Qualcomm announced a multi-generation collaboration with AWS covering customized silicon for AI inference and optical connectivity solutions extending up to 1.6T. In an accompanying 8-K filing, Qualcomm disclosed that it issued warrants to Amazon to acquire 25 million shares at $161.26 apiece—roughly $4 billion—vesting in tranches tied to commercial milestones and up to $60 billion in server chip and technology purchases. Futurum examines how the agreement maps onto the Dragonfly roadmap disclosed at Investor Day 2026, where Qualcomm silicon fits inside AWS’s CPU-heavy agentic infrastructure, and what the warrant structure signals about the depth of the commitment.
What Is Covered in This Article:
- A multi-generation Qualcomm-AWS collaboration on customized silicon at scale for AI inference
- Optical connectivity work extending up to 1.6T built on Qualcomm SerDes and optical DSP technology
- A $4 billion warrant for 25 million Qualcomm shares vesting against up to $60 billion in server chip purchases
- The fit with Qualcomm’s $15 billion fiscal 2029 data center target and its two unnamed hyperscaler custom silicon wins
- AWS’s agentic infrastructure thesis and the competitive field in custom silicon and optics
The News: Qualcomm Technologies, Inc. (NASDAQ: QCOM) announced on September 8 a multi-generation collaboration with AWS to enable customized silicon at scale for large-scale AI data centers, with the two companies working together on AI inference. The agreement extends to high-performance optical connectivity solutions reaching up to 1.6T and future generations, drawing on Qualcomm’s advanced SerDes and optical DSP technologies. As part of the expanded relationship, Qualcomm plans to deepen its use of AWS AI infrastructure, including Amazon Bedrock, for electronic design automation workloads, targeting a reduction in chip design cycles.
In an accompanying 8-K filing with the SEC, Qualcomm disclosed that it issued warrants to Amazon to acquire 25 million shares at an exercise price of $161.26 per share, representing a total investment of approximately $4 billion. The warrant expires on September 3, 2036, with shares vesting in tranches tied to the “execution of certain commercial arrangements” and the purchase of up to $60 billion worth of Qualcomm’s server chips and other technology.
“Qualcomm is pleased to work with AWS on customized silicon and connectivity solutions, bringing decades of leadership in advanced processing and power-efficient compute, to deliver breakthrough performance and enable the next generation of AI infrastructure,” said Cristiano Amon, President and CEO, Qualcomm Incorporated.
“By working together on customized silicon and advanced connectivity, we’re delivering more performant, efficient, and cost-effective infrastructure for our customers,” said Prasad Kalyanaraman, Vice President, AWS.
The press release discloses no product names, no volume commitments, and no deployment timeline beyond the multi-generation framing. The financial architecture of the relationship is disclosed separately in the 8-K.
Qualcomm Lands $60 Billion Custom Silicon and 1.6T Optics Deal with AWS, Backed by a $4 Billion Warrant
Analyst Take: AWS is now the third top-tier hyperscaler publicly attached to Qualcomm’s data center reentry, following Microsoft CEO Satya Nadella’s Investor Day confirmation that Azure will deploy the HBC accelerator and Mark Zuckerberg’s disclosure of a multi-generational C1000 CPU collaboration at Meta. But this one came with a checkbook. The $4 billion warrant vesting against up to $60 billion in server chip purchases puts this agreement in the allegiance tier alongside Marvell-Google and AMD-OpenAI. The announcement centers on the two Dragonfly product lines closest to revenue—custom silicon and connectivity—and it arrives weeks before the first quarter of fiscal 2027, when Qualcomm guided custom silicon revenue to begin. Futurum reads the agreement as the strongest directional validation yet of the Investor Day thesis, with near-memory silicon for decode-heavy inference and Alphawave connectivity earning the right to win with customers that concentrate AI spending. The $60 billion framework ceiling converts a design relationship into a procurement structure with measurable milestones against the $15 billion fiscal 2029 target.
The Agreement Exercises the Two Dragonfly Lines Closest to Revenue
Qualcomm’s June 24 Investor Day laid out four data center product lines that ramp in sequence, from connectivity in fiscal 2026 to custom silicon in the first quarter of fiscal 2027, the AI accelerator in the second half of fiscal 2027, and the CPU in the second half of fiscal 2028. This agreement lands squarely on the first two. The custom silicon leg addresses AI inference, the workload Futurum identified at Investor Day as the cleanest fit for Qualcomm’s near-memory High Bandwidth Compute architecture, and the optical leg puts Alphawave’s SerDes and DSP assets to work inside a network AWS is rebuilding at scale.
Timing sharpens the significance. Qualcomm set revenue targets of $5 billion in fiscal 2027 and $15 billion in fiscal 2029, with at least $1 billion of the fiscal 2027 figure coming from two global-scale hyperscaler custom silicon customers whose identities were withheld. An AWS collaboration announced days before that fiscal year opens—now backed by a warrant structure vesting against up to $60 billion in purchases—invites the inference that AWS is one of the two. What the announcement settles is alignment and scale. Three of the four largest US hyperscalers now have a public, named Qualcomm data center relationship, which was implausible eighteen months ago when the company had neither Alphawave nor a disclosed accelerator roadmap.
AWS’s Disaggregated Inference Posture Makes Room for Near-Memory Decode Silicon
The workload named in the release, AI inference, is the one AWS has been restructuring all year. Futurum’s coverage of AWS Summit New York 2026 documented the posture that 80% of agentic compute lands on the CPU across the orchestration, tool-use, and guardrail steps. AWS has also proven willing to disaggregate the LLM calls themselves, pairing Trainium prefill with Cerebras decode inside Amazon Bedrock. That architecture creates the socket Qualcomm designed for. The HBC roadmap integrates compute with LPDDR memory stacks to attack decode’s memory bandwidth bottleneck, running from the AI200 sampling in fiscal 2026 to the AI250 with HBC Gen 1 in fiscal 2027 and the AI300 in fiscal 2028, with Qualcomm claiming 4 to 8x better decode performance per watt against HBM-based designs.
The release leaves the commercial form open. Customized silicon at an Annapurna-equipped hyperscaler can mean design services, chiplets, and IP where AWS keeps the architecture, or it can mean HBC-derived parts deployed under AWS specifications. Those two outcomes price very differently for Qualcomm, and the announcement supports either reading. The CPU is the line this deal conspicuously omits. AWS built Graviton 5 precisely for the CPU-heavy agentic loop at roughly 40% better price-performance than x86, so the C1000’s path into AWS is narrow, and Meta remains the only disclosed C1000 customer.
1.6T Optics Sends Alphawave Into Marvell and Broadcom Territory
The optical leg is the nearest-term revenue and the hardest competitive ground. Qualcomm closed the $2.4 billion Alphawave acquisition in December, put founder Tony Pialis in charge of the data center business, and told investors that legacy Alphawave hyperscaler wins were already scaling into production. This agreement extends that base at AWS into module-class optical connectivity at 1.6T and future generations, the leading edge of the transition AWS needs as it wires a 20-million-kilometer backbone and rolls out its RGN randomized graph fabric, which AWS claims cuts in-building network power by roughly 40%.
The incumbents are formidable. Marvell built the optical DSP franchise that dominates hyperscale modules and signed a five-year custom silicon agreement covering Trainium-era programs with AWS in December 2024. Broadcom supplies the switching silicon and DSPs threaded through most AI fabrics. Qualcomm’s counterposition bundles SerDes, optical DSP, and custom compute silicon in one engineering organization, sold to a buyer that has shown it prefers multiple qualified sources at every layer. Whether AWS qualifies Qualcomm optics as a second source or a primary socket will show up in module supply chains before it shows up in Qualcomm’s disclosures.
The $4 Billion Warrant Puts This Deal in the Allegiance Tier
The warrant matches the template now standard across hyperscaler custom silicon commitments. Marvell’s Google agreement came with a warrant vesting toward $120 billion in purchases, and AMD’s OpenAI partnership set the precedent with warrants against 6 GW of deployments. Qualcomm’s $60 billion ceiling is smaller than Marvell-Google but larger than any prior Qualcomm data center commitment by an order of magnitude, and it spans the full ten-year window through 2036. The vesting tranches tied to “execution of certain commercial arrangements” and purchase thresholds mean Amazon’s equity upside is directly linked to procurement volume—a structure that aligns incentives and makes multi-sourcing away from Qualcomm silicon more expensive for AWS at the margin.
As part of the expanded relationship, Qualcomm also plans to deepen its use of AWS AI infrastructure, including Amazon Bedrock, for EDA workloads with a stated target of shorter chip design cycles. Futurum’s coverage of agentic EDA at DAC and CadenceLIVE found AI-driven flows compressing design schedules across Synopsys, Cadence, and Siemens tooling, and hyperscaler-hosted compute is where those flows increasingly run. The commitment completes a circular commerce structure that has become the standard architecture of hyperscaler alignment.
Qualcomm’s Alignment with the AWS Roadmap
The $60 billion figure is a ceiling for this deal, not a floor. Vesting in tranches means Amazon earns its equity incrementally as purchase milestones are met, and multi-generation intent at AWS still coexists with relentless multi-sourcing across Annapurna’s internal designs, the Marvell agreement, Broadcom silicon, NVIDIA fleets, and Cerebras capacity on Bedrock. Qualcomm’s own performance metrics remain vendor-supplied. The 4 to 8x decode performance-per-watt claim, the 5 to 7x bandwidth-per-watt claim, and the 1.6T optical capability all await independent validation. The neutrality tension flagged at Investor Day sharpens with each hyperscaler engagement, since a credible custom silicon partner competing with the same buyers in merchant accelerators is running a strategy Broadcom and Marvell deliberately avoid. Fiscal 2027 guidance now functions as the test. At least $1 billion of custom silicon revenue from two hyperscalers is due to start in the quarter that opens this month, and each earnings report will show whether the warrant structure converts into recognized revenue at the pace the $15 billion target requires.
What to Watch:
- Whether Qualcomm names AWS as one of its two hyperscaler custom silicon customers at the Q4 fiscal 2026 earnings call in November
- Whether custom silicon revenue appears in fiscal Q1 2027 results as guided at Investor Day
- The pace at which Amazon’s warrant tranches vest, signaling procurement volume against the $60 billion ceiling
- Whether AWS deploys HBC-derived accelerators in its fleet or offers Qualcomm-powered instances by name
- Whether Qualcomm optics wins 1.6T sockets against Marvell and Broadcom in AWS’s module supply chain
Read the full announcement about the Qualcomm-Amazon collaboration on Qualcomm’s website.
Declaration of generative AI and AI-assisted technologies in the writing process: This content has been generated with the support of artificial intelligence technologies. Due to the fast pace of content creation and the continuous evolution of data and information, The Futurum Group and its analysts strive to ensure the accuracy and factual integrity of the information presented. However, the opinions and interpretations expressed in this content reflect those of the individual author/analyst. The Futurum Group makes no guarantees regarding the completeness, accuracy, or reliability of any information contained herein. Readers are encouraged to verify facts independently and consult relevant sources for further clarification.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Read the full Futurum Group Disclosure.
Other Insights From Futurum:
Qualcomm Q3 FY 2026: Automotive Growth Offsets Handset Weakness
The Software-Defined Vehicle is Winning the Compute War and Losing the Owner
Qualcomm’s Snapdragon Reality Elite Ups the Stakes for Spatial AI
Author Information
Brendan is Research Director, Semiconductors, Supply Chain, and Emerging Tech. He advises clients on strategic initiatives and leads the Futurum Semiconductors Practice. He is an experienced tech industry analyst who has guided tech leaders in identifying market opportunities spanning edge processors, generative AI applications, and hyperscale data centers.
Before joining Futurum, Brendan consulted with global AI leaders and served as a Senior Analyst in Emerging Technology Research at PitchBook. At PitchBook, he developed market intelligence tools for AI, highlighted by one of the industry’s most comprehensive AI semiconductor market landscapes encompassing both public and private companies. He has advised Fortune 100 tech giants, growth-stage innovators, global investors, and leading market research firms. Before PitchBook, he led research teams in tech investment banking and market research.
Brendan is based in Seattle, Washington. He has a Bachelor of Arts Degree from Amherst College.
