The AI chip war is intensifying, and it's no longer just about speed. Alphabet Inc. (NASDAQ:GOOGL) just delivered one of its strongest stock performances in years, riding a wave of investor excitement over next-gen AI models and a powerful in-house chip that now threatens Nvidia Corp.‘s (NASDAQ:NVDA) once-uncontested dominance in the artificial intelligence arms race.
Alphabet shares outperformed Nvidia by an eye-popping 30 percentage points in November, marking their most dominant one-month excess return since 2017.
Alphabet is the top performer among the Magnificent Seven in 2025, with shares up 70% year to date—more than double Nvidia's gains.
Fueling the rally is more than just hype: it's the debut of Gemini 3, Google's most advanced AI model yet, and the rising impact of Tensor Processing Units (TPUs), its custom-built chip designed to rival Nvidia's monopoly in AI compute.
Is this truly the moment when Nvidia's reign over AI chips begins to fade, making room for a new dominant force to emerge?
What Makes Google's TPUs Different Compared To Nvidia’s GPUs?
Gemini 3, released just eight months after its predecessor, signals a shift in the AI landscape. It blends language, vision and logical reasoning in a way that, according to experts, pushes the boundaries of what AI can do.
With 650 million monthly active users, Gemini is rapidly catching up to OpenAI's ChatGPT, and it's powered entirely by Google's TPUv7 chips, built in partnership with Broadcom Inc. (NASDAQ:AVGO), a direct rival to Nvidia's Blackwell GPUs.
Google's TPUs aren't just cheaper or faster—they're fundamentally different in architecture.
"The investor debate has now shifted to competition—between OpenAI and Google Gemini LLMs, and between Google/AVGO-based TPU and incumbent NVDA GPU-based chips," said Bank of America’s semiconductor analyst Vivek Arya in a note Monday.
While Nvidia's Blackwell offers top-tier performance for nearly every type of task, Google's chip shines in certain use cases and are particularly tailored for AI inference and training.
Arya acknowledged that Nvidia's Blackwell chips, including the latest B300/GB300, still deliver the most comprehensive performance across a wide range of AI tasks.
But in specific use cases—particularly for inference using floating point eight (FP8) precision—Google's TPUv7 is proving to be more efficient and cheaper to run.
Arya highlighted that TPUv7 can outperform Blackwell Ultra in performance per watt, delivering 5.42 teraflops per watt in FP8 workloads versus 3.57 teraflops per watt for Nvidia's chip.
In terms of cost, the TPUv7 is estimated at $3.50 per hour to rent internally, compared to Nvidia's $6.30 per GPU per hour.
This can lead to a total cost of ownership up to 40% lower for TPUv7 on the right workloads.
Why Nvidia May Still Have The Upper Hand
Despite its efficiency and price edge on certain metrics, TPUv7 has one major hurdle: limited availability.
While Nvidia's GPUs are available across Amazon.com Inc.‘s (NASDAQ:AMZN) AWS, Microsoft Corp.‘s (NYSE:MSFT) Azure, and Google Cloud Platform (GCP), TPUs are currently exclusive to GCP.
TPUs are locked into Google Cloud, which makes them less attractive to clients with massive datasets stored elsewhere. Moving data between clouds is expensive, so most customers choose compute that matches their data location.
"TPUs have been proven only in Google’s data center, while GPUs are available in multiple cloud environments including Google Cloud,” Arya said.
He also estimated that Google will spend $10 billion in 2025 on both GPUs and TPUs, suggesting that even internally, Alphabet sees a need for hybrid compute.
Arya believes that NVDA’s upcoming Vera Rubin platform, expected in the second half of 2026, could shift the competitive landscape again.
Are Nvidia Shares Cheap Now?
According to Arya, Nvidia shares trade now at compelling valuations following November’s decline.
Arya views Nvidia's current valuation as an opportunity. "NVDA is currently trading ~25x forward PE, right around its prior troughs in October 2023 and July 2022," he wrote. Historically, whenever Nvidia stock reached that level, it rebounded to between 30 and 40 times forward earnings in the following 3 to 6 months, with a five-year median of 37x.
He also noted that Nvidia is now trading at its widest ever discount—around 40%—relative to Broadcom, which currently trades at 42 times forward earnings.
In Arya's view, this discount reflects the market already pricing in a 10-point shift in AI market share toward Broadcom and Google by 2026/2027—a shift he believes may be overstated.
"We continue to believe the current situation is another overstated DeepSeek January 2025-like moment that provided a particularly attractive buying opportunity for AI semis," Arya added, referring to a temporary market panic that preceded a strong rebound in AI semiconductor stocks.
In summary, TPUs are winning in performance-per-watt and cost—but Nvidia's ecosystem, availability, and broad workload support keep it firmly in the lead, for now.
Read Next:
Image created using artificial intelligence via DALL-E.
