Qualcomm and Amazon have announced a long term partnership to develop custom AI inference chips and high speed optical networking technology for Amazon Web Services.
The agreement marks a major expansion of Qualcomm's data center ambitions and could involve up to $60 billion in purchases of Qualcomm hardware, networking products, and manufacturing related services through 2036.
The two companies plan to work together on custom AI accelerators designed for inference workloads, while also developing 1.6 terabit per second optical interconnect technology for hyperscale data centers.
| Detail | Qualcomm and Amazon agreement |
|---|---|
| Main focus | Custom AI inference silicon |
| Cloud partner | Amazon Web Services |
| Optical interconnect target | 1.6Tbps |
| Potential Amazon purchases | Up to $60 billion |
| Agreement horizon | Through 2036 |
| Amazon warrant | Up to 25 million Qualcomm shares |
| Exercise price | $161.26 per share |
| Initial shares vested | 3.75 million |
| Qualcomm cloud usage | AWS infrastructure and Amazon Bedrock |
Custom AI Chips Will Target Inference Efficiency
The partnership is focused on AI inference rather than only training large models.
Inference refers to running trained AI models to generate results, answer queries, process data, or power applications.
At hyperscale, these workloads require large amounts of computing power, memory bandwidth, and electricity.
Qualcomm and Amazon plan to design custom accelerators specifically for these tasks, with an emphasis on high performance and energy efficiency.
Qualcomm will contribute its experience in low power processor design, custom silicon, and system integration.
AWS will provide the cloud environment and infrastructure requirements that shape the chips.
The companies have not yet announced detailed specifications, production nodes, release dates, or expected performance figures for the new processors.
1.6Tbps Optical Networking Is Also Part of the Deal
The agreement goes beyond compute.
Qualcomm will also contribute digital signal processing and high speed SerDes technology to help create optical connections capable of 1.6Tbps data rates.
These links are intended for networking inside large AWS data center clusters.
As AI systems grow, moving information between servers and accelerators can become as important as the processors themselves.
Faster optical connections can reduce communication bottlenecks between systems running large scale AI workloads.
The companies have not detailed when the 1.6Tbps technology will enter production or how widely AWS plans to deploy it.
Amazon Could Spend Up to $60 Billion
The financial structure shows how large the partnership could become.
Qualcomm has issued Amazon a warrant that allows it to purchase as many as 25 million Qualcomm shares at $161.26 each over the next decade.
However, Amazon does not receive all of those rights immediately.
The warrant is tied to purchasing milestones.
An initial 3.75 million shares vested based on Amazon's starting commitments.
Additional portions are expected to vest as Amazon increases spending on Qualcomm server processors, networking equipment, and related services.
According to the supplied filing details, Amazon would need to spend up to $60 billion through 2036 to unlock the entire 25 million share allocation.
That figure represents a potential maximum rather than guaranteed spending.
Qualcomm Will Expand Its Use of AWS
The partnership works in both directions.
Qualcomm plans to rely more heavily on Amazon Web Services for internal development.
This includes using AWS infrastructure for electronic design automation, which is used to design, simulate, and verify complex chips.
Qualcomm also plans to use services including Amazon Bedrock in its development workflows.
This could help the company speed up future semiconductor design work while keeping more of its internal engineering infrastructure in the cloud.
Qualcomm Is Pushing Beyond Mobile
Qualcomm is best known for processors used in smartphones, but the company has been expanding into PCs, automotive technology, connectivity, and data centers.
This agreement gives it a large hyperscale customer for custom AI infrastructure.
The company has also been strengthening its networking capabilities, including through its acquisition of Alphawave Semi.
That combination gives Qualcomm a broader data center portfolio covering compute, connectivity, and custom silicon.
The Amazon agreement suggests that Qualcomm wants to compete more directly in infrastructure markets where NVIDIA, AMD, Broadcom, and cloud providers' internal silicon teams are already active.
The Deal Could Run for a Decade
The agreement is described as multi generational, suggesting that it is intended to cover several chip generations rather than a single product.
The potential spending window runs through 2036.
That gives both companies time to develop multiple AI accelerators and networking products as AWS infrastructure requirements evolve.
Qualcomm has not disclosed how much revenue it expects to recognize in the early years of the partnership.
The $60 billion figure depends on Amazon reaching the highest purchasing thresholds, so actual spending could be lower.
Still, the structure gives Qualcomm one of its clearest opportunities yet to establish a larger position in hyperscale AI infrastructure and reduce its dependence on the mobile processor market.
Discussion (0)
Be the first to comment.