AI surpasses limits? "Diamond cooling" technology solves heat dissipation issues

As AI infrastructure reaches its limits, “diamond cooling” technology is becoming the new solution. As AI systems grow larger and generate more heat, the issue goes beyond simply adding more GPUs; it centers on how to manage the heat and power consumption produced. According to Felix Ezekan, co-founder and CEO of Akash Systems, the company reduces temperatures by directly applying diamond-based thermal conductive materials to server GPUs equipped with NVIDIA and AMD chips, enabling more computations to be performed under the same data center power consumption.

“Akash Systems is a deep-tech company rooted in the Bay Area, dedicated to solving thermal management challenges in data centers and AI,” Ezekan explained. One of the core challenges facing data centers today is limited energy supply, which is the crux of the problem. The company states that by using the world’s most thermally conductive diamond materials, they can help existing and new users efficiently convert energy into the computing power they need.

Ezekan recently discussed in TheCube interview series how diamond cooling technology can solve the energy and heat dissipation issues in AI data centers. Since GPUs generate massive amounts of heat during AI training and inference, operators are using this technology to address the problem of cooling systems consuming too much energy and occupying computing resources.

This diamond cooling technology not only lowers chip temperatures but also reduces operational costs and provides additional computing power. Data center operators can cut their cooling system electricity budgets and redirect that energy to support more computations. Ezekan revealed that this technology has the potential to save millions of dollars per server.

In regions with limited energy supply, such efficiency-enhancing technologies have a profound impact on AI expansion. Akash Systems has begun deploying diamond cooling systems on servers using NVIDIA and AMD GPUs, demonstrating the technology’s potential to integrate into a broad hardware ecosystem.

This innovation, rooted in the fundamentals of energy and computing, reflects a commitment to maximizing the use of the world’s limited energy resources. Whether it can double energy efficiency capacity without building new power plants, leveraging existing infrastructure, remains to be seen and warrants ongoing attention.