Superpower Daily

Superpower Daily

Your reading space

Sign in
Google Moves Chrome to Two-Week Releases to Speed Security FixesOpenAI Lets ChatGPT Work Build Writing Profiles From Connected...Qualcomm and Amazon Partner on Custom Silicon for AWS AI Infere...Arm Introduces Mali G2-Ultra NX, Its First AI-Native Mobile GPUMistral Raises €3 Billion as Samsung Funds Its Push for More AI...Grove City Holds 310-Acre Data Center Plan, Exposing Limits of...Google Moves Chrome to Two-Week Releases to Speed Security FixesOpenAI Lets ChatGPT Work Build Writing Profiles From Connected...Qualcomm and Amazon Partner on Custom Silicon for AWS AI Infere...Arm Introduces Mali G2-Ultra NX, Its First AI-Native Mobile GPUMistral Raises €3 Billion as Samsung Funds Its Push for More AI...Grove City Holds 310-Acre Data Center Plan, Exposing Limits of...
Swipe

Qualcomm and Amazon Partner on Custom Silicon for AWS AI Inference

The collaboration covers customized compute and high-speed optical links for AWS data centers, giving Qualcomm another opening in AI infrastructure.

By 2 min read
Qualcomm and Amazon Partner on Custom Silicon for AWS AI Inference
Qualcomm and Amazon Partner on Custom Silicon for AWS AI Inference

Listen to this story

The audio brief

About 1:36
0:001:36
Read transcript
Qualcomm and Amazon are launching a multigenerational effort to build custom AI inference infrastructure for AWS data centers. The work combines specialized chips for running AI models with optical networking designed to move data quickly between systems. The chip program is aimed at large-scale inference, the phase where trained models generate answers and predictions. But the announcement leaves the first test largely undefined: there is no named processor, deployment date, purchase volume, commercial terms, or deal value. The networking side is more specific. Qualcomm says its SerDes and optical digital-signal-processing technologies will support connections across AWS data centers, with bandwidth scaling as high as one point six terabits. That puts the partnership beyond a processor deal and into the links that connect an AI system together. There is also a two-way technology exchange. Qualcomm plans to use AWS AI infrastructure, including Amazon Bedrock, for electronic design automation during chip development. Shorter design cycles are the goal, not a reported result. For Qualcomm, this creates another route into cloud infrastructure, where Nvidia has been dominant. The company is targeting fifteen billion dollars in data-center sales for fiscal twenty twenty-nine, and Meta has said it plans to use Qualcomm’s Dragonfly C1000 data-center CPU when production begins in twenty twenty-eight. CNBC reported Qualcomm shares rose six percent after the announcement, signaling investor interest rather than booked revenue. The key constraint is still basic: the partnership has no disclosed first product or launch date, so its commercial impact remains unproven.

Story brief

3 key points

Qualcomm and Amazon are starting a multigenerational program for AWS data-center infrastructure, combining custom inference silicon with optical networking that can scale to 1.6T. Qualcomm also plans to use AWS services, including Bedrock, for chip-design automation. The partnership could help Qualcomm build a larger cloud business alongside its announced $15 billion fiscal 2029 data-center sales target, but its...

  1. 01

    The announcement names no processor, deployment date, purchase volume, commercial terms, or deal value.

  2. 02

    The work covers AWS inference silicon plus SerDes and optical DSP connectivity for data-center networks up to 1.6T.

  3. 03

    Qualcomm will use AWS AI infrastructure, including Bedrock, for electronic design automation; shorter design cycles remain a target.

Qualcomm and Amazon have launched a multigenerational effort around custom AI infrastructure for AWS, pairing silicon for inference with optical connections designed for data-center networks. The agreement gives Qualcomm a route deeper into cloud infrastructure, although its announcement does not identify the first product or deployment date.

Customized compute for AWS inference

The central commitment is customized silicon for large-scale AWS data centers focused on AI inference. Qualcomm said the companies will work across multiple generations of hardware rather than around one named processor. The scope places Qualcomm alongside a major cloud provider as it tries to expand its data-center business, where Nvidia has been dominant in AI computing.

The network is part of the deal

The collaboration also reaches beyond chips. Qualcomm said its SerDes and optical digital-signal-processing technologies will support high-bandwidth interconnects within Amazon data-center networks. The companies are working on optical connectivity solutions extending up to 1.6T, along with future-generation connectivity products.

AWS tools enter Qualcomm’s design process

The relationship also runs in the other direction. Qualcomm plans to deepen its use of AWS AI infrastructure, including Amazon Bedrock, for electronic design automation workloads used in chip development. Qualcomm is targeting shorter chip-design cycles, a stated goal rather than a reported outcome.

A larger data-center push

Qualcomm has set a target of $15 billion in data-center sales for fiscal 2029, which is a company target rather than booked revenue. The AWS collaboration follows Qualcomm’s earlier statement that Meta would use its Dragonfly C1000 data-center CPU when production begins in 2028. CNBC reported Qualcomm shares rose 6% after the Amazon announcement, reflecting investor interest in the company’s data-center ambitions.

Editorial analysis

Our Read

The strategic signal is Qualcomm’s attempt to sell into more than one layer of an AI data center. Its collaboration with AWS combines customized silicon with the links that move data around the facility, while Qualcomm also plans to use AWS tools in its own chip-design workflow. That does not establish delivered AWS capacity or commercial scale. It does show that the company’s data-center effort is being framed as a systems business, not solely a CPU business. The next meaningful evidence would be a named component or an indication that Qualcomm’s planned AWS-based design work is reducing chip-design cycles.

Related coverageAWS Plans 2 Million More Nvidia GPUs for 2027–28 Despite Its Own Chip EffortNvidia Says Vera Rubin Represents $40B in Revenue Opportunity per Gigawatt

Sources

  1. cnbc.comQualcomm stock pops 6% on data center infrastructure deal with Amazon
  2. prnewswire.comQualcomm Announces Multi-Generational Product Collaboration with Amazon to Build Next-Generation AI Data Center Infrastructure

Loading discussion...

Related coverage

Continue reading