Microsoft (MSFT 0.50%) and Nvidia (NVDA +1.95%) just announced a partnership with foundational large language model (LLM) maker Anthropic. As part of the deal, Anthropic has agreed to buy $30 billion of compute capacity from Microsoft's cloud computing unit, Azure, and to commit to an additional 1 gigawatt of compute capacity. One gigawatt of compute capacity is presently valued with a cost around $50 billion, with $35 billion of that the cost of graphics processing units (GPUs).
In return, Microsoft will invest up to $10 billion in Anthropic, which now has a $350 billion valuation, while Nvidia will invest $5 billion. Anthropic will also work with Nvidia to optimize its artificial intelligence (AI) model designs for use with Nvidia's future GPU architecture. The $350 billion valuation is a big jump from the $183 billion valuation Anthropic received in its prior funding round in September.
Anthropic already had investments from both Amazon (AMZN +2.16%) and Alphabet (GOOGL +0.14%) (GOOG +0.08%). Amazon is the company's primary cloud computing provider and main training partner, and it will remain in these roles after this deal. Amazon recently opened up Project Rainier, which is an $11 billion AI data center exclusively designed to train Anthropic models and run inference for the company. The data center uses Amazon's custom Trainium 2 AI chips.
Last month, Anthropic struck a deal with Alphabet for access to more than 1 gigawatt of computing capacity that will come online next year. As part of the deal, Anthropic will use Alphabet's custom Tensor Processing Units (TPUs) to help train the next generation of its Claude AI models. At the time, Anthropic said it chose to expand its use of Alphabet's TPUs due to their efficiency and "strong price-performance."
Image source: Getty Images.
What the deal means for Microsoft
For Microsoft, the deal with Anthropic continues the company's quest to diversify away from OpenAI. While OpenAI is a great partner and should remain one, the two companies are now set to forge their own paths in AI. As such, striking a deal with another leading AI model company is a smart move.
NASDAQ: MSFT
Key Data Points
Overall, this is a beneficial deal for Microsoft. First, it gets a nice commitment for Azure, which should help drive revenue growth. In addition, it adds another leading AI model to its platform, giving its customers a viable choice outside of OpenAI. That's a win-win.
Whether or not Microsoft's investment in Anthropic will pay off, given the company's valuation, is up in the air, but at the end of the day, it is a small price to pay for the additional business coming its way.
What the deal means for Nvidia
For Nvidia, the deal with Anthropic increases commitments from a customer that has been using a combination of GPUs and ASICs (application-specific integrated circuits) to power its AI workloads. Microsoft primarily uses Nvidia's GPUs to power Azure, so Nvidia will be selling a lot of its chips as part of this partnership. Nvidia working with Anthropic to optimize designs should also help ensure that the LLM maker remains a user of its GPUs well into the future, despite Anthropic taking a multipronged approach with AI chips.
NASDAQ: NVDA
Key Data Points
The deal is also another example of Nvidia being willing to help finance its own customers. This deal is much smaller than its up to $100 billion investment into OpenAI, but the motivation behind it is similar: ensuring its customers have the cash to buy its chips. Meanwhile, Nvidia has a lot of excess cash, so investing in other parts of the AI value chain isn't a bad idea.
Nvidia and others also don't want to see OpenAI become too powerful, and this investment helps one of its chief rivals. In fact, the more competition in the AI model space, the better it is for Nvidia, which is the main provider with its GPUs.
Overall, the deal is good for both Microsoft and Nvidia investors, as it will add some incremental revenue growth in the coming years. However, it's far from a game changer for either, as OpenAI and Anthropic are both clearly also trying to spread their bets out. That said, as long as there is an AI race, it's good news, especially for Nvidia.
