Nvidia Shares Rise After Saying Blackwell Chips Power OpenAI's New GPT-6 Astra Ultrafast

Nvidia shares rose on Thursday after the chipmaker said its Blackwell GPUs are powering OpenAI's new GPT-6 Astra Ultrafast, which can generate tokens at up to eight times the speed of its predecessor.

The announcement puts Nvidia's latest AI architecture at the centre of another major OpenAI deployment as demand shifts towards faster inference for coding, tool use, and interactive AI applications.

Nvidia stock gained 1.09% on 1st October, closing at $230.86. Shares traded as high as $232.29 during the session after opening at $229.98, according to historical market data.

GPT-6 Astra Ultrafast Receives Inference Boost

Nvidia said GPT-6 Astra Ultrafast runs on its Blackwell GPUs and is available through the OpenAI API, as well as to eligible ChatGPT Work and Codex users. The company said the Ultrafast mode can deliver up to eight times faster token generation than Astra Standard.

The performance improvement is aimed at workloads where latency compounds across multiple steps. Coding agents, for example, can repeatedly generate code, call tools, inspect results, and make further decisions. Hence, faster inference can reduce the waiting time between those steps.

OpenAI is also using its own models to optimise the inference software running on Nvidia GPUs, with the companies highlighting Blackwell's programmability as part of the performance gains.

Note that the collaboration between the companies is not simply about putting a new AI model on faster chips.

OpenAI inference lead Philippe Tillet said the company has used Nvidia's tooling and documentation to make its models effective at programming Blackwell and Rubin GPUs. OpenAI CTO of compute Uday Ruddarraju added in the press release that the company's internal models were used to optimise inference on Nvidia GPUs.

That creates a hardware-software optimisation loop: OpenAI can use its models to improve inference performance, while Nvidia's programmable platform provides the underlying compute infrastructure.

For Nvidia, the announcement also highlights the importance of inference as AI workloads increasingly move beyond model training. Faster response times are particularly relevant to agentic applications that perform multiple operations before completing a task.

Nvidia's AI Infrastructure Push Takes Shape

The GPT-6 Astra development adds to Nvidia's expanding role in OpenAI's infrastructure strategy.

In September 2025, Nvidia and OpenAI announced plans for OpenAI to deploy at least 10 gigawatts of Nvidia systems, including the company's Vera Rubin platform. Nvidia also said it intended to invest up to $100 billion in OpenAI progressively as the infrastructure is deployed.

The latest announcement is narrower: Nvidia said GPT-6 Astra Ultrafast is currently running on Blackwell GPUs, but it does not establish how many Nvidia GPUs OpenAI is using for the service or how much revenue Nvidia will generate from the deployment.

Nvidia's Blackwell platform is being positioned as infrastructure for both high-performance inference and increasingly complex agentic workloads. The AI leader had said earlier this year that its Blackwell Ultra systems had delivered leading results on AgentPerf, an industry benchmark for agentic AI, with its GB300 NVL72 system running up to 20 times more agents per megawatt than the tested Hopper platform.

Disclaimer: Our digital media content is for informational purposes only and does not constitute investment advice. Please conduct your own analysis or seek professional advice before investing. Remember, investments are subject to market risks, and past performance does not guarantee future returns.