Nvidia Claims GPUs “Generation Ahead” of Google’s AI Chips

Nvidia asserted on Tuesday that its AI technology remains a generation ahead of the competition, addressing Wall Street’s concerns that the company’s leadership in AI infrastructure could be challenged by the rise of in-house AI chip development at companies like Google.

“We’re delighted by Google’s success — they’ve made great advances in AI, and we continue to supply to Google,” Nvidia stated in a post on X. “NVIDIA is a generation ahead of the industry — it’s the only platform that runs every AI model and does it everywhere computing is done.”

This communication came after Nvidia’s stock experienced a 3% dip on Tuesday, triggered by a report highlighting the possibility of Meta, a significant Nvidia client, potentially partnering with Google to leverage its Tensor Processing Units (TPUs) for its expansive data centers. This scenario raises questions about the future diversification of the AI chip supply chain and its impact on Nvidia’s stronghold.

Nvidia countered these concerns by emphasizing the superior flexibility and power of its chips compared to Application-Specific Integrated Circuits (ASICs), like Google’s TPUs, which are tailored for specific companies or dedicated tasks. Nvidia’s current flagship architecture is the Blackwell. While TPUs are undeniably powerful and optimized for Google’s AI workloads, Nvidia argues that its GPUs offer broader utility across various AI models and deployment environments.

“NVIDIA offers greater performance, versatility, and fungibility than ASICs,” Nvidia declared in its post. This contention focuses on the ability of Nvidia’s GPUs to adapt to evolving AI algorithms and diverse computational needs, contrasting with the more rigid nature of ASICs.

Currently, Nvidia commands over 90% of the artificial intelligence chip market with its high-performance GPUs, according to industry analysts. However, Google’s internally developed chips have garnered increasing attention from the broader tech community, positioned as a formidable and potentially cost-effective alternative to Nvidia’s Blackwell series. This situation underscores a strategic trend among hyperscale cloud providers to achieve greater vertical integration and control over their AI infrastructure.

A key differentiator is that Google doesn’t sell its TPUs commercially to other companies. Instead, it utilizes them for its internal operations and provides access to them through its Google Cloud Platform, positioning TPUs as a service rather than a product. This strategy reinforces Google’s overall cloud ecosystem and allows customers to leverage cutting-edge AI hardware without the upfront investment.

Earlier in November, Google unveiled Gemini 3, its highly anticipated AI model that has been lauded for its capabilities. Notably, Gemini 3 was trained on Google’s TPUs, rather than Nvidia GPUs. This demonstrates the increasing viability of non-Nvidia hardware in training state-of-the-art AI models and presents a growing challenge to Nvidia’s dominance.

“We are experiencing accelerating demand for both our custom TPUs and Nvidia GPUs,” a Google spokesperson stated, underscoring the company’s balanced approach to AI infrastructure. “We are committed to supporting both, as we have for years.”

Nvidia CEO Jensen Huang addressed the rising TPU competition during a recent earnings call, acknowledging that Google remains a customer for Nvidia’s GPUs and that Gemini can be deployed on Nvidia’s technology. This reinforces the existing, albeit evolving, partnership between the two tech giants.

Huang also shared insights from his communication with Demis Hassabis, the CEO of Google DeepMind, regarding the industry’s observation that scaling AI models with more chips and data leads to improved performance. This principle, known as “scaling laws,” suggests that increased computational power directly translates to more powerful AI, potentially fueling sustained demand for advanced chips and systems like those offered by Nvidia.