Qualcomm, Amazon Partner on AI Data Center Chips
Qualcomm Technologies and Amazon will collaborate on customized silicon and optical networking for large-scale artificial intelligence data centers, a deal that broadens Qualcomm’s push beyond mobile processors and into the fast-growing market for AI infrastructure.
The companies said that they will work across multiple product generations on custom chips for AI inference, the process of using trained models to generate responses, predictions or other outputs. They also will develop optical connectivity solutions capable of supporting speeds of up to 1.6 terabits per second, along with future-generation technologies.
The partnership pairs Amazon Web Services’ AI infrastructure with Qualcomm’s semiconductor design, power-efficient processing and system-integration capabilities. For AWS, customized silicon can help control costs and improve performance for customers running AI applications at scale. For Qualcomm, the collaboration offers a strategic foothold in data centers, where Nvidia has held a dominant position in AI computing hardware.
“As AI demand accelerates, data center infrastructure will require advances in both computing and connectivity to deliver greater performance with more efficiency,” Cristiano Amon, Qualcomm’s president and CEO, said.
Qualcomm and Amazon will also develop high-performance optical interconnects to support the increasing bandwidth requirements of AI clusters. The effort will use Qualcomm’s serializer-deserializer, or SerDes, and optical digital-signal-processing technologies.
Qualcomm’s data-center roadmap, unveiled in June, includes support for 800-gigabit and 1.6-terabit connectivity across optical, active optical cable and active electrical cable applications, including links extending up to 20 kilometers.
As part of the agreement, Qualcomm will expand its use of AWS AI infrastructure, including Amazon Bedrock, for electronic design automation workloads. EDA tools are used to design, simulate and verify complex chips; moving those compute-intensive workloads to cloud infrastructure can shorten design cycles and give chip engineers access to more flexible processing capacity.
“By working together on customized silicon and advanced connectivity, we’re delivering more performant, efficient and cost-effective infrastructure for our customers,” Prasad Kalyanaraman, an AWS vice president, said.