accounts this morning see losses, because pensions and 401(k)s are heavily exposed to these firms. The Atlantic Council guest gave the magnitude: without AI, US growth this year would sit somewhere between 0 and 1%; with it, above 3%, and there is no debate about that. Whether the productivity gains show up is the proof still to come. ![Two bars compare US economic growth this year: without artificial intelligence it is 0 to 1%, with artificial intelligence it is above 3%, and the right bar is several times taller.](/figures/us-growth-with-without-ai-en.svg) ## Which part of my own position is being repriced? When bad news breaks, the hard question is which slice to cut. BlackRock's global head of investment and portfolio solutions offered a split I found useful: look back over the past four to five weeks and the market has priced constraints on the build-out — financing conditions and build costs. Demand has not been reassessed, and the evidence is that prices moved in both directions: neoclouds and semis priced down, while hyperscalers with capacity today and GPU rental rates priced up. It is a rotation inside one trade, sorted by who controls the supply of compute. ![A single line forks from the left into an upper and a lower branch: the upper branch is marked as repriced up and leads to players that already hold compute, while the lower branch is marked as repriced down and leads to neoclouds and semiconductors.](/figures/ai-trade-repricing-fork-en.svg) He threw in a counterintuitive observation too: token volumes are up, while spend on tokens has not risen anywhere near as much. That line matters, because it separates "used more" from "earned more". ![Two lines run right from the same starting point and separate: the upper line, token usage, climbs steeply, while the lower line, money spent on tokens, climbs flat, opening a gap between them.](/figures/token-usage-vs-spend-gap-en.svg) Two readings come out of this for an ordinary investor. First, when your position drops, ask whether this is a supply-side repricing or a demand-side one; the holding period and the stop logic differ. Second, he said that unless you are short-horizon and have high conviction about which layer wins, you want exposure across the whole stack rather than concentrated in one layer — and that in their surveys the use of market-neutral strategies is rising, because some people have no conviction on the big call about real usage demand and would rather not take direction at all. I took that as a prompt: is my conviction in the handful of names I own strong enough to justify not diversifying? ## How do you spot a company resting on one person? The closer-to-home worry is whether a company I own works this way and I cannot see it. McKinsey global managing partner Bob Sternfels gave three questions you can take straight to a filing. First, does the company have a playbook. He said having one, even an imperfect one, beats figuring it out in the heat of the moment, and that boards and management teams should have walked through succession, geopolitics and credit market disruption before any of them arrive. Second, he called succession planning a narrow topic; the real one is leadership development — generation after generation, and not at the CEO level alone but at multiple levels, so there is more than one option. That, he said, is where the gap is, because everyone focuses on the narrow version. Third, governance needs someone whose job is to put challenge into the system; the seat can be the board or the CEO's top team, and what counts is that somebody is prepared to present the contrary view. His trade metaphor stuck with me as well: trade is like water, and if you try to dam it, it will find other ways to flow. Global trade was up in 2025, yet his institute's work points at deeper Achilles heels — too much trade passing through pinch points, too much occurring between countries that are geopolitically distant, too much concentrated in a few suppliers. AI tools give you more transparency into your supply chain than ever, he said, while the AI supply chain itself carries concentration risk, and where there is no choice, that is your heel. The Atlantic Council guest added the same shape: leadership this personalised is not unique to this company, and the risk sits elsewhere too. ![Six lines on the left all converge through a single narrow neck in the middle and then fan out to the right, with the neck labelled as the layer that has no alternative.](/figures/supply-chain-chokepoint-neck-en.svg) I ran those three questions over my own holdings this week. The phrasing was crude: if this CEO left, can I name who takes over? Has anyone on an earnings call pushed back on management? Is the critical supplier one firm or three? I can usually answer the third. The first two I often cannot — and not being able to answer is itself the answer. ## Further reading - The Circuit, Emily Chang's interview with Jensen Huang (last month; on YouTube, Bloomberg.com and the Bloomberg Terminal) - Bloomberg's The Pulse, 12 November 2026, hosted by Francine Lacqua - McKinsey Global Institute's latest work on global trade, referenced by Bob Sternfels on the show ## One thing to take away One idea: get something to the point where only you can do it, and your risk peaks at the exact moment your value does. The cult, the posters, the key rings, the 60 direct reports, the 33 years of supply-chain relationships built by hand — every one of them was an asset while things ran well, and all of them turned into the same problem in a single week, because there was no second person who could catch it. This has nothing to do with company size; a one-person studio, a household and a team of six all work the same way. ![Two panels side by side: on the left, one person with a dashed box of steps hovering above their head; on the right, two people with a solid arrow between them and a solid box of steps.](/figures/only-i-can-do-it-handoff-en.svg) Here is something I have tried. Write down three things that only you can currently do — the tax filing at home, a parent's medication schedule, the update on some report at work. Take the one you least want to hand over, spend twenty minutes this week writing the steps in an order somebody else can follow, then find a person to do it once while you watch. Wherever they get stuck is the part you thought you had written down and had only kept in your head.