NVIDIA's counterattack…"A generation ahead of Google's AI chips"[Stocks+]

As news of Meta's investment in Google's AI chips emerged

NVIDIA: "Better performance, and used in many places"

With reports that Meta is considering a multibillion-dollar investment in Google's artificial intelligence (AI) chips, NVIDIA has struck back, saying its GPU technology is "a generation ahead of the industry." Analysts say this marks a direct response to the recently highlighted competition over Google's AI chips.

On the 25th (local time), NVIDIA said on X (formerly Twitter), "We are pleased with Google's success. They have achieved remarkable results in the AI field, and we will continue to supply Google going forward." It added, "NVIDIA is the only platform a generation ahead of the industry, and it has the technology to run all AI models in all computing environments."

Earlier, it was reported that Meta is discussing plans to introduce Google's Tensor Processing Units (TPUs) into its data centers from 2027. It is also reported that next year it is considering the option of leasing chips from Google Cloud.

NVIDIA holds more than 90% of the AI chip market based on graphics processing units (GPUs). However, in recent weeks Google's in-house developed chips have been mentioned as a high-performance, high-efficiency alternative, drawing investors' attention.

Reports of Meta's possible adoption of TPUs raised concerns that NVIDIA's market dominance could be affected. In response, NVIDIA emphasized that ASIC-based chips like TPUs are designed for specific functions and corporate uses and therefore lack versatility, stressing that they are less flexible compared to its chips. NVIDIA's latest GPU generation is called 'Blackwell.' NVIDIA added, "NVIDIA offers better performance, versatility, and replaceability compared to ASICs."

Google does not sell TPU chips to external companies, but uses them to operate its own services and offers them to corporate customers through Google Cloud in a leasing form. Earlier this month, Google unveiled its latest AI model 'Gemini 3' trained on TPUs, which received favorable reviews from experts.

A Google spokesperson said in a statement that "Demand is rapidly expanding for both custom TPUs and NVIDIA GPUs," and "We have supported both technologies simultaneously and will continue to do so."

Meanwhile, Jensen Huang, NVIDIA's CEO, said at an earnings announcement earlier this month in response to a question about intensifying competition from TPUs, "Google is a major customer of NVIDIA GPUs, and the Gemini model can also run on NVIDIA technology." He also said he is in close contact with Demis Hassabis, CEO of Google DeepMind.

Huang said that in a recent message Hassabis sent him, he emphasized that the industry's core premise, the 'scaling laws,' "are still valid." Scaling laws refer to the concept that the input of more chips and data produces more powerful AI models. NVIDIA expects that these scaling laws will support long-term demand growth for its chips and systems.

New York = Park Shin-young, correspondent nyusos@hankyung.com