TLDR
- Google’s TPU chips have gained attention after Gemini 3’s release, with Alphabet stock up 12% while Nvidia dropped 3.4%
- Meta Platforms is reportedly in talks to buy Google’s TPU chips for AI data centers, traditionally Nvidia’s territory
- Google’s Alphabet is now worth $3.86 trillion, making it the world’s third largest company behind Nvidia and Apple
- Nvidia still dominates with 80% market share and 73% gross profit margin, protected by its CUDA software ecosystem
- Alphabet raised 2025 capital expenditure guidance to $91-93 billion from $85 billion, with further increases expected in 2026
Google has suddenly become the center of attention in the artificial intelligence chip market. The company’s stock has jumped 12% since launching its Gemini 3 AI model on November 18.
Nvidia stock fell 3.4% during the same period. The shift came after reports that Meta Platforms is considering buying Google’s Tensor Processing Units for AI data centers.
Alphabet now holds a market value of $3.86 trillion. This makes it the third largest company globally, trailing only Nvidia at $4.38 trillion and Apple at $4.1 trillion.
Google developed its first TPU chip in 2015. The company faced the same challenge other tech giants experience today: Nvidia hardware was expensive and hard to obtain in large quantities.
The TPU powered Google products like Maps, Photos, and Translate before anyone outside the company knew it existed. Google now operates its seventh generation of TPU chips.
Apple trained its Apple Intelligence models on TPUs. AI startup Anthropic, valued at $350 billion, also uses TPUs as part of its multicloud strategy.
Nvidia’s data center revenue has grown 2,400% since 2021. The company achieved a 73% gross profit margin in the third quarter, representing nearly a 300% markup.
The TPU Advantage
TPUs focus on one task: matrix math for deep learning. They handle this specific workload more efficiently than Nvidia’s GPUs under the right circumstances.
Nvidia’s graphics processing units offer more versatility. They split tasks into many pieces and run them simultaneously, useful for gaming and various AI calculations.
Deep learning has driven AI research for over a decade. This focus led to the large language models powering current chatbots and coding assistants.
Customers want to reduce their dependence on Nvidia. But alternatives like TPUs still don’t meet all their needs, which explains Nvidia’s continued dominance.
Intel previously controlled the data center chip market with 65% revenue share in early 2021. By 2024, Nvidia captured over 80% of the market while Intel dropped to single digits.
Software Protection
Nvidia built its CUDA software starting in 2004. The program allows developers to use common programming languages that GPUs can understand.
Most AI researchers know how to use CUDA. Far fewer understand Google’s software, which remains less mature.
Nvidia’s market share depends on CUDA, not just its hardware. Competition will become material when Nvidia’s gross margins start declining, indicating price cuts to protect sales.
Alphabet reported third quarter earnings of $2.87 per share, beating estimates by 27%. Revenue reached $102.35 billion, up 16% year over year.
Google Cloud revenue surged 33.5% to $15.16 billion. The cloud division signed more deals worth over $1 billion through Q3 2025 than in the previous two years combined.
Google Cloud’s backlog reached $155 billion, up 46% from the prior quarter. About 70% of Google Cloud customers now use Alphabet’s AI products.
Search and other revenues increased 14.5% to $56.57 billion. YouTube advertising revenue grew 15% to $10.26 billion.
Broadcom stock has risen 16% since the Gemini 3 launch. The company helps design Google’s TPU chips.
Alphabet raised its 2025 capital expenditure guidance to $91-93 billion from $85 billion. The company expects capital expenditure to increase further in 2026.


