Nvidia is gearing up for the inference war by developing a chip built on Groq technology for OpenAI and AI agents.

New Nvidia chip – a step toward faster AI applications

Nvidia plans to unveil a new processor in March that is specifically designed to accelerate inference (computations of already trained models) and aimed at clients such as OpenAI. This marks a shift for the company, which had previously focused mainly on training AI systems.

What will appear
- The new platform will be showcased at the GTC (Graphics Technology Conference) developers’ conference.
- The core of the platform will include a chip from startup Groq – one of the fastest inference solutions available.
- The platform is intended to compete with Google’s and Amazon’s own accelerators, which have already achieved significant success.

Why it matters
With the growing popularity of “web‑coding” (generating code via AI from text), demand for fast and energy‑efficient infrastructure is rising.

For OpenAI the new processor becomes a key factor:
- In January the company signed an agreement with Cerebras for an inference‑oriented chip that, according to their assessment, outperforms competitors.
- On the eve of OpenAI’s announcement of a major deal with Nvidia, they secured “dedicated inference capacity” and invested $30 billion from the “green” manufacturer.

Competitive landscape
Company Product Focus
Nvidia GPU families Hopper, Blackwell, Rubin Training and inference
Google TPU Inference
Amazon Trainium Inference
Cerebras Specialized chip Inference
Groq Chips with “language processing blocks” Inference

Nvidia holds over 90 % of the GPU market share, but its chips are often considered expensive and power‑hungry for inference tasks. This attracts attention from companies seeking alternatives.

Key applications
- Code generation: services Claude Code (Anthropic) and Codex (OpenAI) use Amazon and Google cloud infrastructures, but Codex plans to move to Nvidia’s new platform.
- Ad targeting: Meta collaborates with Nvidia for AI system inference, where central processors prove most effective.

Conclusion
Nvidia is not only expanding its product line but also entering direct competition with industry giants. The new chip, integrated with Groq’s startup technology, could become the decisive factor for companies striving for faster and more economical AI model processing.