Nvidia Was Already in Trouble Before the Jensen Huang Mess
AI-assisted digest, checked against the original source before publishing. How we work
Nvidia hands the wheel to CFO Colette Kress as rivals press on training, inference, and China.
Nvidia is changing hands at a moment when competitors are closing in on the market it has dominated. Colette Kress, the company's CFO, is running it following Jensen Huang's departure, and the handoff lands on a fast-shifting schedule. "Everything sold today was designed two-to-three years ago," the CEO of a large AI company told WIRED on condition of anonymity, adding that whoever leads Nvidia permanently must now chart where the chip industry will be two years from now.
The shape of demand is shifting too. Patrick Moorhead, chief analyst at MoorInStrat, says the rise of AI agents makes inference — running models that have already been trained — more important, an area where Nvidia holds less of an edge over cloud hyperscalers and startups. He also expects growing demand for general-purpose CPUs as assistants and agents spread, a category Nvidia barely competes in. Nvidia did acquire a startup making specialized chips, called Grok, in December 2025.
Nvidia's lead traces back to Huang's own foresight. He saw AI's promise in the mid-2010s and backed it early: funding academic research, developing software to support AI engineering, and handing hardware to labs and startups. When ChatGPT took AI mainstream in 2022, chip sales skyrocketed. The challenges now accumulating may not be existential, but they add up, especially on a global scale.
What to watch: whether Kress keeps the seat or a permanent successor emerges, and how that leader steers products locked in years earlier; whether agent-driven inference and CPU demand actually erode Nvidia's training advantage; and whether export controls — plus the uncertain flow of chips routed through third countries — leave Nvidia's China floor where it stands.
Key facts
- • AMD's AI-chip share rose from 2% in 2022 to 7–10% in 2026.
- • Nvidia revenue grew from $26.91B in January 2022 to $215.94B in January 2026.
- • AMD recently acquired Taalas, which bakes model weights into silicon.
- • Nvidia was founded in 1993 to design PC graphics chips.
Why it matters
Builders should expect a more multi-vendor AI stack: hyperscaler silicon and specialized startups are viable for serving and inference, but porting costs (code rewrites, custom tooling) remain real. Planning horizons matter — today's shipping accelerators were designed two to three years ago, so roadmap bets made now by Nvidia's next leader determine the competitive field of 2028.
Spotted an error? Send a correction.
Get the day's AI news in one email
The Daily Epoch · 60-second read · unsubscribe anytime