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

In a move that signals a deeper convergence between silicon design and cloud infrastructure, Qualcomm and Amazon have struck a partnership to craft bespoke AI inference chips for Amazon Web Services’ (AWS) data‑center fleet. The two tech giants said they will not only co‑develop several generations of custom silicon, but also work on high‑speed optical connectivity that can keep up with the surging data‑traffic of modern AI workloads.

“AI demand is accelerating faster than anyone expected,” said Qualcomm President and CEO Cristiano Amon. “That means our data‑center hardware has to evolve – both in raw compute power and in the way that data moves around.” He added that Qualcomm will bring “decades of leadership in power‑efficient processing” to the table, hoping to unlock “breakthrough performance” for the next wave of AI services.

From the Amazon side, AWS Vice‑President Prasad Kalyanaraman emphasized the practical upside: “When we talk about custom silicon and advanced connectivity, we’re really talking about delivering more performance at lower cost for our customers.” In other words, the collaboration is meant to make the cloud cheaper to run while still delivering the horsepower that large‑scale generative‑AI models crave.

The technical details are still under wraps, but Qualcomm’s press release hinted at optical links that can reach up to 1.6 terabits per second, leveraging the company’s SerDes and digital signal‑processing know‑how. Those connections are expected to plug directly into AWS’s internal networking fabric, shaving latency and easing bandwidth bottlenecks that often plague AI inference at scale.

Beyond the chips themselves, Qualcomm disclosed plans to lean more heavily on AWS’s own AI services, such as Amazon Bedrock, for its electronic‑design‑automation (EDA) workflows. By doing so, the chipmaker hopes to trim the time it takes to design and verify new silicon – a win‑win that could accelerate future iterations of the custom processors.

While no financial terms were disclosed and a rollout timetable remains vague, the partnership clearly marks Qualcomm’s push beyond its traditional mobile‑chip dominance into the realm of massive‑scale, energy‑efficient AI compute. If the collaboration lives up to its promise, the cloud could soon see a new class of AI‑specific hardware that is both faster and greener.