Google LLC is investing more than $1 billion in Anthropic PBC, the Financial Times reported today.

The cash infusion comes on top of the $2 billion that Google has already provided to the artificial intelligence developer. Separately, Anthropic is said to be raising $2 billion from a group of institutional investors led by Lightspeed Venture Partners. The latter deal is expected to value the company at $60 billion.

Anthropic develops a series of enterprise-focused large language models called Claude. The models can generate text, craft software code and perform a range of other tasks. Anthropic provides its LLMs via an application programming interface that makes it possible to integrate them into third-party software. 

The company debuted its most capable mode, Claude 3.5 Sonnet, last June. Anthropic claims that the algorithm outperforms OpenAI’s general-purpose GPT-4o model across a range of reasoning tasks. Claude 3.5 Sonnet can also perform actions in business applications on users’ behalf.

At a Wall Street Journal event this week, Anthropic Chief Executive Officer Dario Amodei shared new details about the company’s AI roadmap. He detailed that Anthropic plans to equip its LLMs with the ability to browse the web and remember user preferences across projects. Additionally, Anthropic will raise its API’s rate limits to let developers send more prompts to its LLMs.

According to Amodei, Anthropic limits API usage partly because its backend infrastructure is struggling to keep up with demand. The company is currently working to upgrade that infrastructure. The reported $1 billion investment from Google could make it easier for Anthropic to buy more AI hardware. 

This past November, the LLM developer raised a $4 billion funding round from Amazon.com Inc. As part of the partnership, Anthropic will use Amazon Web Services Inc.’s custom AWS Trainium chips to train future foundation models. It’s unclear if the investment from Google might see Anthropic increase its use of the search giant’s competing public cloud to support LLM projects. 

Last month, AWS detailed that it’s collaborating with Anthropic to build an AI supercomputer called Project Rainier. The system is set to include hundreds of thousands of Trainium chips. Once online, it’s expected to provide more than five times the compute capacity of the cluster that Anthropic used to train its current flagship LLMs.

In addition to upgrading its infrastructure, Anthropic plans to more directly compete with OpenAI’s growing lineup of reasoning-optimized LLMs. During this week’s WSJ event, Amodei said the company is working on new models that can take on more complex tasks. He hinted that Anthropic is not building LLMs focused solely on reasoning, but rather plans to equip general-purpose models with better problem-solving skills.

Image: Anthropic