Google’s $90bn AI bet rides on custom chips as bubble fears loom

Google's $90bn AI bet rides on custom chips as bubble fears loom

Google CEO Sundar Pichai has staked the company's future on a high-risk, high-reward gamble: a $90 billion annual investment in artificial intelligence, anchored by its proprietary Tensor Processing Units (TPUs)-custom-built chips designed to dominate the next era of computing. Speaking from Google's California headquarters, Pichai framed AI as "the most profound technology humanity has ever worked on," but acknowledged even tech giants like Google may not be "immune" to a potential market correction.

The TPU: Google's secret weapon in a chip arms race

Tucked behind trees at the edge of Google's Mountain View campus, a restricted lab hums with the sound of cooling systems struggling to tame the heat of trillions of calculations. Here, Google's TPUs-specialized chips optimized for AI workloads-undergo testing before deployment. Unlike general-purpose CPUs or GPUs (like those from Nvidia), TPUs are application-specific integrated circuits (ASICs), tailored to accelerate Google's AI models, including its newly launched Gemini 3.0.

Pichai's confidence in TPUs reflects a broader industry scramble. From Elon Musk and Larry Ellison reportedly begging Nvidia's Jensen Huang for more GPUs at a Palo Alto dinner to Meta and Microsoft racing to build "AI factories"-data centers packed with high-performance chips-the sector's mantra is clear: scale or fail.

Bubble warnings clash with $15 trillion valuations

The AI gold rush has minted trillions in market value. Nvidia alone is now worth over $5 trillion, while OpenAI's valuation hit $500 billion. Collectively, the "Magnificent 7" tech stocks (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, Tesla) account for one-third of the S&P 500's total value-a concentration surpassing even the dotcom bubble, per the IMF.

Yet cracks are showing. The Bank of England warned of "stretched" tech valuations, and OpenAI's Sam Altman admitted parts of the industry feel "bubbly." Coreweave, an AI infrastructure startup supplying OpenAI, saw its shares plummet 26% this month. Pichai conceded the sector may be "overshooting" but insisted Google's deep pockets-funding TPUs and data centers without reliance on debt-insulate it from the worst risks.

"It's both rational and there are elements of irrationality through a moment like this."

Sundar Pichai, CEO of Google

Energy dilemma: Can AI and climate goals coexist?

By 2030, global data centers could consume as much electricity as all of India did in 2023, the IMF projects. Pichai, when pressed on reconciling AI's energy demands with Google's pledge to run on 95% low-carbon power by 2030, called it "possible" but urged governments to "scale up infrastructure" to avoid "constraining the economy."

The tension underscores a broader paradox: AI's promise of transformative benefits-from healthcare to education-collides with its voracious appetite for power and resources. "You don't want to constrain an economy based on energy," Pichai said. "That will have consequences."

AGI: The 'glittering prize' driving the frenzy

Beyond profits, Silicon Valley's elite are chasing artificial general intelligence (AGI)-machines that match or surpass human cognition. OpenAI's Sam Altman has floated an $8 trillion global investment in AI over the next decade, while Pichai positions Google's TPUs as the hardware backbone for that future.

Yet history offers cautionary tales. The dotcom crash wiped out firms like WorldOfFruit.com but spared Amazon, which rebounded from a $4 billion valuation to $2.4 trillion today. "Even in the toughest crashes," Pichai implied, "the strongest players endure." The question is whether AI's physical infrastructure-data centers brimming with TPUs and GPUs-will become the new oil rigs of the 21st century, or relics of a burst bubble.

Geopolitical stakes: US vs. China

While Beijing funds AI centrally, the U.S. relies on a "messy but productive" free-market approach, one Silicon Valley insider noted. Nvidia's GPUs and Google's TPUs currently give the U.S. an edge in silicon, but the race for supremacy-with implications for military, economic, and technological dominance-leaves little room for hesitation.

As Pichai put it: "Every decade, there's an inflection point. Now it's clearly the era of AI." Whether that era ends in revolution or ruin may hinge on who controls the chips-and whether the world can afford their cost.