Qualcomm is entering the high-performance server market with its first dedicated CPU, the Dragonfly C1000. This move matters to data center operators because the chip targets the severe power and efficiency bottlenecks that currently limit AI infrastructure scaling. By focusing on performance per watt, Qualcomm aims to reduce operational costs for hyperscale cloud providers.
Meta signs multi-generation agreement to use the chip for data centers
The Dragonfly C1000 utilizes Qualcomm's self-developed Oryon architecture in a multi-chiplet design. The processor packs more than 250 cores with clock speeds exceeding 5GHz. Meta has confirmed it will become the first public customer by signing a multi-generation strategic agreement to use Qualcomm as a data center CPU supplier.
Specifications
- Architecture: Oryon
- Core Count: >250
- Clock Speed: >5GHz
- Interface: PCIe Gen 7, CXL
- Memory: High-bandwidth computing memory
Technical specifications include support for PCIe Gen 7 with bandwidth over 2TB/s and the CXL interface. The chip is compatible with high-bandwidth computing memory to address data movement constraints. It also meets the OCP ORv2 standard for open server hardware.
Qualcomm claims the C1000 delivers more than double the performance per watt compared to existing server CPUs. The first production platforms will support both air and liquid cooling solutions. Mass production is scheduled for the second half of 2028.
Two unnamed hyperscale customers have already placed orders for custom chips based on this technology. The product faces competition from AMD, Intel, and Nvidia in the server space. The announcement marks a significant shift in the supply chain for AI data center processors.
Discussion
0 comments
Log in to join the thread with a thoughtful take, question, or correction.