NVIDIA claims liquid-cooled Blackwells are 25x more energy efficient, 300x more water efficient – techAU
As the top AI providers scale out their training and inference clusters, it comes with a big, hot problem. Traditional data centres just weren’t built for the kind of heat today’s massive AI models generate, pushing air cooling to its absolute limits.
NVIDIA believes the answer isn’t more fans, but plumbing, turning to liquid cooling for its latest generation of AI hardware. It’s a necessary step as server racks move from handling 20kW to well 135kW, so air-cooled solutions just won’t cut it, unable to shift that much heat efficiently.
One key solution is liquid cooling. By reducing dependence on chillers and enabling more efficient heat rejection, liquid cooling is driving the next generation of high-performance, energy-efficient AI infrastructure.
NVIDIA is betting big on this with its new Blackwell platform systems.
NVIDIA GB200 NVL72 & GB300 NVL72
These are full rack-scale systems, designed from the ground up with liquid cooling to tackle trillion-parameter AI models and demanding reasoning tasks. They promise significant leaps in performance and efficiency compared to older air-cooled setups.
Compared to traditional air cooling, NVIDIA claims the GB200 NVL72 system is a 25x more energy efficiency and an incredible 300x more water efficiency. The newer GB300 NVL72 pushes this further, aiming for 30x energy efficiency gains.
Serious Cost Savings
Cooling can eat up a whopping 40% of a data centre’s power bill. NVIDIA estimates that switching to their liquid-cooled GB200 system could save a large 50MW data centre over A$6 million annually (US$4 million+) just on cooling energy and water costs.
This shift isn’t just about saving money; it’s about enabling scale. The tight integration needed for modern AI, using tech like NVIDIA NVLink for rapid GPU communication, results in power densities (up to 120kW per rack) where air cooling becomes impractical.
Liquid, being almost 1,000 times denser than air, is simply far better at carrying that intense heat away. This allows more power to go to actual computing, rather than just trying to prevent a meltdown.
Industry partners are already jumping on board, developing reference architectures and cooling components specifically for these new NVIDIA systems. Companies like Vertiv, Schneider Electric, CoolIT Systems, and Boyd are rolling out solutions designed to handle the heat load efficiently.
Even cloud giants like AWS are collaborating on next-gen liquid cooling, reporting significant compute power increases while cutting energy use. It seems clear that cooling is becoming as crucial as the silicon itself for the future of AI.
While specific Australian availability and pricing for the NVIDIA GB200 NVL72 and GB300 NVL72 systems haven’t been detailed yet, expect them to arrive via NVIDIA’s usual enterprise partners and system integrators.
This technology represents a fundamental shift in how high-performance data centres will be built and operated.
For more information, head to the NVIDIA Blackwell Platform page:
