Direct-to-chip cooling market to hit $22.5 billion by 2035
The global Direct-to-Chip Cooling Systems market is projected to reach $22.5 billion by 2035, growing at a 26% CAGR, driven primarily by the thermal demands of AI accelerators and High-Performance Computing (HPC) clusters. This growth impacts streaming as content processing and AI-driven services increasingly rely on high-density data centers requiring advanced thermal management for their processors. Hyperscalers like Google, Microsoft, and Amazon Web Services are leading the transition to liquid cooling, which will become standard in new data center builds.
Key Takeaways
- Global market valuation for direct-to-chip cooling expected to grow eightfold from $2.8 billion in 2025 to $22.5 billion by 2035.
- Data center applications drive 60–65% of global demand, led by hyperscalers like Google, Microsoft, Meta, and AWS.
- Direct-to-chip cold plates provide thermal resistance below 0.1 °C/W, necessary for cooling processors that breach the 600 W TDP mark.
- Asia-Pacific holds the largest market share at 40%, fueled by aggressive data center builds in China, Japan, and South Korea.
- PFAS regulatory restrictions in Europe are emerging as a significant compliance hurdle for cooling chemistry and supplier qualification.
Why It Matters
The shift to liquid cooling marks a fundamental change in the streaming infrastructure stack as AI-driven personalization and real-time encoding demand high-density compute power that air can no longer regulate. For streaming platforms, this necessitates a transition toward 'AI factory' architectures where thermal management is as critical as networking fabric. Service providers must now account for specialized hardware costs—ranging from $1,000 to $2,500 per kilowatt of cooling capacity—in their long-term CAPEX planning. Watch for the standardization of Open Compute Project (OCP) liquid cooling designs to determine how quickly vendor lock-in will diminish for mid-tier colocation users.
Additional Context
The move toward direct-to-chip cooling is being accelerated by the sheer physical heat flux of next-generation silicon. Per Global Market Insights (June 2026), the broader data center liquid cooling market is set to expand rapidly as rising AI rack densities and stringent Power Usage Effectiveness (PUE) thresholds make air cooling economically unviable above 35 kW per rack. Strategic shifts by major infrastructure providers confirm this trajectory; for instance, Vertiv reported a 60% year-over-year increase in organic orders by Q3 2025 as it transitioned from traditional cooling to high-density thermal management. Hardware roadmaps and environmental pressures are also forcing the hand of hyperscalers. Per TrendForce (September 2024), NVIDIA's Blackwell platform was expected to drive liquid cooling penetration from 10% to over 20% by 2025, largely because a single GB200 NVL72 rack can pull up to 140 kW. Simultaneously, Microsoft and Google have faced intense scrutiny over data center water consumption. Per Tom's Hardware (June 2026), Microsoft has pivoted toward closed-loop liquid systems that CEO Satya Nadella claims can operate with near-zero water consumption after the initial fill. This focus on water-neutrality is now a prerequisite for new builds in drought-stressed regions like Texas and Arizona. Energy efficiency benchmarks are also shifting to reflect these thermal realities. While traditional air-cooled facilities struggle to maintain PUEs below 1.5, best-in-class liquid-cooled sites are targeting 1.08. Schneider Electric and Microsoft jointly published reference designs in late 2024 that integrate power distribution with direct-to-chip loops to enable predictable performance at scale. As AI workloads increasingly pivot from training to inference—scheduled to account for 75% of AI energy demand by 2030 per SentiSight—the ability to deploy these high-density cooling modules at the edge will become the primary competitive differentiator for streaming latency and reliability.
Read full article at indexbox.io