Seamless application of computing power

Infinigence AI, a rising star in China's AI infrastructure landscape, is revolutionizing access to computational resources under the leadership of its 32-year-old co-founder and CEO, Xia Lixue.

Founded just over two years ago in May 2023, the company champions a bold vision: making powerful AI computing resources as readily available and simple to use as turning on a water tap or flipping an electrical switch.

This mission comes at a pivotal moment, as Xia identifies AI as "China's biggest technological opportunity since the industrial revolution", positioning the nation at the center of a transformative global stage.

Xia's path to entrepreneurship is deeply rooted in academic excellence. Spending a decade at Tsinghua University, progressing from undergraduate studies to a doctorate degree, he focused on a critical challenge.

His research team addressed the fundamental communication barrier between diverse AI chipsets, each operating with distinct architectural "languages". Their breakthrough invention of a chip "Esperanto" effectively created an information superhighway, enabling seamless collaboration between heterogeneous processors. "To others, entrepreneurship might seem fraught with uncertainty," Xia reflected. "For me, it's the natural culmination of years of preparation."

This deep integration of expertise spanning algorithms, hardware and real-world application scenarios has fueled rapid progress at Infinigence AI. Within its first year, the company achieved a significant milestone by launching the industry's inaugural one thousand-card hybrid chip training platform. This innovation boasts an unprecedented computing utilization rate of 97.6 percent, maximizing the potential of domestic hardware.

Looking ahead, Xia sets an even more ambitious target: reducing chip operating energy consumption to merely one-tenth of current levels through innovative co-design of hardware and software.

Central to Xia's vision is the development of an agile middle-ware platform. This platform acts as a universal adapter, seamlessly connecting large, complex AI models with diverse underlying chips. It empowers model developers and application providers to effortlessly access and harness the combined computational power from industry leaders like Huawei, Biren and Cambricon, with a single, simplified command.

"Our fundamental aim is to make using computing power as convenient and ubiquitous as using water or electricity," Xia emphasized. "We're essentially building a 'Taobao'-like online market place for computing power resources — an online platform designed to help clients swiftly find the perfectly suited computational service while simultaneously enabling a vastly more efficient allocation and utilization of the nation's entire computing resource pool."

The company made a debut at the 2025 World Artificial Intelligence Conference in Shanghai in July. The centerpiece of its showcase was a stack of three distinct "boxes", representing the launch of its latest solution suite, which provides comprehensive, full-scenario support, encompassing everything from sophisticated model scheduling and performance optimization to seamless application deployment, all unified across diverse and complex computing architectures.

Xia explained the strategic imperative. "Building large-scale computing infrastructure resembles constructing a 'giant power plant', but the actual needs of AI applications demand readily available 'plug-and-play sockets'.Current AI computing infrastructure must fundamentally adapt to effectively serve the multitude of small, widespread AI application enterprises that require agile, on-demand production resources."

Li Wei, deputy head of the cloud computing and big data research institute at the China Academy of Information and Communications Technology, said that traditionally if companies want to use computing power, they either build the infrastructure themselves or rent the computing power via cloud subscriptions, which is expensive. But now with platforms such as Infinigence AI, they can easily find and use computing power resources that are scattered across China in a more efficient way.

Wu Hequan, an academician at the Chinese Academy of Engineering, said that every yuan ($0.14) invested in computing power drives 3 to 4 yuan in GDP growth.

"In the global race for AI leadership, expanding computing power supply is critical," Wu added.

Xia passionately highlighted the vital role of young talent and China's unique market advantage in driving innovation. He recalled a pivotal early breakthrough significantly boosted by a 24-year-old intern, whose insight tripled algorithm efficiency on hardware. The company fosters this spirit through weekly open tech forums, where young engineers freely exchange ideas, sparking continuous innovation.

Xia views China's position as the world's largest AI application market as an unparalleled catalyst for growth. "This vast potential naturally energizes young practitioners and injects genuine, lasting momentum into the entire industry," he said.