Josh Claman is the CEO of Accelsius, he’s an authority global executive with over 30 years of experience in data center technology. He’s driven growth and innovation at NCR, AT&T, and, as a Dell Executive, he managed significant business units across Europe and the Americas, including as General Manager of Dell UK. Post-Dell, he has led ventures in 3D printing and digital health, championing transformative tech. He holds degrees with honors from the University of Illinois and advanced business qualifications from the University of South Carolina and Oxford University.
Accelsius was formed in 2022 to develop and commercialize a cooling solution system based on years of scientific thermal research conducted by Bell Labs. Founded by Innventure — a technology commercialization company that focuses on identifying, funding, and scaling disruptive technologies — our experts have created a brand new liquid cooling system that’s serviceable, scalable and resilient: NeuCool.
Are you able to tell us about the way you became the CEO of Accelsius to commercialize its liquid cooling technology?
Over the past 18 months, the information center industry recognized that traditional, brute force air cooling techniques had reached their limits for data centers as recent emerging AI, machine learning and other compute-intense applications began flooding data centers from Hyperscalers to enterprise and COLO operations. To maintain pace, chip and server performance and resulting wattages, and the associated thermals, were growing at an unanticipated rate. Innventure, which funds progressive and sustainable technology firms, identified an IP from Nokia Bell Labs that delivered a two-phase, direct-to-chip solution that had the potential with further advanced engineering development and productization to revolutionize the industry.
Innventure bought the IP and asked me to return on board to construct the corporate and commercialize the NeuCool technology into an enterprise-class solution. Soon after that, ChatGPT was launched just just a few months later. This accelerated us into an amazing growth phase with high-TDP, AI chips that demanded liquid cooling.
How does Accelsius’ two-phase, direct-to-chip liquid cooling system differ from traditional air cooling methods?
Air cooling is a standard but remarkably energy-inefficient solution used to chill data centers, particularly in dry, hot climates. As chips turn into more powerful and compute demands grow, air cooling becomes increasingly unsustainable and in-efficient. Air just isn’t a very good conductor of warmth compared with liquid, which has as much as 4,000 times greater amount of warmth rejection compared with air.
With NeuCool, Accelsius direct-to-chip technology, the A1 safety-rated liquid coolant is pumped to Vaporators (or engineered cold plates) instead of the heatsinks on electronic components in an air-cooling configuration. While many of the heat (80%) is taken away by the NeuCool 2 phase Vaporators on the chip-level (CPU, GPU) and via your complete cooling loop, fans are still needed to assist remove heat on the board level, but at a really low volume and velocity typically.
Direct to chip liquid cooling has an estimated 50% savings in energy costs and CO2 emissions compared to traditional air cooling.
Are you able to explain the environmental advantages of your cooling technology, particularly by way of CO2 emissions and water usage?
We use as much as half as much electricity as an air-cooled data center – which implies we reduce CO2 emissions by the identical amount. We also devour zero water in our system, resulting in greater environmental impacts.
How does Accelsius Ascent Model help data center operators transition from air to liquid cooling?
Many recent data center and edge computing build-outs are being architected and constructed to facilitate liquid cooling from the beginning. For brownfield or existing data centers, NeuCool is uniquely capable of be easily retrofitted with the familiar form factor of a direct-to-chip solution.
The Accelsius Ascent Model is a collaborative program that allows, empowers and actively guides data center operators through a progressive transition to liquid cooling. This unique program ensures that customers systematically receive maximum support, value and all key advantages of the NeuCool System.
We ensure a smooth transition into liquid cooling from air cooling potentially reaching its limits. The method involves modifications or retrofits to the server, rack and even facility. Our Kickstart program, a part of the Ascent model, offers customers willing to try liquid cooling a probability to plan, test and gain experience in liquid cooling and forecast a hybrid IT environment for a smooth transition away from air cooling.
What challenges did you face while commercializing your technology, and the way did you overcome them?
The historical supply chain crisis impacted our sector particularly hard, and this was something I used to be keenly aware of as I used to be working to scale Accelsius in 2022. We’ve prioritized a multi-layer AVL that may protect us in the long run, and have also committed to having 85% of our top vendors be NAM based to enable a more responsible supply chain.
We even have a dynamic manufacturing and assembly model which blends lower volume, early stage in-house production with an integrated tier 1 global manufacturing partner, which supplies us more control and adaptability.
How do you see the role of liquid cooling evolving as AI and machine learning technologies proceed to advance?
As AI and ML advancements increase, direct to chip liquid cooling will eventually turn into the de-facto cooling technology in data centers. Leaders at NVIDIA and Intel have already declared this in recent announcements, supported by this blog here.
What advice would you give to other entrepreneurs seeking to tackle complex technical challenges within the hardware technology space?
Don’t get too far ahead of yourself and run out of runway: that is certainly one of the important reasons hard tech businesses fail.
While the phrase “Move slow, don’t break things” works well for firms that emphasize speed and experimentation, the precise opposite is true for those pursuing operational innovation.
It’s important to do not forget that simply because a technology or recent way of doing business is novel doesn’t mean it’s at all times appropriate for the client or the market. Operational innovation results from a cautious approach built with iterative steps moderately than transformational ones.
Are you able to share any success stories or case studies where Accelsius’ technology has made a major impact on data center operations?
One noteworthy partner is Texas Advanced Computing Center (TACC). We now support Vista, which is a brand new leading-edge, high-performance supercomputer, that manages TACC’s AI workloads. Vista uses NVIDIA GH100 Grace Hopper Superchips with TDPs as much as 1000 watts per socket.
As per confidentiality standards, we plan to release additional customer profiles soon.
What do you’re thinking that are the most important misconceptions about liquid cooling technology within the industry today?
There are just a few misconceptions about liquid cooling within the industry that I’m glad to deal with.
First, now we have seen some hesitation in combining liquids and servers inside data center racks. Our direct-to-chip architecture is heavily engineered to stop leakage, subsiding this popular fear. Within the rare event that there’s a leak, the dielectric fluid we use will not destroy or damage servers.
One other common misconception is that liquid cooling technologies are difficult to implement into current brownfield data centers. If done accurately, direct-to-chip cooling technology should fit into traditional racks without the necessity for special mounting or difficult retrofitting.
How does Accelsius plan to scale its operations to satisfy the growing demand for advanced cooling solutions?
We’ve a sturdy product portfolio, enabling modular constructing of liquid cooling solutions from one rack to many – with increasing heat removal capability as rack densities proceed to grow. We even have an expansive partner eco-system that allows us to assemble, deploy, integrate and repair at volume and across a large geographic coverage area. Lastly, in 2024 we have launched our presence in Europe, and can proceed to expand our services there.
