Centamil Introduces 5,000W Hayagreeva HEx Cold Plate for AI Data Centers

May 28, 2026
Centamil has launched the Hayagreeva HEx, a copper-matrix cold plate designed for AI data center infrastructure, capable of dissipating over 5,000 watts of heat with a thermal resistance of 0.005°C/W.

Centamil has launched the Hayagreeva HEx, its first product in AI thermal management, announced in a press release. The copper-matrix cold plate is designed for AI data center infrastructure and has been validated to remove more than 5,000 watts of continuous heat at a rated thermal resistance of 0.005°C/W.

The product targets upcoming AI server silicon with power dissipation levels exceeding current cooling capabilities. According to Centamil, AI accelerator power has grown from 300 watts per chip three years ago to 1,200 watts today, with projections reaching over 4,400 watts within two years.

Hayagreeva HEx uses a copper-matrix wetted-surface architecture providing over 4,000 square meters per cubic meter of surface area for heat exchange. The design is protected by 14 patent filings covering its thermal structure, manufacturing process, and integration system. Production samples are expected to ship within four to six weeks following a letter of intent, and a pilot evaluation program is open to qualified data center operators, server manufacturers, and integration partners.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like Silicon Brief or Daily AI Brief.

Also, consider following us on social media:

AI Chips & Datacenters AI Brief AI Brief (X)
Free newsletter

Silicon Brief

Weekly coverage of AI hardware developments including chips, GPUs, cloud platforms, and data center technology.

AI Token Webinar

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more

You may also like

Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.