The News
Nvidia on Tuesday unveiled plans for more powerful chips, an artificial intelligence model for humanoid robots, and “personal AI supercomputers.”
The US semiconductor giant’s announcements at an event CEO Jensen Huang dubbed the “Super Bowl of AI” — some also called it “AI Woodstock” — come at a critical time for the company.
Nvidia’s stock sank about 3% Tuesday after Huang delivered his keynote speech — weaving product announcements into what was essentially a lecture on the economics and technical challenges of the artificial intelligence industry.
The overarching thesis was that the appetite for AI is only growing, and the only real limiting factor is the raw power needed to run increasingly gargantuan data centers.
Nvidia has faced increasing skepticism since late January, when Chinese company DeepSeek released R1, an innovative AI model that highlighted how the technology could be run more efficiently. The uncertainty of the Trump economy and stepped up competition in the form of custom-designed chips from hyperscalers like Amazon, Google, and Microsoft have raised even more questions.
Reed’s view
The tech industry is crazy about Nvidia and its leather-clad CEO. Wall Street, however, is not.
Huang’s keynote was not an overt response to Nvidia’s critics, but the point was hardly subtle.
Instead of ignoring DeepSeek, which caused Nvidia to set the record for the largest ever single-day stock market loss, he used R1 as the benchmark for performance gains in Nvidia’s own chip architecture, showing how well it ran on different versions. The message: even the world’s most efficient models are not anywhere close to affecting the demand for compute power.
At one point, he showed a graphic of a data center. In the next slide, half of the server racks had vanished, representing how his customers can now buy fewer Nvidia chips and get the same amount of compute.
“Buy fewer of my products” isn’t something CEOs usually say. But the illustration provided a setup for a joke. Nvidia’s customers would use those savings to buy more, filling their data centers to the brim. “I’m the chief revenue destroyer,” he joked.
It was a brutal jab at the post-DeepSeek logic that efficiency would reduce the demand for compute power. Instead, efficiency is a product Nvidia sells to customers, and that won’t change until the industry is no longer constrained by the limits of energy production. The pie is also growing rapidly, and Nvidia is unusually nimble for a company that’s been around since 1993.
Its share price is a way to gauge how the world outside that stadium sees the fundamental dynamics of the AI industry. And right now, it appears it could probably use some more lectures.