Qualcomm Introduces Its First AI Data Center Processor To Challenge Nvidia

Qualcomm, the maker of the Snapdragon processors found in most Android phones, has entered a new market: data centers. Today they announced the Dragonfly C1000, their first processor targeted at running AI agents in servers. They’ve also secured a contract with Meta to start using it in 2028, causing Qualcomm’s stock to jump 15% after hours.

This is a huge bet for Qualcomm. They’re adjusting their long term revenue expectations by a large margin: they now expect their non-phone business to reach $40 billion in 2029, almost doubling their previous estimate. A big part of that growth will come from data centers, which they expect to generate $15 billion of revenue. Instead of dipping their toes in a new market, they’re doubling down to create a major new line of business.

1The chip is built for how AI actually runs today

The Dragonfly C1000 is tailored for AI agent use cases, which involve running an AI model in realtime to handle ongoing interaction. These have become popular as AI assistants, code generators, and self-driving cars move from prototype to product. There’s a difference between training an AI model (Nvidia’s specialty) and running it efficiently in production, handling millions of requests a day.

In addition to announcing the Dragonfly, Qualcomm is buying Modular for about $4 billion. Modular creates software tools that allow AI programs to run on multiple types of hardware without modification. Together with the Dragonfly, this allows Qualcomm to offer both the hardware and software needed for data centers. That helps them compete with Nvidia, which enjoys strong lock-in thanks to their CUDA software stack.

2Why this matters for the cost of AI

Currently, Nvidia controls the majority of the AI server chip market, giving them plenty of leverage. Almost all the chips powering the AI models used by ChatGPT, Claude, Gemini, and other services are Nvidias. One reason why AI APIs are still so expensive compared to their raw computational cost is because of Nvidia’s monopoly.

With Qualcomm entering the data center space, and a major player like Meta committing to using their product, it’s a clear sign of competition in the market. AMD has been slowly eroding Nvidia’s market share, and now Qualcomm offers another viable option. In the past, more competition in AI hardware has led to cheaper options further down the stack, which trickles down to developers and builders paying by the token or API request. It won’t happen overnight (Meta’s contract doesn’t begin until 2028), but the trend is clear.