Qualcomm has announced new artificial intelligence accelerator chips aimed at competing with Nvidia and AMD in the data centre market, causing its shares to jump 11%.
Traditionally focused on mobile and wireless chipsets, Qualcomm is now targeting AI inference workloads with its AI200 chip launching in 2026 and the AI250 in 2027. These chips will be part of fully liquid-cooled server racks designed for high-performance AI tasks.
Qualcomm’s data centre processors are based on the Hexagon neural processing units found in its smartphone chips. Durga Malladi, general manager for data centre and edge, said, “We first wanted to prove ourselves in other domains, and once we built our strength over there, it was pretty easy for us to go up a notch into the data centre level.”
The AI chip market is rapidly growing, with McKinsey projecting $6.7 trillion in global data centre spending by 2030, mostly focused on AI hardware. Nvidia currently dominates over 90% of this sector, powering AI models like OpenAI’s GPT. However, OpenAI and others are diversifying suppliers, including partnerships with AMD, while giants like Google and Microsoft develop in-house AI accelerators.
Qualcomm’s chips are designed primarily for inference rather than training. The company claims its racks consume around 160 kilowatts, comparable to Nvidia’s, but with lower operational costs. Qualcomm will offer chips both as full racks and individual components, allowing clients flexibility.
“What we have tried to do is make sure that our customers are in a position to either take all of it or say, ‘I’m going to mix and match,’” Malladi said.
Qualcomm has also partnered with Saudi Arabia’s Humain to supply AI inference chips to the region’s data centres, with deployments planned up to 200 megawatts of power. The firm highlights advantages in power efficiency, cost, and memory capacity, with AI cards supporting 768 gigabytes of memory—higher than Nvidia and AMD offerings.