← Accueil actu | Salle du marché
2026-09-10 · Source：HotHardware
Qualcomm's largest cloud data center push to date will pair its power-efficient architecture with custom inference silicon and optical interconnects built alongside AWS.
Qualcomm and Amazon have announced a multi-generational strategic collaboration to co-develop custom AI inference silicon and high-bandwidth optical interconnects for hyperscale data centers. The agreement represents Qualcomm's largest expansion into cloud data center infrastructure to date, embedding the chipmaker's power-efficient processing architecture within Amazon's cloud operations.
The silicon side of the partnership targets AI inference — the phase in which trained models are put to work serving real-world requests. Inference at hyperscale is dominated by energy and throughput economics, which is where Qualcomm's power-efficient processing architecture becomes the centerpiece of the collaboration: the companies are pairing that design philosophy with Amazon's data center footprint to build custom chips tuned for AWS workloads.
The second pillar, high-bandwidth optical interconnects, addresses the movement of data inside hyperscale facilities. As AI clusters scale across thousands of servers, the links between chips, nodes and racks become a defining constraint on performance, and optical technology is the standard answer for pushing bandwidth while containing power draw. Co-developing interconnects alongside the inference silicon signals that Qualcomm and Amazon intend to optimize the full data path rather than the processor alone.
The multi-generational framing is equally significant. Rather than a one-off chip engagement, the agreement is structured to carry across successive hardware cycles, giving both companies a durable framework for refreshing custom silicon and interconnect technology as AI workloads evolve inside Amazon's hyperscale data centers.
For Qualcomm, the deal marks its deepest step yet into cloud data center infrastructure, extending a processing architecture built around power efficiency into the center of the AI buildout. For Amazon, it adds a co-development partner for custom inference hardware and the optical fabric that connects it, spanning both compute and the data path across its hyperscale fleet.
Mots-clés: Qualcomm · Amazon AWS · AI Inference · Optical Interconnects
萬安算交所 AIXX · AIXX · Bourse compute vérifiée · Accéder à la salle