Over 100 million in funding.
A new player has emerged in AI chips!
News from June 24th: On June 23rd, US AI chip startup Snowcap Compute made its first public announcement, revealing it had secured $23 million (approximately 165 million RMB) in seed funding led by Playground Global. It also announced that former Intel CEO Pat Gelsinger has joined its board of directors.
Snowcap plans to develop a new commercial AI computing chip using superconductors, aiming to create computers that can surpass today’s most advanced AI systems while consuming minimal power.
According to foreign media reports, Snowcap plans to launch its first foundational chip by the end of 2026, but the complete system will be released later.
According to the press release, Snowcap is a startup building the first commercially viable superconducting computing platform. Its chip architecture is designed for extreme performance and energy efficiency, enabling new data centers optimized for AI, quantum, and high-performance computing, providing the performance and efficiency required to support advanced AI inference and training, high-performance computing, and quantum-classical hybrid workloads.
The company’s founding marks the first time superconducting technology will compete commercially with CMOS.
- Former Intel, Google, Tesla, Nvidia Executives Join
Snowcap CEO Michael Lafferty previously served as the head of the Beyond Moore Engineering group at US EDA giant Cadence, where he pioneered superconducting and quantum technologies. Its founding team also includes Chief Scientist Anna Herr and CTO Quentin Herr, who are leading researchers in practical superconducting computers.
Several seasoned Silicon Valley tech veterans have joined the startup. Brian Kelleher, former Senior Vice President of GPU Engineering at Nvidia; Phil Carmack, former Vice President of Chip Engineering at Google; and Liam O’Conner, former Vice President of Global Supply Management and Supplier Industrialization Engineering at Tesla, are all advisors to Snowcap.
Former Intel CEO Pat Gelsinger and former Nvidia Vice President of Business Development Rick Hyman have both joined its board of directors.
Lafferty revealed that Snowcap is significantly improving performance while reducing power consumption. Even accounting for the energy consumed by cooling, its chips will offer approximately 25 times the performance per watt of today’s best chips.
“Superconducting logic allows us to go beyond the limitations of existing CMOS technology, achieving orders of magnitude improvements in processing speed and efficiency,” Lafferty said. “This performance is crucial for the future of AI and quantum computing.”
Gelsinger posted on LinkedIn: “Excited to share the launch of Snowcap Compute Inc., a company building the first commercially viable superconducting compute platform, backed by a $23M seed round led by Playground Global. This is my first public investment as a Playground General Partner, and I couldn’t be more thrilled to back a team redefining the frontier of computing performance—classic, AI, and quantum—the trifecta of computing will all benefit from superconductivity and Snowcap.”
- Replacing Transistors with Josephson Junctions, Much Lower Energy Consumption Than Traditional Chips
According to Gelsinger’s sharing, scientists and engineers have been researching superconductivity for decades, but only now has the physics matured and manufacturing become feasible, with recent advancements making commercialization possible.
“Snowcap is laying the groundwork for a post-CMOS era, where significant improvements in performance and power efficiency will be required for AI, high-performance computing, and quantum-classical hybrid workloads. Its platform architecture offers orders of magnitude improvements in processing speed and efficiency, achieved through decades of R&D and some of the industry’s most experienced talent,” Gelsinger wrote. “This is deep tech at its best, solving bottlenecks, opening new energy frontiers, and pushing the boundaries of what’s possible with silicon.”
Snowcap’s official website states that superconducting computing involves cooling silicon chip materials to superconducting temperatures, resulting in zero resistance in wires and greatly reduced switching energy consumption for gates.
Both quantum computers and the Snowcap platform utilize superconducting manufacturing and cooling technologies. The difference is that Snowcap is building a platform suitable for standard digital chip designs, such as traditional CPUs, GPUs, or AI inference chips.
According to its official website, Snowcap’s AI chips replace transistors with Josephson junctions, whose switching gates consume 5 orders of magnitude less energy than today’s transistors.
A Josephson junction, also known as a superconducting tunnel junction, is a tiny quantum device where superconducting electrons can tunnel from one side through a thin film of semiconductor or insulator to the other. When cooled to 4.5 Kelvin using standard cryogenic systems, these circuits can switch in picoseconds, consuming astonishingly little energy per operation—only one hundred thousandth of that of CMOS.
At the same time, Snowcap’s chips can be manufactured using existing 300mm (12-inch) semiconductor processes, without the need for special manufacturing processes.
- Conclusion: Overcoming Multiple Key Engineering Challenges, Using Superconducting Technology to Alleviate AI Computing Energy Consumption Pain Points
According to foreign media reports, scientists have been theorizing about using superconductors to build computer chips since at least the 1990s, but they faced a major challenge: the chips need to be kept at extremely low temperatures in a cryogenic cooler, which itself consumes a lot of power.
For decades, superconducting chips did not develop until AI chatbots generated a huge demand for computing power, while at the same time, traditional chip performance was nearing its limits, and massive computing power consumption burdened power infrastructure.
Snowcap’s innovative approach aims to use superconducting materials to deliver computing power with higher energy efficiency, thereby reducing the energy consumption burden of next-generation computing systems. The startup has also addressed key engineering challenges that previously hindered the widespread commercialization of superconducting technology, including scalability, wafer fab compatibility, EDA challenges, and system architecture.
This injects new blood into the gradually stable AI chip industry.