[News] JFS Announced New Breakthrough in GaN Power Module, Poised for Pilot-Scale Validation

On December 4, 2025, Jiufengshan (JFS) Laboratory announced a major technological breakthrough — the successful development of a gallium nitride (GaN) power module. Roughly the size of a thumb, these “black boxes,” each about as large as a million thumbnails, can be integrated into a 1-gigawatt (1 billion watts) hyperscale AI compute cabinet. Once deployed, a full rack is expected to save nearly 300 million kWh of electricity annually, equivalent to about CNY 240 million in power costs.

According to project lead Dr. Li Sichao, the technology has completed proof-of-concept and is about to enter pilot-scale validation. Mass production is expected within 3–5 years, targeting a market with potential demand in the hundreds of billions of yuan.

Surging Gigawatt-Scale Compute Drives Growing Energy Gaps in AI Data Centers

Today’s AI compute centers operate under a model essentially defined by “trading power for compute.” A 1-GW AI data center consumes 8.76 billion kWh of electricity per year—roughly the annual full-load output of a large nuclear power plant. Among all subsystems, the power modules that convert high-voltage electricity into the low-voltage supply required by XPUs (specialized processors) alone account for 1 billion kWh annually, or 11% of total consumption, making them one of the most energy-intensive components in the system.

As global compute capacity accelerates into the “gigawatt era,” energy demand is rising in lockstep. Massive infrastructure investments by tech giants and national-level compute initiatives underscore the urgency of the challenge, making energy efficiency a core imperative for the compute industry.

In October 2025, Google and AI unicorn Anthropic sealed a multibillion-dollar compute partnership, including a dedicated 1-GW power allocation supporting one million TPU chips—energy consumption on par with an entire large-scale nuclear plant.

OpenAI is expanding even more aggressively, advancing two 10-GW custom compute projects with Broadcom and AMD, while jointly developing the 8-GW “Stargate” compute campus with Oracle and SoftBank. These publicly disclosed projects alone exceed 28 GW of power demand—more than twice New York City’s peak electricity load.

Yet most existing compute centers still rely on silicon-based power modules, a legacy architecture with well-recognized limitations: high conversion losses that waste large amounts of electricity, bulky module size that impedes high-density rack deployment, and rising long-term system costs.

GaN Technology Sparks an “Energy Reduction Revolution,” Addressing Gigawatt-Scale Power Challenges

GaN, a new-generation semiconductor material, offers superior electron mobility and high breakdown voltage. When used to create silicon-based GaN chips inside power modules, it reduces conversion losses by up to 30%—the key to the significant energy savings. GaN’s material advantages also shrink module volume by roughly 30%, freeing valuable rack space and enabling higher-density compute deployment.

To seize the next strategic technology frontier, JFS Lab established a cross-disciplinary R&D team early this year, led by Dr. Li Sichao and focused specifically on GaN power modules. Despite the complexity of GaN semiconductor fabrication—which involves more than a thousand process steps and nanometer-level precision—the lab’s seven PhDs completed hundreds of rounds of testing and optimization, ultimately achieving a full closed-loop GaN technology chain.

JFS Lab has already secured partnerships with multiple domestic data-center power suppliers, with commercial orders exceeding CNY 10 million. The lab has also collaborated with the Optics Valley state-owned capital platform to incubate eight compound-semiconductor startups, accelerating the transfer and commercialization of frontier technologies. With the rapid global expansion of AI compute infrastructure, the GaN power module is expected to address a market opportunity valued in the hundreds of billions of yuan, signaling enormous growth potential for the power-module sector.

(Photo credit: The Government of Wuhan)