SUNON is taking a chip-to-rack approach to liquid cooling as AI workloads push data center power density higher. At the 2026 OCP Global Summit, the thermal-management company will showcase cooling technologies spanning chip, server and rack levels, positioning integrated thermal infrastructure as an increasingly important part of AI data center design.
AI infrastructure is running into a physical problem that cannot be solved by faster processors alone: more computing power produces more heat.
As accelerator performance and rack densities increase, data center operators are having to rethink how heat is removed from servers. Air cooling remains widely deployed, but liquid cooling is gaining importance for high-density AI systems because it can move heat away from processors more directly and efficiently.
SUNON is using the upcoming 2026 OCP Global Summit to showcase a broader liquid cooling portfolio aimed at this transition. Under the theme “Build to Chill,” the company plans to demonstrate cooling technologies covering chip-level, server-level and rack-level infrastructure.
The event will take place October 12–15 in San Jose, California, with SUNON exhibiting at Booth A19. The Open Compute Project says this year’s summit is focused on “Scaling Innovation for the AI Era,” reflecting the industry’s broader effort to redesign computing and data center infrastructure around increasingly demanding AI workloads.
Why AI is making cooling a bigger infrastructure issue
The thermal challenge is closely connected to the growth of AI compute.
Gartner forecasts worldwide data center electricity consumption will reach 565 terawatt-hours in 2026, a 26% increase from 2025. AI-optimized servers are expected to account for 31% of data center power consumption this year, while their power consumption is projected to surpass that of conventional servers in 2027.
The International Energy Agency is seeing a similar structural shift. Its analysis estimates that data centers consumed around 415 TWh of electricity in 2024, equivalent to roughly 1.5% of global electricity consumption. It also notes that AI is accelerating the deployment of high-performance accelerated servers, increasing power density inside data centers.
Power consumed by computing equipment ultimately becomes heat that must be managed.
That makes thermal infrastructure an increasingly important part of AI infrastructure alongside GPUs, CPUs, networking, storage and power distribution.
SUNON takes a chip-to-rack approach
Rather than presenting liquid cooling as a single component, SUNON is organizing its OCP Summit portfolio around three infrastructure layers: chip, server and rack.
At the chip level, cooling technologies can be designed around the thermal characteristics of processors and accelerators. At the server level, the challenge becomes integrating thermal management with system architecture, airflow and power delivery. At rack level, cooling becomes a facility-scale consideration involving coolant distribution, heat rejection and interactions between multiple servers.
SUNON says its approach is intended to accommodate different power densities, environments and system requirements.
The company is also showing what it describes as a broader liquid cooling ecosystem. Its current OCP exhibition information references closed-loop liquid cooling systems and coolant distribution units alongside its chip-, server- and rack-level strategy.
That distinction matters because cooling performance at one level does not automatically solve thermal constraints elsewhere in the infrastructure.
A high-performance cold plate, for example, still needs an appropriate coolant-distribution architecture. A rack-level system also needs to fit into the facility’s broader power, control and heat-rejection design.
From component cooling to infrastructure architecture
The move toward liquid cooling reflects a wider change in how data centers are being designed.
Traditional enterprise workloads often allowed operators to treat servers as relatively independent thermal loads. AI clusters are different. High-density accelerators can concentrate substantial computing capacity into comparatively small physical footprints, making thermal behavior a rack-level and facility-level concern.
That is why the OCP community has been examining infrastructure beyond individual servers. Its 2026 program includes work around AI systems, machine-learning infrastructure and power and cooling for AI, while previous OCP workshops have explored designs for 1-megawatt AI racks.
The implication is that liquid cooling is increasingly becoming an architectural decision rather than simply a cooling-product choice.
Operators need to consider coolant distribution, redundancy, maintenance, leak detection, heat rejection, facility water requirements and compatibility with existing data center infrastructure. Equipment manufacturers likewise need to make cooling technologies adaptable to different server and rack configurations.
Efficiency becomes a design requirement
Energy efficiency is another reason cooling has moved closer to the center of AI infrastructure planning.
The IEA reported that global data center electricity demand increased 17% in 2025, while electricity consumption from AI-focused data centers grew 50%. It also notes that the energy requirements of AI applications vary considerably as models and use cases become more sophisticated.
Cooling is only one component of total data center energy use, but improving thermal management can affect the efficiency and operating envelope of the entire facility.
For hyperscalers and other large operators, that creates a need for cooling systems that can scale alongside compute deployments. For system manufacturers, it also creates demand for modular technologies that can be integrated into different server generations and rack architectures.
Companies such as NVIDIA, AMD, Intel and major cloud providers are increasing the computing density of AI systems, while the data center industry is simultaneously working on power delivery and thermal infrastructure capable of supporting them. The result is a more interconnected AI infrastructure stack in which compute, power and cooling increasingly have to be designed together.
The OCP ecosystem provides a testing ground
SUNON’s presence at OCP Global Summit is therefore less about a single cooling product and more about where thermal technology fits within the next generation of open data center infrastructure.
OCP describes its community model as centered on openness, efficiency, sustainability and scalability as data center operators adapt infrastructure for AI at increasingly large scale.
That environment gives vendors an opportunity to position cooling technologies alongside the broader hardware and infrastructure ecosystem.
SUNON’s chip-to-rack strategy also reflects a practical reality: AI data center cooling cannot be optimized in isolation. Processor thermal design, server architecture, rack density, coolant distribution and facility infrastructure all influence one another.
As AI workloads continue to grow, the competition around data center performance will increasingly involve the infrastructure that keeps processors operating within their thermal limits.
Liquid cooling is consequently moving from a specialized technology for selected high-density deployments toward a more central component of AI infrastructure planning. SUNON’s OCP showcase illustrates that shift, with the company presenting thermal management as a connected system extending from individual chips to complete racks.
Market Landscape
AI data center expansion is increasing pressure on both electricity supply and thermal infrastructure. Gartner forecasts global data center electricity consumption of 565 TWh in 2026, up 26% year over year, with AI-optimized servers representing 31% of data center power consumption.
The IEA says AI-focused data center electricity consumption grew 50% in 2025, highlighting how quickly high-performance computing is changing infrastructure requirements.
This is expanding the AI infrastructure market beyond accelerators and networking into liquid cooling, coolant distribution, power management, rack architecture and facility-level thermal systems. OCP’s emphasis on open, scalable infrastructure further increases the importance of interoperable cooling technologies that can support different AI server and rack configurations.
Top Insights
- SUNON will showcase liquid cooling technologies spanning chip, server and rack levels at the 2026 OCP Global Summit.
- AI accelerator density is increasing thermal loads, making cooling architecture an increasingly important part of AI infrastructure design.
- Gartner forecasts global data center electricity consumption will reach 565 TWh in 2026, up 26% year over year.
- SUNON’s portfolio includes liquid cooling systems and coolant distribution technologies intended for high-performance computing environments.
- OCP’s AI infrastructure work increasingly connects computing, power, cooling and rack-level system design.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
