Qualcomm and Amazon Web Services have outlined a multi-generation collaboration to develop custom AI chips for data centers, with inference as the central focus. The agreement could give AWS hardware tailored to its own infrastructure while providing Qualcomm with a major cloud customer as it expands beyond its smartphone-centric Snapdragon business.
The work will also cover high-speed optical interconnects for moving data between infrastructure components. Qualcomm has cited links of up to 1.6 Tbps, drawing in part on technologies obtained through its AlphaWave acquisition. Such connectivity is increasingly important as AI systems require large volumes of data to move across servers and accelerators.
AWS is also expected to run more of Qualcomm's chip-design workloads on its cloud infrastructure, a move Qualcomm says is intended to shorten development cycles. On the commercial side, Amazon could purchase up to $60 billion in Qualcomm hardware over time, although that figure is a potential ceiling rather than a guaranteed purchase commitment.
The arrangement further includes rights that could allow Amazon to acquire up to $4 billion in Qualcomm stock. Those provisions are long-term contractual mechanisms, not an immediate $4 billion equity purchase.
Qualcomm has been building its data-center AI strategy since entering the market in 2025, and it expanded that push in 2026 with its Dragonfly family for AI infrastructure. The company has set a target of $15 billion in data-center revenue by fiscal 2029.
AWS is not replacing its in-house silicon efforts with Qualcomm hardware: Trainium, Graviton and Nitro remain part of its portfolio. Instead, the deal adds another custom-hardware option as cloud providers secure AI infrastructure years in advance. The key remaining details are which products will reach deployment first and how much of the potential hardware commitment ultimately materializes.
