Tuesday 10 November 2026 950 stories on file Full archive
Daily Edition
newscms

Volume III Edition Daily

Equinix Opens Five Liquid Cooling Innovation Centres to Accelerate Enterprise AI Deployments

On October 21, 2026, Equinix introduced the Equinix Enhanced Performance Innovation Centers (EPIC), a global network of production-ready liquid cooling environments across five International Business Exchange (IBX) data…

Cloud & Edge Computing 1,330 words 7 min read

Equinix Opens Five Liquid Cooling Innovation Centres to Accelerate Enterprise AI Deployments — Cloud & Edge Computing No Image Cloud & Edge Computing
Lead image · Filed 23 October 2026, 05:43

Equinix Opens Five Liquid Cooling Innovation Centres to Accelerate Enterprise AI Deployments

The EPIC Network: Five Sites, Live Data Centres

On October 21, 2026, Equinix introduced the Equinix Enhanced Performance Innovation Centers (EPIC), a global network of production-ready liquid cooling environments across five International Business Exchange (IBX) data centre facilities. The five EPIC locations are live and operational in:

  • Frankfurt (FR6)
  • London (LD5)
  • Paris (PA10)
  • Osaka (OS3)
  • Sydney (SY5)

Each EPIC occupies space within an existing Equinix IBX facility, exposing customer equipment to the same cooling, power and networking conditions it will meet in production. Enterprises can bring in liquid-cooled GPU and CPU platforms from Equinix's hardware partners and run AI training and inference workloads under real-world conditions, with Equinix engineers on hand.

Network equipment installation at a compute facility — the rack and cabling layer underpinning modern data centre infrastructure.

Server rack in a data centre facility — the kind of rack environment where liquid-cooled GPU platforms must coexist with existing air-cooled infrastructure.

The launch addresses a growing enterprise gap: as AI workloads move beyond pilot stages, more powerful GPU platforms are pushing rack densities upwards and increasing demand for liquid cooling, yet few enterprises have the expertise to make cooling, power, networking and facility operations work together. A gap in any one of those areas can hold up a deployment.

From Prototype to Production: Testing AI Infrastructure at Scale

EPIC helps address that challenge by giving enterprises a place to prove it out. Inside live Equinix data centres, enterprises can trial the latest liquid-cooled GPU and CPU platforms with leading hardware partners and test AI training and inference workloads under real-world conditions. With Equinix experts alongside, they can identify infrastructure and operational issues before production and make informed decisions as they scale. The result is a faster, lower-risk path from proof of concept to production.

"Liquid cooling is an essential component of modern enterprise infrastructure and a hard requirement for customers deploying AI," said Brian Stein, Senior Vice President, Infrastructure Products and Services at Equinix.

EPIC can help enterprise teams:

  • Work with leading technology partners: Work with Equinix and its ecosystem of technology partners and original equipment manufacturers (OEMs) to assess combined liquid cooling stacks under field conditions. See how the systems integrate to help get to deployment faster.
  • Get hands-on infrastructure experience: See liquid cooling technologies in action, understand how each fits into specific environments and address operational questions that matter before deployments.
  • Test their infrastructure where it will run: Each EPIC location operates inside a live data centre, giving customers exposure to liquid cooling under real production conditions.

Industry Context: The Cooling Readiness Gap

Equinix positions EPIC as a response to an operational shortfall. More powerful GPU platforms are pushing rack densities upwards — a rack of current-generation GPUs can draw more than 200kW, whereas most data centres built a few years ago were designed for 5-10kW per rack. That orders-of-magnitude jump means existing infrastructure is often insufficient without a fundamental rethink of cooling architecture.

Liquid cooling has become a more prominent issue for data centre operators and enterprise IT teams as AI workloads move beyond pilot stages. Higher-density compute systems place greater demands on cooling and power design, prompting operators to examine whether existing air-cooled setups are sufficient for newer hardware.

For enterprises, the challenge is often practical rather than theoretical. Deploying liquid-cooled systems requires coordination across several layers of infrastructure, from the hardware platform itself to networking, power distribution and the building environment in which the equipment runs. The EPIC model is intended to let enterprise teams test infrastructure in the same type of live environment in which it will ultimately operate. Customers can review how different parts of the stack interact before making long-term investment decisions.

Criteo, an early EPIC user, put it this way: "Liquid cooling was new territory for us, so we built our confidence step by step. Our AI runs across a distributed footprint, and the infrastructure must perform the same way wherever it sits. We saw how liquid cooling performed in practice, learned what it took to run it well, and then committed to scale. Equinix was with us at every stage, with the expertise and the live environment to prove it out first. That's what turned liquid cooling from an open question into a core part of our distributed AI infrastructure."

The Supply Side: CDUs and the Density Squeeze

The EPIC launch lands as the hardware layer of liquid cooling matures into products that can be specified rather than prototyped. At the OCP Global Summit in October 2026, Cooler Master unveiled a 2,500kW coolant distribution unit (CDU) aimed squarely at AI data centres — a single unit sized to feed a row of the 200kW-plus racks that EPIC is built to host. A CDU is the interface between a facility's primary water loop and the secondary loop that serves each rack manifold, and its capacity effectively sets how much compute a given data hall can carry.

That matters because the density curve has broken the assumptions most colocation facilities were designed around. A hall built for 5-10kW per rack can be retrofitted with rear-door heat exchangers and in-row cooling, but those approaches top out well below the 130-200kW densities that current GPU platforms demand. Direct-to-chip liquid cooling, where cold plates sit on the processor packages and carry heat away in a sealed loop, is the architecture that closes the gap — and it requires plumbing, leak detection, quick-disconnects, manifolds and controls that most enterprise IT teams have never operated.

Equinix is not the only operator building validation capacity. The broader pattern across 2026 has been colocation providers converting floor space into test beds, because the alternative is customers signing multi-year leases and discovering during fit-out that the hall cannot take the load. For an enterprise, the cost of that discovery is measured in delayed AI programmes and stranded capital, not in cooling hardware. EPIC's commercial logic rests on that asymmetry: charge for the validation step so the production commitment is de-risked.

The programme also fits Equinix's wider pivot toward higher-value interconnection and AI infrastructure. The company operates more than 260 IBX data centres globally, and the EPIC sites are drawn from that existing estate rather than new construction — a capital-light way to productise expertise the operator has accumulated through its Co-Innovation Facilities and Solution Validation Centres since 2022.

Partner Ecosystem: Who's Already Testing

EPIC builds on Equinix's existing innovation spaces. In 2022 Equinix announced its first Co-Innovation Facility where partners could collaborate on testing new technology before it entered Equinix data centres. In 2024, Equinix rolled out Solution Validation Centres for customers to validate deployments and see how Equinix supports distributed environments. EPIC brings that same experiential approach to liquid cooling and high-density AI deployments.

Key partners already engaged with EPIC include:

  • Criteo — The ad-technology company is testing liquid cooling with Equinix before scaling it across its distributed AI infrastructure.
  • NVIDIA — In January 2024 Equinix launched a fully managed private cloud service that lets enterprises run NVIDIA DGX AI infrastructure in its IBX data centres.
  • ZutaCore — The liquid cooling specialist tested its direct-to-chip, two-phase rack systems, rated for more than 100kW per rack, at Equinix's Co-Innovation Facility from 2022.
  • Bloom Energy — Bloom trialled solid oxide fuel cells for generator-less and UPS-less designs at the Co-Innovation Facility and has supplied fuel cells to Equinix sites including SV5 in San Jose.
  • NUS Centre for Energy Research & Technology — The National University of Singapore partnered with Equinix in 2024 on a Co-Innovation Facility at SG6 to incubate liquid cooling and other technologies for tropical climates.

What's Next: Expansion to 2027 and Beyond

Equinix plans to open more EPIC locations in 2027, expanding the network beyond the initial five sites. The company is also deepening its partner ecosystem, with additional OEMs and cooling specialists expected to join the programme. For enterprises, the message is clear: liquid cooling is no longer a theoretical consideration but a production necessity for AI infrastructure at scale, and EPIC provides the fastest, lowest-risk path from evaluation to deployment.

This article is based on Equinix's official press release (October 21, 2026) and Data Centre Magazine's coverage of the OCP Global Summit 2026, where Equinix showcased the EPIC portfolio alongside Cooler Master's 2,500kW CDU. More on Equinix infrastructure at https://www.equinix.com/newsroom and the Cloud & Edge section at /category/cloud-edge-computing.