Equinix to Pilot Liquid-Cooled AI Infrastructure With Launch of Global EPIC Centers

On the 27th, Equinix unveiled EPIC, the Equinix Enhanced Performance Innovation Centers, a global network of innovation centers. EPIC is designed to support testing and deployment of high-density AI infrastructure. EPIC is operating at five sites: Osaka, Japan; Sydney, Australia; Frankfurt, Germany; London, the United Kingdom; and Paris, France.

Equinix plans to open additional sites in 2027. The announcement comes as AI moves from the experimental stage to actual service operations. As a result, integrated design and operation of power supply, cooling, and network connectivity are becoming increasingly important core capabilities for data centers.

Behind this trend is higher rack-level power density driven by improvements in GPU performance. Demand for liquid cooling is also expanding. As a result, companies face growing pressure to tightly integrate cooling, power, network, and facility operations.

Because some infrastructure components may not be fully validated, deployment schedules could be delayed and unexpected issues could arise during operations. Accordingly, there is a growing need to verify infrastructure in advance under conditions similar to the actual operating environment, and pre-validation is emerging as a key challenge in building AI data centers.

EPIC was designed to meet this need by enabling testing of liquid cooling-based infrastructure in a real operating environment. Companies can tap support from Equinix experts and key technology partners, and can validate AI training workloads and AI inference workloads under conditions similar to real operations using the latest liquid cooling-based GPU and CPU platforms as test targets.

Through such testing, companies can assess cooling performance before full-scale deployment, review the system integration process, and identify operational issues. They can also use the test results to determine whether to expand the infrastructure and how to build it, and the source said these capabilities can reduce uncertainty in the process of moving from proof of concept (PoC) to actual service operations.

EPIC is an expanded space built on the innovation facilities Equinix has already operated. Equinix announced plans in 2022 to establish a Co-Innovation Facility, which was intended to jointly test new technologies from technology partners before introducing them into its own data centers.

Equinix also introduced Solution Validation Centers in 2024. These centers are responsible for helping customers validate deployments and see how Equinix supports distributed infrastructure.

Building on this foundation, Equinix says EPIC supports collaboration across an ecosystem made up of key technology partners and OEMs. Customers can verify not only the performance of individual equipment but also whether integrated infrastructure, including liquid cooling, works in real-world settings.

Brian Stein, Equinix senior vice president of infrastructure products and services, said liquid cooling is an essential element of modern enterprise infrastructure and a requirement for AI deployment customers. He said EPIC provides a space where companies can reliably test, validate, and build before deciding on large-scale deployment, and that Equinix is helping customers speed the transition from planning to actual operations through its global infrastructure and ecosystem.

EPIC's experience- and validation-focused approach is expanding into liquid cooling and high-density AI infrastructure deployment. By validating technologies in real operating environments, companies can secure hands-on experience needed for infrastructure adoption and define operating conditions and requirements before large-scale deployment.

Equinix customer Criteo shared its experience validating liquid cooling in this manner. Mathieu Bloomberg, Criteo's senior vice president of engineering infrastructure, security, and internal IT and chief information security officer (CISO), said liquid cooling was a new area for Criteo. He said the company needed to build trust step by step, and that AI operating across a distributed environment required the same performance regardless of infrastructure location.

Accordingly, before deciding to expand, Criteo confirmed how liquid cooling actually worked and identified what was needed to operate it effectively. Criteo decided to expand after that confirmation and assessment. Mathieu Bloomberg said Equinix supported Criteo at every stage with expertise and real operating environments so the company could validate first, and that as a result liquid cooling shifted from a topic under review to a core element of Criteo's distributed AI infrastructure.

Source: TECHWORLD · Park Kyu-chan
Original: https://www.epnc.co.kr/news/articleView.html?idxno=408797

References

• https://www.epnc.co.kr/news/articleView.html?idxno=408797