$500bn Stargate Initiative: The key to solving AI infrastructure and cooling challenges?

Capes highlighted how AI is reshaping power and cooling infrastructure. “AI compute power density has jumped by an order of magnitude in the last 12 months,” he explained, noting that the surge in demand is pressing supply chains to meet the infrastructure needs of high-power GPU and CPU clusters. Liquid cooling, a key enabler of this shift, is expected to scale rapidly to support the growing demand.

Capes acknowledged that while the Stargate Initiative is transformative, it faces hurdles in execution. “The biggest challenges are designing facilities for non-homogeneous applications that may include a combination of AI and non-AI related workloads, time to power and grid interconnection, and capacity issues within the supply chain for cooling and power-related infrastructure,” he said.

One of the most pressing issues is cooling. Many existing data centres rely on air-cooled systems, which are ill-equipped to handle the demands of modern AI workloads.

“Today, most AI workloads are being deployed in air-cooled data centres with insufficient infrastructure to support new power and liquid cooling requirements,” Capes explained. This mismatch often results in higher Power Usage Effectiveness (PUE) and energy inefficiencies.

Capes emphasised that new data centres designed for homogeneous AI applications can optimize liquid cooling systems for higher water temperatures, significantly reducing energy and water consumption.

LiquidStack has emerged as a leader in liquid cooling technologies, providing cutting-edge solutions to address the unique challenges of AI infrastructure. “We supply direct-to-chip and immersion liquid cooling solutions,” Capes said, adding that the company is delivering multi-megawatt Coolant Distribution Units (CDUs) for AI data centre buildouts.

These CDUs act as the critical heat exchange interface between the data centre’s Technology Cooling System and its primary mechanical plant.

The Stargate Initiative underscores a broader trend of prioritising energy efficiency and sustainability in AI infrastructure. Liquid cooling plays a pivotal role in these efforts.

“By leveraging liquid cooling with higher operating fluid temperatures, we can reduce the need for mechanical cooling and evaporative water use,” Capes explained. This approach not only lowers PUE but also reduces Water Usage Effectiveness (WUE), contributing to a smaller carbon and water footprint for data centres.

As AI continues to drive unprecedented demand for infrastructure, initiatives like Stargate and innovative solutions from companies like LiquidStack will be critical in ensuring that growth is both scalable and sustainable. “Liquid cooling has a positive impact on global ecosystems by reducing direct and indirect carbon footprint, and water use,” Capes concluded.

RELATED STORIES

Softbank, OpenAI, Oracle & Microsoft back $500bn Stargate AI project

Musk attacks OpenAI’s $500bn Stargate project claiming it ‘doesn’t have the money’