Compute & Infrastructure

Private equity firm KKR is seeking a multibillion-dollar exit from CoolIT Systems, a data centre cooling technology provider, as the AI infrastructure boom drives up valuations for thermal management specialists, according to Financial Times. The potential sale reflects investor recognition that cooling has emerged as a critical constraint on data centre expansion, particularly for GPU-intensive AI workloads that generate significantly more heat per rack than traditional compute infrastructure. KKR's expected return multiple signals that infrastructure enablers — not just primary compute providers — are capturing significant value from the buildout cycle.

The sale process comes as hyperscalers and colocation providers face increasing thermal density challenges. Advanced cooling solutions including liquid cooling and immersion systems are becoming necessary rather than optional for high-performance AI clusters, creating sustained demand for specialised providers. The valuation KKR achieves will serve as a benchmark for other infrastructure subsector investments.

AI infrastructure buildout is creating a historic memory chip shortage that will raise prices across consumer electronics including smartphones, vehicles, and computing devices, Bloomberg reports. The shortage stems from fundamental capacity constraints as manufacturers allocate production toward high-bandwidth memory (HBM) required for AI accelerators rather than commodity DRAM and NAND for consumer products. Meeting exponential AI demand may prove physically impossible without multi-year capacity expansion, creating a structural supply-demand imbalance.

The allocation conflict is particularly acute because HBM production requires similar fabrication equipment but commands significantly higher margins, incentivising manufacturers to prioritise AI customers. Samsung, SK Hynix, and Micron face strategic decisions about capacity allocation that will determine both near-term profitability and long-term customer relationships. Consumer electronics manufacturers are already experiencing allocation constraints and price increases, with automotive suppliers particularly vulnerable due to lower pricing power.

GEO Group, a private prison and ICE detention facility operator, is targeting the AI data centre market by providing temporary worker housing or 'man camps' for construction crews building facilities in remote locations, TechCrunch reports. The approach mirrors housing solutions developed for oil field workers in remote areas, reflecting the scale and remote geography of many AI data centre projects. The involvement of detention facility operators illustrates both the magnitude of construction labour demand and the secondary infrastructure requirements created by AI buildout.

Data centre developers are increasingly siting facilities in locations with available power capacity and favourable energy costs rather than proximity to urban centres, creating logistical challenges for housing construction workforces during multi-year buildout timelines. The economics of temporary worker camps become viable when construction timelines extend beyond 18-24 months and commuting distances exceed practical limits.