GOOGL Stock Leans On TPU Strength Amid Shaky 2026 — Even Elon Musk Admits They're Underrated

• “TPUs are underrated,” Musk said on X.
• Custom chips are becoming an increasingly important growth engine for Google as well as its peer, Amazon.
• GOOGL was the top performer among the “Magnificent Seven” last year and is up 8.5% year to date, outpacing the S&P 500.

Google announced its eighth-generation Tensor Processing Unit (TPU) chips on Wednesday, featuring impressive specifications, and earned a hat tip from Tesla CEO Elon Musk. 

“TPUs are underrated,” Musk wrote on X, in response to Alphabet CEO Sundar Pichai’s post showing off the new TPU 8t and TPU 8i chips. 

Custom chips are becoming an increasingly important growth engine for Google. Google Cloud revenue surged 48% in the fourth quarter, with management pointing to AI infrastructure — including its TPU lineup — as a key driver. The company introduced its first TPU back in 2016.
 

Peers are moving in the same direction. Amazon, Microsoft and Meta Platforms are all expanding their custom silicon efforts to reduce reliance on costly Nvidia hardware. Meta rolled out its next-generation MTIA chips last month, while Tesla recently said its upcoming data center chips, designed for its Terafab Austin complex, will be built on Intel’s advanced 14A process.

Big Tech Chip Wars

Morgan Stanley, in a December 2025 note, projected that Google is on track to ship 5 million TPUs in 2027 and 7 million in 2028, and that every 500,000 units sold externally could add about $13 billion in revenue, 11% to Google Cloud, and roughly 3% to Alphabet’s bottom line.

Earlier this month, Amazon disclosed that its chips business had reached an annualized revenue run rate of over $20 billion, doubling from the $10 billion reported alongside fourth-quarter results.

2016–2020: Google leads (TPUs), Tesla builds FSD chip

2020–2023: Amazon scales Trainium and Inferentia chips portfolio

2023–2026: Microsoft and Meta aggressively enter with their custom chips

2026: Major players are launching chips for specific AI workloads

GOOGL’s AI, Chip Momentum

Google’s latest TPUs reflect that split: the TPU 8t is designed for AI model training, while the TPU 8i is optimized for inference. Inference refers to the stage where a trained model applies what it has learned to make real-world predictions or decisions — essentially, what happens in the background when AI responds to user queries and performs tasks.

Google claims that the 8th-gen chips offer three times faster AI model training, 80% better performance per dollar, and the ability to connect 1 million+ TPUs in a single cluster.

The TPU roadmap has helped drive gains in Google’s stock, fueled largely by optimism around the Gemini 3 AI model, its AI partnership with Apple, and strong growth in the cloud unit. GOOGL was the top performer among the “Magnificent Seven” last year and is up 8.5% year to date, outpacing the S&P 500’s 4.3% gain.

For updates and corrections, email newsroom[at]stocktwits[dot]com.

Read Next: Why Did TXN Stock Jump Over 11% Overnight?