NextFin News - In a move that signals a paradigm shift in the power semiconductor landscape, Innoscience (Zhuhai) Technology Co., Ltd. has officially secured a major design win for Google’s next-generation AI hardware platforms. According to DIGITIMES Asia, this collaboration, confirmed on February 5, 2026, involves the integration of Innoscience’s 8-inch Gallium Nitride-on-Silicon (GaN-on-Si) power devices into the power delivery systems of Google’s high-performance AI servers. The deal marks the first time a major U.S. hyperscaler has deeply integrated 8-inch GaN technology from a Chinese-founded firm into its core AI infrastructure, highlighting the technical maturity of Innoscience’s high-volume manufacturing capabilities.
The partnership centers on the deployment of GaN power chips to manage the extreme power densities required by modern Tensor Processing Units (TPUs) and AI accelerators. As AI workloads grow exponentially, traditional silicon-based power management has reached its physical limits in terms of heat dissipation and efficiency. By utilizing Innoscience’s 8-inch GaN wafers, Google aims to reduce energy loss in power conversion by up to 20%, a critical metric for data centers currently facing unprecedented electricity costs and cooling challenges. This win is particularly significant as it validates Innoscience’s ability to meet the stringent reliability and volume requirements of the world’s largest cloud service providers.
The rise of Innoscience is a testament to the strategic transition of GaN from a niche material used in consumer fast chargers to a foundational component of the global AI economy. Founded by former NASA scientist Luo Weiwei, Innoscience has aggressively expanded its 8-inch GaN-on-Si capacity, reaching a milestone of 20,000 wafers per month by the end of 2025. This scale is pivotal; while competitors like Infineon and STMicroelectronics have historically dominated the power market, Innoscience’s early and singular focus on 8-inch GaN manufacturing has allowed it to achieve economies of scale that are difficult to replicate. The 8-inch format provides roughly 1.8 times more chips per wafer than the older 6-inch standard, significantly lowering the cost per die and making GaN price-competitive with high-end silicon MOSFETs.
From a broader industry perspective, the Google win underscores a "GaN-first" mentality emerging in the AI sector. As U.S. President Trump continues to emphasize American technological leadership and energy independence, the efficiency gains provided by GaN technology align with national interests in reducing the carbon footprint of the digital economy. However, the reliance on Innoscience also highlights a persistent dilemma for Western tech giants. Despite ongoing trade tensions and the "China Plus One" strategy, the sheer manufacturing clout and technical lead held by Innoscience in the 8-inch GaN space make them an indispensable partner for companies like Google that cannot afford to lag in the AI arms race.
The impact of this deal extends beyond Google. It serves as a market signal to other hyperscalers—including Amazon and Microsoft—that GaN is now ready for mission-critical data center applications. We are likely to see a surge in similar partnerships as the industry moves toward 48V power architectures, where GaN’s high-frequency switching capabilities are most advantageous. For Innoscience, the Google win provides the ultimate "stamp of approval," likely accelerating its plans for global expansion and potentially a public listing, as it now sits at the intersection of two of the most lucrative trends in technology: AI infrastructure and wide-bandgap semiconductors.
Looking ahead, the competition in the GaN space will intensify as traditional powerhouses ramp up their own 8-inch lines. However, Innoscience’s first-mover advantage in high-volume 8-inch production gives it a formidable moat. The primary risk remains geopolitical; as U.S. President Trump’s administration monitors semiconductor supply chains, Innoscience may face pressure to further localize manufacturing outside of China to maintain its standing with U.S. clients. Nevertheless, for the immediate future, the road to efficient AI power runs through the 8-inch GaN foundries of Innoscience, marking a new era where material science is as critical to AI as the algorithms themselves.
Explore more exclusive insights at nextfin.ai.
