Sometimes the market sells first and asks questions later. That's basically what happened Tuesday when Nvidia Corp. (NVDA) and Advanced Micro Devices, Inc. (AMD) took a hit on news that Meta Platforms Inc. (META) might start using Alphabet Inc.'s (GOOGL) AI chips instead of relying exclusively on the usual suspects.
But here's where it gets interesting: a Google DeepMind researcher who actually builds these chips thinks Wall Street completely missed the point.
The Numbers Tell a Story, But Maybe Not the Right One
Nvidia closed down 2.59% at $177.82 on Tuesday, while AMD slid 4.15% to $206.13. After hours, Jensen Huang's chip giant dropped another 0.52%, and AMD fell 1.04% according to market data.
The selloff came after reports surfaced that Meta might tap Google's (GOOG) tensor processing units, or TPUs, for its data centers. Google has been quietly ramping up its push to compete directly with Nvidia in the AI chip arena. These TPUs are designed to deliver optimized performance and cost efficiency across the full spectrum of AI workloads, from training massive models to running inference at scale.
To most traders, this looked like classic market share erosion. If Meta diversifies away from Nvidia, that's fewer chips sold, right? Simple math.
Except One Problem: The Entire Premise Is Wrong
That's the view from Amit Yazdan, a Google DeepMind researcher who works on TPU design. He took to X to share his professional opinion of the market's reaction: "The selloff shows how clueless the market is about hardware and the demand."
It's not every day you see someone who actually engineers AI chips calling out Wall Street's logic in real time. Yazdan's point is straightforward: demand for AI hardware is so absurdly high right now that one company building more chips doesn't mean another company sells fewer. The pie is growing faster than anyone can bake it.
Nvidia's Response: Gracious Winner Energy
Nvidia played this perfectly. The company issued a public statement congratulating Google on its AI progress while casually mentioning it still supplies Google with GPUs. Very diplomatic. Very classy.
Then came the flex. Nvidia emphasized that it remains "a generation ahead" of competitors and is the only platform capable of running every major AI model wherever computing happens. Translation: congratulations on your nice chips, but we're still the gold standard.
And Then There's Broadcom
While everyone focused on Nvidia and AMD, Jim Cramer pointed out that Broadcom Inc. (AVGO) might be the real winner here. Broadcom CEO Hock Tan sits on Meta's board, and if Meta needs custom chip design work for Google's TPUs, Cramer thinks they'll likely contract with Broadcom to make it happen.
It's a reminder that in the AI chip world, the picks-and-shovels businesses can sometimes outperform the headline names.
Meanwhile, market data shows that GOOGL continues to exhibit strong price momentum across short, mid, and long-term timeframes, suggesting investors remain bullish on Google's broader AI strategy even as traders temporarily freaked out about competitive dynamics.
The bigger picture? AI infrastructure spending isn't a zero-sum game right now. When a DeepMind engineer who designs these chips tells you the market doesn't understand demand, maybe it's worth listening.