Tae Kim on Nvidia post-Jensen: succession candidates, deal-making moat at risk, and Anthropic's IPO timing

Nvidia Post-Jensen: Succession, Moat, and Anthropic's IPO

Tae Kim, who spent over 100 interviews researching Nvidia's culture for his book, argues Jensen Huang's departure is final — and that the second affair allegation, involving a direct report, was the decisive moment. Nvidia's code of conduct makes relationships with subordinates an unambiguous conflict of interest. When Reuters added a third allegation, Kim says, "it was just game over."

The portrait of Jensen that emerges from Kim's reporting is not the one skeptics are now offering. The theory that Jensen was essentially a dealmaker-frontman while others ran the technical operation is, in Kim's view, "completely false." Jensen spent weekends writing 15 to 20 short, operational emails a day to executives, served simultaneously as chief marketer, chief business strategist, and HR chief, and micromanaged across all of it. The harder question is what happens to a company built around that singular operating style.

“There's no replacement. Just don't care for AI. There has never been and there can never be a true succession for a founder like him. The two candidates I did put out there publicly are Renee Haas, the CEO of ARM, and Ian Buck, the creator of CUDA. Financial Times reported, like, an hour ago that they are talking to Renee Haas.”

Succession

Kim puts two names at the top of the list. Rene Haas, CEO of Arm Holdings, is his preferred external candidate — a former Nvidia executive who the internal team "loved working with" and who could manage the public-facing and business functions while leaving technical leadership to those already running it. The Financial Times reported, while they were speaking, that Nvidia's board is already in conversations with Haas.

The internal candidate is Ian Buck, creator of CUDA and head of Nvidia's data center business. Kim points to a 2019 podcast in which Buck was discussing large language models, scaling laws, and natural language processing years before the market caught up — as evidence he has Jensen's ability to see the technology future.

Colette Kress, currently interim CEO, is not viewed as a permanent appointment. Separately, multiple sources told Kim they were surprised by how quickly Kimberly Powell was promoted through Nvidia's ranks, noting it didn't follow the normal process — a detail that becomes more complicated in the current context.

Kim's framework for stabilizing the company is straightforward: no single successor can replicate Jensen, so the goal is to split the role across two or three people, retain the deep technical bench, and hope the new CEO knows their own limits. He identifies roughly 10 people the board needs to lock down immediately — and flags Jonah Alton, head of GPU engineering, alongside Kress as priorities. "Throw a billion dollars at these people," he says.

The two-to-three-year product roadmap is, in Kim's view, already set. The critical variable is deal-making strategy in the three-to-five-year window, where Jensen's relationships and instincts were irreplaceable.

Moat at risk

The concern isn't engineering continuity — it's the deal-making layer. Kim acknowledges the DHH-style argument: Boeing, Intel, and Apple all degraded after the founding operator handed off to someone without engineering depth. His counter is structural rather than dismissive. If Haas handles conferences and business management and operators like Buck and Alton run autonomously on the technical side, Nvidia doesn't need to find another Jensen. It needs to avoid giving a single non-technical executive control over strategic direction.

China and the loophole

On the geopolitical side, Kim expects the data center loophole — which has allowed Chinese customers to access Nvidia GPUs through international data centers in Malaysia and Singapore — to close. Legislation is likely. The bigger constraint on Chinese AI labs, though, may be distillation access rather than raw compute. Kim argues the main reason Chinese models have been competitive is that they've been distilling from Anthropic and OpenAI outputs. Both labs have been systematically cutting off those accounts. As recursive self-improvement cycles accelerate, Kim expects the gap between US and Chinese frontier models to widen significantly over the next six months.

Anthropic's IPO

The IPO timing, which looked shaky a few weeks ago, has been rehabilitated by a single data point and deliberate positioning. OpenAI's aggressive push against Anthropic from July through September compressed Anthropic's growth numbers in that period. Then Opus 5 launched and the growth curve from September to October "looks great," Kim says. Someone leaked October ARR to the Wall Street Journal — Kim says it came in above $80 billion annualized — sentiment flipped, the valuation was reset lower, and the book is now six or seven times oversubscribed. Kim expects the IPO to "go bonkers."

The read on RSI — recursive self-improvement — is bullish for semiconductors broadly, not just Nvidia. Meta and OpenAI executives Kim has spoken to say they're taking every chip they can get regardless of vendor. OpenAI's head chip architect told him agents have been "a dramatic multiplier and accelerant" in chip design itself.

Jensen faces congressional testimony and ongoing investigations before any next chapter becomes possible. Kim doesn't think he's done — "work is the way he has fun" — but the timeline is at minimum six to twelve months out.