Nvidia’s board spent Monday investigating its own CEO. Google spent the same day launching more AI chips aimed at Nvidia’s customers. One company fought a crisis. The other expanded its catalog.
Nvidia shares fell more than 4% on Monday and closed at their lowest level since April 7. The drop followed news that Nvidia’s board would probe CEO Jensen Huang’s relationship with an alleged Chinese spy.
The rest of the sector rallied. The iShares Semiconductor ETF jumped more than 7%, and Alphabet rose about 1%. Investors didn’t dump AI chips. They dumped uncertainty at one company.
Google Launches Four AI Chips in a Single Year
Google unveiled two new tensor processing units on Monday, the TPU 8tx and the TPU 8i Max. They join the TPU 8t and TPU 8i, which Google introduced at Cloud Next in April.
For the first time, Google has launched four TPU systems in one year. It used to release about one data center TPU annually. Both new chips ship next year.
Each AI chip serves a different buyer:
- TPU 8tx: keeps data encrypted and pairs with AMD processors. It targets governments, banks, and other regulated firms.
- TPU 8i Max: fuses two chips into one package. It handles long context windows, the amount of information an AI model juggles at once.
Nvidia sells one powerful engine for nearly every job. Google now offers a menu, with each chip tuned to a specific AI workload.
AI Chip Sales: Forecast Meets Track Record
Bloomberg Intelligence expects TPU sales to grow from about $12 billion this year to more than $135 billion by 2028. That estimate leaves out chips Google rents through its cloud. It implies more than 11-fold growth in two years.
Nvidia’s data center business grew nearly eightfold in the first two years of the AI boom. Revenue climbed from $15 billion in fiscal 2023 to $115 billion in fiscal 2025.
The comparison needs context. Nvidia earned those numbers while building the AI chip market. Google’s numbers remain a forecast in a market Nvidia has already proven. Forecasts can slip. Booked revenue can’t.
The timing matters for Alphabet too. The forecast lands weeks after Alphabet’s $500 billion selloff, when investors balked at the cost of its AI buildout. Strong TPU demand gives Google a clearer path to earn that spending back.
Factory data supports the demand side. Epoch AI estimates that TPU shipments doubled in computing power in a single quarter this year, reaching roughly 800,000 chips. Raw compute still shapes what makes AI smarter, so every jump in supply feeds the next wave of models.
Anthropic Anchors Google’s AI Chip Demand
Anthropic leads the customer list. Its IPO filing shows about $161 billion in lease payments tied to TPUs. Raymond James says that makes Anthropic the largest custom-chip customer of Broadcom, Google’s design partner. Meta, OpenAI and Oracle also buy TPUs.
That list changes how the market views custom silicon. It no longer looks like a side project for cloud giants. Frontier AI labs now stake their compute budgets on it.
AI Agents Rewrite the Chip Math
Nvidia still wins on raw power. Its next-generation Rubin chip delivers 35 petaflops, against about 10 for the TPU 8i and about 20 for the TPU 8i Max, per Yahoo Finance calculations. Nvidia also sells far more than chips today, all the way up to a container-sized Rubin AI factory.
Google competes on price. SemiAnalysis found that Google’s Ironwood TPU produces up to nearly three times more AI output per dollar than Nvidia’s Blackwell chips when running AI agents. Nvidia’s top system still leads when speed matters most.
That gap explains Google’s strategy. AI agents run nonstop. They chain tasks, call tools, and loop through steps for hours. Under that load, cost per task beats peak benchmark scores.
Cheaper agent runs also help companies stuck in the AI ROI gap, where usage runs high but returns lag. If agents become the default way businesses use AI, efficiency sells better than petaflops. Google designed its 2026 lineup around that bet.
Nvidia’s AI Crown Faces a Nov. 17 Test
Nvidia remains the market leader. It still sells far more AI chips than any rival, and forecasts aren’t sales.
Its next test arrives Nov. 17, when Nvidia reports third-quarter results. A requested closed-door House briefing on Huang follows a day later. Investors will watch for signs that customers now split orders between suppliers.
Bernstein’s Stacy Rasgon framed the stakes in a client note: “Jensen is not an ordinary CEO; he IS NVIDIA.”
What Changes for the AI Chip Market
Nvidia’s moat never rested on speed alone. It rested on the lack of a real alternative. Google now offers four, each designed and priced for a specific AI job.
The AI chip race isn’t crowning a new winner. It’s splitting one supplier’s market among several. Google sits closest to the biggest share of whatever Nvidia gives up.
Related: Google Put AI Chips in Space. The Real Problem Isn’t Power
