Solving AI Data Center Power at the Point of Load
AI data center power sits at the top of the modern infrastructure agenda. Hans Hasselby-Andersen, CEO of Lotus Microsystems, joins Alan Shimel on Techstrong TV. Furthermore, they show how AI data center power efficiency now beats what legacy voltage regulators can deliver.
About Hans Hasselby-Andersen
Hans has spent more than 25 years in the semiconductor industry. In addition, he founded Merus Audio, which Infineon acquired in 2018. Consequently, he brings a founder mindset to a Danish deep tech spinout from the Technical University of Denmark.
Inside AI data center power
Lotus Microsystems designs fully integrated power modules for AI and data center processors. As a result, its silicon power interposer places the converter, inductors and capacitors on one substrate. Meanwhile, that layout moves huge currents through the interposer instead of long copper traces.
Hans explains that modern accelerators run at very low voltages near 0.6 to 0.7 volts. Furthermore, a two kilowatt part at that voltage pulls a kiloampere class current. Consequently, any resistive loss on the board turns into heat right beside the processor.
Why the vStrata platform matters
Meanwhile, the first vStrata module, the LSC0580, sits on the motherboard next to the xPU as the last stage of conversion. Therefore, engineering samples ship in the next few months to hyperscaler and xPU partners. In addition, Lotus is talking to most of the hyperscalers that design their own AI silicon.
Hans argues that AI data center power efficiency is the real cure for heat, not just better cooling. Furthermore, roughly 35 to 40 percent of data center energy goes to cooling today. As a result, higher compute per watt eases the pressure on grids, water and NIMBY politics at once.
Explore more artificial intelligence coverage and the latest Techstrong TV interviews.
For more information please visit lotus-microsystems.com
