LiquidStack’s low-power, closed-system server-cooling technology, has caught the eye of Microsoft Azure. Credit: KTSimage / Polygraphus / Getty Images Ever since Bitcoin was introduced back in 2009, this cryptocurrency has had the distinction of being something you could mine with your computer—putting your hardware to use helping the blockchain technology Bitcoin is based on record and verify transactions by solving complex math problems. As a reward, you’d get Bitcoins. But it was a very slow process for a single PC, and the necessary component for success was a high-end GPU. One GPU brought to bear on Bitcoin could take years to find one coin, so miners started building massive farms akin to data centers but without enclosures. The result was that Bitcoin farms bought up all the GPUs, causing severe shortages and infuriating gamers. GPUs run notoriously hot and for some time GPU cards have come with three fans on them. It is possible to replace the heat sink with a water cooling system, but The Bitfury Group, a blockchain services provider, went one better and instead used liquid immersion—putting the hardware in a dielectric fluid. It bought the technology by purchasing Allied Control in 2015, and now that division of Bitfury is being set up as an independent company called LiquidStack. Wiwynn, a hyperscale server vendor from China, is investing $10 million in Series A funding to launch LiquidStack and its immersion product, DataTank. DataTank immerses the computer electronics in a non-conductive bath that eventually comes to a boil as the chips heat up. The liquid boils off as a vapor that carries away the heat, but is captured within the sealed DataTank and condensed back into liquid. In 2014, LiquidStack (then known as Allied Control Limited) built a 500kW data center in Hong Kong using its two-phase cooling, saving more than 95% on energy for cooling when compared to air cooling. LiquidStack said it was able to run like this continuously for eight months without needing to add more coolant. That data center achieved a power-usage effectiveness (PUE) of 1.01, absolutely unheard of in cooling. That means that virtually all of the power consumed by the data center went into powering the servers and almost none of it went into cooling. The evaporation/condensation method means no pumps or fans. Allied Control was acquired by Bitfury in 2015 and the companies jointly deployed over 160MW of two-phase immersion cooled data centers, including its specially designed DataTank units operating at densities of up to 252kW per 48U rack. An air-cooled 48U rack would be lucky to reach 10kW. So you are talking incredible power density and no noise because no fans are needed. LiquidStack’s technology enables at least 21 times more heat rejection per IT rack compared to air cooling, with no water consumed for outside heat rejection. This cooling method results in a 41% reduction in energy used for cooling and up to 60% reduction in whitespace used for compute infrastructure. LiquidStack has at least one powerful ally. Microsoft is experimenting with DataTank as an option to support higher power densities in its Azure cloud computing operation and Wiwynn is a bit player in the hyperscaler market. Although hyperscalers and HPC are its main targets, LiquidStack is initially aiming at smaller entities by offering products for edge computing and 5G wireless infrastructure deployments with smaller enclosures to support micro data centers. Related content news Supermicro unveils AI-optimized storage powered by Nvidia New storage system features multiple Nvidia GPUs for high-speed throughput. By Andy Patrizio Oct 24, 2024 3 mins Enterprise Storage Data Center news Nvidia to power India’s AI factories with tens of thousands of AI chips India’s cloud providers and server manufacturers plan to boost Nvidia GPU deployment nearly tenfold by the year’s end compared to 18 months ago. By Prasanth Aby Thomas Oct 24, 2024 5 mins GPUs Artificial Intelligence Data Center news Gartner: 13 AI insights for enterprise IT Costs, security, management and employee impact are among the core AI challenges that enterprises face. By Michael Cooney Oct 23, 2024 6 mins Generative AI Careers Data Center news Network jobs watch: Hiring, skills and certification trends What IT leaders need to know about expanding responsibilities, new titles and hot skills for network professionals and I&O teams. By Denise Dubie Oct 23, 2024 33 mins Careers Data Center Networking PODCASTS VIDEOS RESOURCES EVENTS NEWSLETTERS Newsletter Promo Module Test Description for newsletter promo module. Please enter a valid email address Subscribe