NVIDIA says its GPUs remain 'generation ahead' of Google chips
What's the story
NVIDIA has defended its position as a leader in the artificial intelligence (AI) hardware space, despite growing investor concerns over Google's increasingly powerful tensor processing unit (TPU) chips. The tech giant maintains that its own AI hardware is a full generation ahead of the industry standard and continues to support every major AI model across different sectors. This comes after a 3% fall in NVIDIA's share price following reports that Meta might consider using Google's TPUs in its data centers.
Hardware superiority
NVIDIA highlights versatility and performance of its AI hardware
In a post on X, NVIDIA acknowledged Google's advancements but emphasized that its platform is unparalleled in terms of versatility and performance. The company stressed that its chips can run every major AI model across various environments, unlike application-specific chips such as Google's TPUs which are meant for narrower use cases. This claim was backed by the example of NVIDIA's latest Blackwell chips, which form the backbone of its accelerated computing stack.
Chip advantages
NVIDIA's GPUs offer superior flexibility and fungibility
NVIDIA asserted that its GPUs provide greater flexibility and "fungibility" than ASIC-style hardware. Analysts estimate that the company still holds over 90% of the AI chip market. However, Google has been making strides with its state-of-the-art Gemini 3 model, which was trained entirely on TPUs instead of NVIDIA GPUs. The tech giant also provides access to TPUs through Google Cloud services.
Market response
Google acknowledges rising demand for TPUs and NVIDIA GPUs
A Google spokesperson confirmed the increasing demand for both TPUs and NVIDIA GPUs, reiterating the company's commitment to supporting both technologies. During a recent earnings call, NVIDIA CEO Jensen Huang addressed the TPU competition by noting that Google is still a major customer of his company's GPU chips. He also revealed his discussions with DeepMind CEO Demis Hassabis, who reaffirmed that AI scaling laws still hold true.