Zyphra, an American artificial intelligence model developer, is raising $500 million in a new funding round that includes commitment from chipmaker AMD, sources told Forbes. The valuation will be at least $5 billion, some of them said.

AMD did not respond to requests for comment. Zyphra declined to comment.

San Francisco-based Zyphra develops advanced, open-weight artificial intelligence models and offers cloud infrastructure services. While most AI labs still default to Nvidia chips, Zyphra says it trains and runs inference for its models entirely on AMD hardware—pitching that as both a cost and supply-chain advantage.

In recent months, investors have been increasingly interested in funding U.S.-based alternatives to DeepSeek. This includes Zyphra rival Reflection AI, which is reportedly in talks for a $25 billion valuation.

Zyphra has two arms: a research lab has released about a dozen open-weight models, including a model designed to decode brainwave activity and a series of advanced reasoning models named Zaya. Its second business, Zyphra Cloud, is a ‘neocloud’ that specializes in running and hosting advanced open-weights AI models, all on AMD hardware. The startup has told investors Zyphra Cloud will provide computing capacity to major, compute-constrained frontier laboratories like Anthropic, Meta, and OpenAI.

According to Pitchbook, the startup last raised $110 million at a valuation of $1 billion last year, with investors including Steve Jurvetson’s Future Ventures, Skype founder Jaan Tallinn and Bison Ventures.

The boom in generative AI has turned computing power—“compute”—into the limiting ingredient. The bottlenecks aren’t just chips. They’re also power, data center space, and the slow grind of supply chains. Long manufacturing lead times for specialized AI chips have left AI model makers facing extensive queues for critical processors. In the last few years, AI labs have tried to diversify their infrastructure. OpenAI is developing its own custom silicon, in addition to using Nvidia alternatives like AMD. Anthropic has also deployed a diversified strategy, using Amazon’s Trainium chips, Google TPUs and it is also reportedly planning to work with AMD’s chips.