Snowcap Compute's audacious plan: superconducting chips to power future AI

2025-06-25 general

San Francisco, Wednesday, 25 June 2025.
Snowcap Compute, a startup aiming to revolutionize AI processing, just secured $23 million in funding. Fueling this investment is their plan to create superconducting AI chips. These chips promise 25x performance per watt compared to today’s best. Pat Gelsinger, former Intel CEO, is betting big, joining Snowcap’s board. The company hopes to launch its first basic chip by the end of 2026. Could this be the end of the road for traditional chipmakers?

Superconducting Technology

Superconducting chips represent a potentially disruptive technology [2]. These chips use materials that allow electricity to flow with almost no resistance [2]. Over the years, scientists have explored superconducting chips, but keeping them at extremely low temperatures required significant power, offsetting the benefits [2]. Michael Lafferty, Snowcap’s CEO, believes that if performance gains are substantial enough, the power used for cooling becomes justifiable [2]. He notes that Snowcap’s chips aim for 25 times better performance per watt, even after considering cooling energy [2][1].

Market Implications for Chip Manufacturers

Snowcap’s entry arrives as AI demands surge and traditional chips face efficiency limits [2]. NVIDIA’s upcoming ‘Rubin Ultra’ AI data center server, expected in 2027, could consume 600 kW per hour [2]. This amount approximates two-thirds of the average American household’s monthly electricity usage [2]. Pat Gelsinger sees power limitations as a critical issue for data centers [1]. If Snowcap delivers on its promises, companies like NVIDIA may face pressure to adopt alternative architectures or risk losing market share. Investors should monitor these developments.

The talent behind snowcap

Snowcap’s team includes Anna and Quentin Herr, who bring experience in superconducting chips from Imed and Northrop Grumman [1]. The company also has former executives from NVIDIA and Google [1]. Advisors include Brian Kelleher (ex-NVIDIA), Phil Carmack (ex-Google), and Liam O’Conner (ex-Tesla) [3]. The leadership of Playground Global in the funding round, coupled with Gelsinger’s board membership, signals confidence in Snowcap’s vision [5]. This influx of expertise and capital could accelerate development and enhance investor confidence.

Challenges and Considerations

Despite the promise, Snowcap faces hurdles. The chips require niobium titanium nitride, sourced from Brazil and Canada [1]. Supply chain vulnerabilities could impact production [1]. The first basic chip is expected by the end of 2026, but complete systems will take longer [1]. MRAM, another emerging memory technology, also faces challenges in capacity and speed compared to existing solutions [7]. Investors must weigh these risks against the potential rewards of superconducting AI chips.

Bronnen

AI chips Superconducting computing