OpenAI raises $110B at $730B valuation, expands AWS and Nvidia partnerships – News

OpenAI Group PBC today announced that it has raised $110 billion in funding at a pre-money valuation of $730 billion.

Nvidia Corp. and SoftBank Group PBC provided $30 billion each. The remaining $50 billion came from Amazon Web Services Inc., which has also expanded its technical partnership with OpenAI.

In November, the ChatGPT developer committed to using $38 billion worth of AWS services. The deal announced today will add an additional $100 billion worth of cloud consumption over eight years. In particular, the Amazon.com Inc. unit will use its AWS Trainium artificial intelligence chips to provide OpenAI with 2 gigawatts of computing capacity.

OpenAI is adopting not only the newest chip in the product series, Trainium3, but also a planned successor called Trainium4. The companies disclosed today that Trainium4 will provide a “major performance gain,” particularly when running FP4 workflows. FP4 is a data format commonly used by AI models. OpenAI expects its Trainium4 clusters to start coming online in 2027.

The company will use some of the chips’ computing capacity to power a new offering called the Stateful Runtime Environment. It will run on AWS’ Amazon Bedrock managed AI service. The offering is intended to power AI agents that can perform multistep tasks. According to OpenAI, the Stateful Runtime Environment will enable agents to carry over data from one step to another.

The company will also use its Trainium computing capacity to power Frontier, a software platform it debuted this month. The platform makes it easier to build and maintain AI agents. AWS will become the “exclusive third-party cloud distribution provider” of Frontier under the partnership.

The Amazon unit’s largest competitor in the cloud market, Microsoft Corp., is also a major OpenAI backer. It has provided the ChatGPT developer with $13 billion worth of funding and infrastructure since 2019. The companies said in a joint statement today that their “partnership remains strong and central.” Their revenue sharing agreement will remain in place following the funding round, as will a contract that gives Microsoft access to some of OpenAI’s intellectual property.

The tech giant will also continue providing infrastructure to OpenAI. The ChatGPT developer uses Nvidia Corp. chips installed in Microsoft, Oracle Corp. and CoreWeave Inc. data centers to power many of its workloads.

Nvidia will broaden its chip partnership with OpenAI following the funding round. Currently, the LLM developer’s graphics card clusters mainly use Hopper and Blackwell chips. The expanded partnership will see OpenAI add 5 gigawatts of computing capacity powered by the Vera Rubin architecture. The architecture comprises the Rubin graphics card, which Nvidia debuted last month, and a central processing unit called Vera.

OpenAI says that it expects several more backers to join the funding round. CNBC cited a source as saying that Microsoft has the option to participate. The raise could be OpenAI’s last before its initial public offering, which is expected to take place as soon as this year. 

“Through this additional investment, we will accelerate OpenAI’s research and ecosystem expansion, while advancing our own ASI strategy,” said SoftBank Chief Executive Officer Masayoshi Son.

OpenAI disclosed that more than 50 million consumers and 9 million business users are now paying for ChatGPT. It added that January and February are “on track to be the largest months for new subscribers in our history.” Some of that growth is driven by OpenAI’s Codex programming tool, which has seen its weekly user count more than triple since the start of the year.

Photo: Focal Foto/Flickr