If you pay attention to what the biggest tech companies are saying about the demand for AI, you’ll notice a common thread: they don’t have computing power. This means that the large language models that underpin today’s AI products need even more data centers for training and inference, and therefore need more power. In this context, energy efficiency suddenly became a critical priority for semiconductor manufacturers.
PowerLatticea startup founded by veteran electrical engineers from Qualcomm, NUVIA and Intel in 2023 claims to have developed a breakthrough approach that reduces the power needs of computer chips by more than 50%. On Monday, the startup came out of stealth with a $25 million Series A funding round led by Playground Global and Celesta Capital, bringing its total funding to $31 million.
“That’s the hard thing: How do you power the device? There are very few groups and individuals who can do that,” said Pat Gelsinger, general partner of Playground Global. “We’ve assembled what I would argue is the dream team in power delivery.”
As a former Intel CEO, Gelsinger wields significant authority in the semiconductor world, making his involvement a strong stamp of approval for PowerLattice.
In fact, when the startup’s CEO Dr. Peng Zou and her founding team pitched their idea to Playground’s offices in March, so impressed by Gelsinger’s reputation that they asked him for a selfie, Gelsinger told TechCrunch. The admiration proved mutual, as Gelsinger came away truly impressed with PowerLattice’s technology.
The startup’s technology sounds simple in concept: a tiny power chiplet designed to bring power closer to the processor, significantly minimizing power loss.
After two years, PowerLattice has already achieved its first milestone: Its first batch of chiplets is being produced by TSMC, in partnership with an unnamed manufacturer testing the startup’s functionality, Gelsinger said.
Techcrunch event
San Francisco
|
13-15 October 2026
Beyond its initial customer, the startup plans to roll out its product for testing to other customers in the first half of 2026. The trials will prove instructive, given that PowerLattice’s potential customer set includes major chipmakers Nvidia, Broadcom, and AMD, as well as specialized AI chip developers such as Cerberus, Grok, and Playground-Startupx, and Next-Masterback.
Although every chip company has internal teams working to improve energy efficiency, Gelsinger hopes PowerLattice’s innovative approach will pique their interest.
“They might say, ‘I’m going to add some volume to this approach, some volume to my more traditional approach,'” he said. “But we believe our ability to capture substantial share will emerge quickly.”
PowerLattice isn’t the only startup trying to help chipmakers tackle the energy problem. The company would compete more closely with Empower Semiconductor, a startup that built one $140 million Series D;led by Fidelity Management & Research Company, in September.
However, Gelsinger is confident that PowerLattice’s 50% energy efficiency gain is an “excellent” result and expects the company to soon raise a much larger round of funding to fund production.
“The idea is bold, the benefits are great, and I expect others to say, ‘That’s a great idea. Let me try, too,” Gelsinger said.
