TMTPOST -- Nvidia Corp. shares plunged Tuesday following reports that Google is gaining ground in the artificial intelligence ( AI ) chip market, with its tensor processing units ( TPUs ) emerging as a viable alternative to Nvidia's dominant graphics processing units ( GPUs ) . The sell-off reflects growing investor concern about intensifying competition in the AI semiconductor space, though analysts suggest the market remains large enough to support multiple players.
The stock decline came after The Information reported Monday that Meta Platforms is considering using Google's TPUs in its data centers in 2027, and may rent TPUs from Google Cloud as early as next year. Nvidia shares fell as much as 7.1% before recovering to close down 2.6%, erasing $115 billion in market value. Shares of Advanced Micro Devices ( AMD ) dropped 6%, while Broadcom, which helps Google design its TPUs, rose nearly 1.9%.
Google parent Alphabetstock gained 1.6% Tuesday, building on Monday's 6% rally. The company has added nearly $1 trillion in market capitalization since mid-October, propelling it past Microsoft to become the third-largest company by market cap and closing in on Apple.
The market moves underscore shifting dynamics in the AI infrastructure race, with Google emerging from what analysts have called a "sleeping giant" status following the launch of its well-received Gemini 3 model and expanding commercial adoption of its custom chips.
Google Expands TPU Strategy Beyond Cloud
Google is intensifying efforts to compete directly with Nvidia by pitching customers on using TPUs in their own data centers, according to The Information on Monday. The company has approached Meta and major financial institutions about the arrangement as part of a program called TPU@Premises.
Meta is currently in talks with Google about spending billions of dollars to use TPUs in its data centers in 2027, the report cited a person involved in the discussions. The social media giant, which projects capital expenditures of $70 billion to $72 billion this year, has indicated its 2026 spending will be "noticeably larger" as it expands AI efforts.
Some Google Cloud leaders have suggested the expanded TPU business could capture as much as 10% of Nvidia's annual revenue, potentially generating billions of dollars in additional sales. Google has marketed TPUs as more cost-effective than Nvidia chips and better suited for customers with stringent security and compliance requirements.
Wall Street Debates Market Share Versus Market Growth
Analysts are divided on whether Google's momentum threatens Nvidia's dominance or simply reflects surging overall demand for AI compute capacity. Bernstein analyst Stacy Rasgon argued that investors are fixating on the wrong question by debating whether GPUs or TPUs will prevail.
"Right now the real question should really be, 'Is the opportunity in front of us still big, or is it not?'" Rasgon wrote. He noted that the AI hardware market likely hasn't reached a "mature, saturated" stage, making market size more important than market share. "If it's big, both GPU and ASIC should thrive," he said.
Mizuho analyst Jordan Klein pushed back against "winner-take-all" sentiment, calling the AI competition "a marathon that will see many different lead changes." Nvidia CEO Jensen Huang disclosed last month that the company has visibility into $500 billion in revenue from its Blackwell and Rubin AI platforms through 2026.
Google's AI Resurgence Reshapes Competitive Landscape
Google's recent advances signal a broader comeback in the AI race after facing criticism for falling behind OpenAI following ChatGPT's 2022 launch. The company's Gemini 3 model has risen to the top of closely watched AI leaderboards, earning praise from OpenAI founding member Andrej Karpathy as "clearly a tier 1 LLM."
Bloomberg reported that Google has become "a sleeping giant that is now fully awake," leveraging advantages including its search data corpus, profitable core business, and integrated computing infrastructure. The company consolidated its AI efforts under DeepMind leader Demis Hassabis in early 2023, focusing on foundational models that compete with OpenAI and Microsoft.
Google's TPUs represent one of the few successful application-specific integrated circuit programs to reach commercial scale. A deal announced in October with Anthropic to provide up to one million TPUs validated the technology's viability. However, TPUs remain primarily attractive to large-scale customers and can only be accessed through Google's cloud service, potentially limiting adoption compared to Nvidia's more flexible GPUs.
The chip market dynamics extend beyond technology to business relationships. According to The Information, Huang has moved aggressively to counter Google's TPU expansion, announcing investment deals with both Anthropic and OpenAI after they explored using Google chips.
