Qualcomm and Amazon Join Forces to Build Custom AI Chips for AWS Data Centers

In a move that signals the growing importance of specialised hardware for artificial‑intelligence workloads, Qualcomm has inked a partnership with Amazon Web Services (AWS) to co‑develop custom AI inference chips. The two giants said they’ll work together over a series of generations, tailoring silicon to the unique demands of AWS’s expanding AI data‑center fleet.

“As AI demand accelerates, data‑center infrastructure will need advances in both computing and connectivity to deliver greater performance with more efficiency,” said Qualcomm President and CEO Cristiano Amon. He went on to note that Qualcomm’s decades‑long expertise in power‑efficient processing will help bring breakthrough performance to the next generation of AI infrastructure.

The collaboration isn’t limited to chips alone. Qualcomm and AWS will also tackle high‑speed optical connectivity, a critical piece of the puzzle when you’re moving petabytes of data across racks in a split second. According to the press release, the teams are targeting optical solutions that can push up to 1.6 Tbps today, with an eye on even faster, future‑generation technologies.

AWS Vice President Prasad Kalyanaraman echoed the sentiment, saying the joint effort will focus on “improving the performance and cost of AI infrastructure.” He added that by marrying custom silicon with advanced optical links, the partnership aims to give customers a more performant, efficient and, crucially, cost‑effective platform for running AI models at scale.

Beyond building hardware, Qualcomm plans to lean more heavily on Amazon’s AI services, such as the Bedrock foundation model platform, for its own electronic‑design‑automation workloads. The goal? To shave time off chip‑design cycles and accelerate innovation across the board.

While the announcement marks a clear step for Qualcomm to diversify beyond its traditional mobile‑chip business, the companies have kept the financial details and exact rollout timeline under wraps. What’s clear, however, is that the partnership could reshape how cloud providers power AI workloads, delivering a blend of custom silicon and ultra‑fast interconnects that rivals the performance of today’s most advanced supercomputers.

For anyone watching the AI hardware race, this joint effort is a reminder that the future of cloud AI isn’t just about more GPUs—it’s about purpose‑built chips and the networks that connect them.