The man who made Nvidia a $3 trillion company just walked out the door, and the vultures are already circling.

The Summary

The Signal

Jensen Huang didn't just run Nvidia. He WAS Nvidia. For 33 years, the leather jacket-wearing CEO turned a graphics chip company into the backbone of the AI revolution. His abrupt departure leaves Colette Kress, the company's CFO since 2013, holding the bag as interim CEO. Bloomberg's framing is telling: "interim" suggests the board doesn't know what comes next.

The markets hate uncertainty, and founder exits are the ultimate uncertainty event. Kress brings financial discipline and institutional knowledge, but she's inheriting a company at an inflection point. Nvidia's dominance in AI training chips, the H100s and H200s powering every major AI lab, isn't guaranteed forever.

"Nvidia will remain dominant in the near term, but its market share is likely to decline as more specialized AI chips emerge."

Callosum co-founder Dan Akarca isn't predicting Nvidia's collapse. He's predicting fragmentation. His company builds software that routes AI workloads to the processors best suited for each task. Translation: the one-size-fits-all GPU era is ending. When Amazon builds custom Trainium chips, Google runs on TPUs, and startups like Groq target inference workloads, Nvidia's 90%+ market share becomes a target, not a moat.

The leadership vacuum accelerates this. Huang's genius wasn't just technical vision. It was ecosystem control. CUDA, Nvidia's software layer, locked developers into its hardware. That stickiness bought time while competitors played catch-up. But ecosystems need constant tending, and Kress inherits a garden that was entirely Huang's design.

Key competitive pressure points:

  • Specialized inference chips undercutting Nvidia on cost and energy efficiency
  • Hyperscalers building custom silicon to reduce dependence
  • Software abstraction layers like Callosum's making it easier to switch between chip architectures

The Implication

Watch what Nvidia does in the next 90 days. If Kress stays interim for more than a quarter, the board is paralyzed. If they move fast to name a permanent CEO from outside, they're admitting Huang's bench wasn't deep enough. Either scenario gives competitors a window.

For anyone building in the agent economy, this matters more than it looks. Nvidia's chips are the rails. If those rails fragment, if inference gets cheaper and training diversifies across architectures, the cost curve for running AI agents changes. That's the real story. Not who sits in Huang's chair, but whether the throne still matters.

Sources

Bloomberg Tech