Qualcomm Lands $60 Billion Amazon Deal for Custom AI Chips and Optical Networking

14 Sources

Share

Qualcomm announced a multi-generational partnership with Amazon Web Services to develop custom silicon for AI inference and 1.6 Tbps optical networking hardware. The deal includes $4 billion in warrants tied to up to $60 billion in purchases through 2036, marking Qualcomm's major push into AI data centers alongside its existing Meta partnership.

Qualcomm Secures Major AWS Partnership Worth Up to $60 Billion

Qualcomm announced a sweeping multi-generational partnership with Amazon Web Services (AWS) to develop custom AI chips and optical networking solutions for AI data centers

1

2

. The Qualcomm Amazon partnership includes a complex financial arrangement where Amazon receives warrants to purchase 25 million Qualcomm shares at $161.26 each, representing a $4 billion position

3

5

. These warrants vest in tranches tied to commercial milestones and Amazon purchasing up to $60 billion of Qualcomm server chips and technology through September 2036

1

. Qualcomm's stock surged more than 9% in premarket trading following the announcement, offering relief after the stock slipped about 1% this year due to weaker smartphone demand

2

.

Source: SiliconANGLE

Source: SiliconANGLE

Custom Silicon for AWS Targets AI Inference Market

The collaboration centers on developing custom AI chips optimized specifically for AI inference workloads, a fast-growing market that has become a key battleground among semiconductor firms

2

4

. While Qualcomm offered few concrete details about the custom silicon for AWS, the arrangement could involve fully custom silicon, chiplets, or customized versions of its emerging datacenter accelerator portfolio

1

. Qualcomm's approach includes its rack-scale AI platform that swaps expensive high-bandwidth memory (HBM) for what the company calls high-bandwidth compute, achieving higher effective memory bandwidth using cheaper LPDDR memory by shifting compute to the memory's base die

1

. This technology is expected to launch next year as part of Qualcomm's AI250-series Dragonfly rack systems.

Dragonfly C1000 CPU and Expanding Data Center Ambitions

Qualcomm is simultaneously developing the Dragonfly C1000 CPU, which will scale to more than 250 cores and reach speeds of 5 GHz when it arrives in the second half of 2028

1

. The company unveiled this datacenter processor in June with Meta as a launch customer from 2028, setting an ambitious $15 billion data center revenue target for fiscal 2029

3

4

. The AWS agreement gives Qualcomm a second major hyperscaler partner alongside Meta, both of which commit capital expenditures on AI infrastructure running into hundreds of billions of dollars annually

4

. This represents a significant step for Qualcomm as it works to establish itself as a supplier of AI data centers chips alongside Nvidia, leveraging its reputation built on smartphone and mobile handset processors.

Source: The Register

Source: The Register

High-Speed Optical Connectivity Addresses Infrastructure Bottlenecks

Beyond processors, Qualcomm will supply Amazon with 1.6 Tbps optical networking hardware using its serializer-deserializers (SerDes) and optical digital signal processors (DSPs), likely acquired through its purchase of Alphawave Semi late last year

1

5

. This high-speed optical connectivity component addresses a major supply chain hurdle for scaling AI infrastructure, particularly as compute scales from a rack with a few dozen accelerators to hundreds or thousands spanning multiple racks

1

. AWS vice president Prasad Kalyanaraman stated that "by working together on customized silicon and advanced connectivity, we're delivering more performant, efficient, and cost-effective infrastructure for our customers"

5

.

Amazon's Expanding Custom Silicon Strategy

Amazon has been aggressively expanding its custom silicon ambitions across multiple fronts. CEO Andy Jassy raised the prospect of selling chip racks to outside customers in his April 2026 shareholder letter, noting annualized chip revenue at roughly $20 billion across its Trainium, Graviton, and Nitro product lines and suggesting external sales could push that toward $50 billion annually

4

. Amazon's senior vice president Peter DeSantis confirmed the company is in talks to sell its Trainium AI chips to other companies' data centers, positioning Amazon as a direct peer to Nvidia in chip development

4

. Optical networking remains a major focus, with Amazon agreeing to buy billions of dollars worth of hardware from Corning in June and detailing an internally developed network design dubbed RNG that uses fiber-optic cables in a semi-random configuration

5

.

Source: Benzinga

Source: Benzinga

Reciprocal Benefits Through Amazon Bedrock Integration

The customer-supplier relationship runs both ways in this arrangement. Qualcomm plans to expand its use of Amazon Bedrock, AWS's managed AI service, for electronic design automation workloads

1

5

. Electronic design automation applications help engineers create and test chip blueprints, and Qualcomm expects this initiative to speed up its product development efforts and shorten development cycles

2

5

.

Market Implications and Competitive Landscape

Bank of America expects the server processor market to more than double from $27 billion in 2025 to $60 billion by 2030, with Nvidia, Intel, and AMD all pushing aggressively into the space

3

. Qualcomm's deal with Amazon demonstrates how equity arrangements are becoming ordinary in AI infrastructure deals, with Amazon being paid in warrants to become a customer

3

. This multi-generational partnership spanning through 2036 signals long-term strategic alignment between the companies as AI inference workloads continue growing. Watch for how Qualcomm's customized versions of its AI250-series chips and the Dragonfly C1000 perform against established players, and whether the company can meet its ambitious $15 billion datacenter revenue target by fiscal 2029.

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved