Dev.to AI ๐Ÿค– Ai ๐Ÿ‘ 0 ๐Ÿ“– 2 min read

Nexalus and Renda launch Nova8, a liquid-cooled AI server that drops into air-cooled data centers

๐Ÿ“… November 17, 2026 โ€ข Business Wire Nexalus, a liquid cooling and thermal energy solutions company from Cork, Ireland, and RENDA, a UK maker of bespoke high-performance servers and workstations, announced a partnership a

Nexalus and Renda launch Nova8, a liquid-cooled AI server that drops into air-cooled data centers

Infographic cover

๐Ÿ“… November 17, 2026 โ€ข Business Wire
Nexalus, a liquid cooling and thermal energy solutions company from Cork, Ireland, and RENDA, a UK maker of bespoke high-performance servers and workstations, announced a partnership and launched the RENDA Nova8, a drop-in ready liquid-cooled AI server. Per the joint press release, the system is built around eight Nvidia RTX PRO 6000 Blackwell GPUs with 96 GB of memory each, packaged as a single 4U server where GPUs, CPU, memory, storage, power and cooling are specified, integrated and tuned together. The design is co-developed and jointly patented.

Cooling is the heart of the design. Per the press release, Nexalus applies its patented jet-impingement method on every GPU and the CPU, driving jets of coolant onto the plate directly above the hottest areas of each chip instead of along the channels of a conventional cold plate, so heat is removed where it is generated. Each GPU runs at its full 600 W, and the server runs all eight at full power. Per AEC Magazine, the Nvidia RTX PRO 6000 Blackwell card is normally dual-slot, and its bespoke liquid-cooling design lets each GPU occupy a single slot, so eight cards fit in a 4U chassis.

The drop-in claim speaks directly to AI capacity planning. Per the joint statement, also cited by DCD, Nova8 fits a standard 19-inch rack in an existing air-cooled data center, so operators can add GPU capacity inside current racks and power budgets without a facility rebuild or new piping. Per the press release, GPUs run more than 20 degrees below their maximum operating temperature, converting thermal headroom into sustained performance, while 10 pumps and 16 fans are designed to maintain reliability. The companies already cool GPU servers for 8 of the 11 constructors in Formula One racing.

In internal testing reported by the companies, Nova8 delivered 55,000 tokens per second on a Llama 3.3 benchmark and up to 20% more throughput than comparable air-cooled systems, plus 8.6 tokens per watt and up to 22% better energy efficiency. Per Unite.AI, the figures are company claims, not independently verified, and a rack fit does not remove the need for a site review of electrical, thermal, networking and service requirements. Buyers can validate on their own models via a remote Test Drive, with no pricing or shipping volumes disclosed ahead of the Data Centre World Paris showcase.

Sources:
businesswire.com/news/home/20261117063662/en/
datacenterdynamics.com/en/news/nexalus-and-renda-launch-liquid-cooled-ai-server-for-data-center-retrofit-market/
unite.ai/nexalus-renda-nova8-liquid-cooled-ai-server
aecmag.com/workstations/renda-and-nexalus-launch-liquid-cooled-ai-server
renda.ai

Key figures

IMPACT checklist

Continue the conversation: https://t.me/AIAutomationArchitecture ยท Bluesky

AI Automation

๐Ÿ“ฐ Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes โ€” full credit and traffic to the original publisher.