Normal Computing Raises $50M from Samsung Catalyst to Tackle the AI Energy Crisis with Thermodynamic Chips (March 2026)

On March 25, 2026, Normal Computing announced a $50 million strategic funding round led by Samsung Catalyst, bringing the startup's total raised to over $85 million. The New York-based company is building two linked products aimed at solving what CEO Faris Sbahi calls "the AI energy crisis" — an AI-powered chip design platform already used by the world's largest semiconductor firms, and the CN101, the world's first thermodynamic computing chip.

What Happened

Normal Computing closed a $50 million round with Samsung Catalyst Fund as lead investor. New participants include Galvanize, Brevan Howard Macro Venture Fund, and ArcTern Ventures, alongside existing backers Celesta Capital, Drive Capital, Micron Ventures, and Eric Schmidt's First Spark Ventures. The company confirmed the round in a press release on March 25, 2026.

The funding comes seven months after a significant technical milestone: in August 2025, Normal completed the tape-out of CN101 — the world's first thermodynamic computing chip. Tape-out is the point at which a chip's design is finalized and sent to a fab for manufacturing, a major validation step. The CN101 targets multi-modal diffusion generative AI workloads and is the first implementation of Normal's Carnot architecture, which the company claims could achieve up to 1,000× gains in energy efficiency on targeted AI tasks compared to conventional GPU-based systems.

Key Details

• $50M raised, $85M+ total: Samsung Catalyst led the round; other investors include Galvanize, Brevan Howard Macro Venture Fund, ArcTern Ventures, Celesta Capital, Drive Capital, Micron Ventures, and Eric Schmidt's First Spark Ventures.
• Normal EDA platform: An AI-native electronic design automation (EDA) tool using "auto-formalization" — combining large language models with formal logic to accelerate chip verification. Already used by more than half of the top 10 semiconductor companies by revenue.
• CN101 thermodynamic chip: Tape-out completed August 2025. The CN101 harnesses the natural thermal dynamics of physical systems to compute more efficiently than deterministic GPUs, which expend large amounts of energy suppressing inherent physical randomness in transistors.
• Target workloads: Diffusion-based generative AI (image and video generation), where the stochastic nature of thermodynamic computation is a natural fit for diffusion model inference.
• 1,000× energy efficiency target: Normal's roadmap targets up to 1,000× lower energy consumption on targeted workloads versus H100-class GPUs, though the CN101 tape-out is the first step on that roadmap, not the final product.
• Team pedigree: Founded in 2022 by Faris Sbahi (CEO), Matthias Tan, and Antonio Martinez, with engineers and researchers from Google Brain, Google X, Meta, Palantir, Apple, Broadcom, NVIDIA, and national laboratories.

What Developers and Users Are Saying

The funding has drawn attention across the semiconductor and AI infrastructure community. Samsung Catalyst's involvement is particularly notable — Samsung is both a major chip manufacturer and a leading foundry, and a strategic investment of this kind signals that Samsung is closely watching thermodynamic computing approaches for future manufacturing partnerships.

Reaction from the broader AI hardware community has been guardedly optimistic. The CN101 tape-out in 2025 gave the company real credibility — claims of 1,000× efficiency gains from startups without silicon in hand are easy to dismiss, but Normal now has a physical chip to point to. However, technical discussions on hardware-focused forums note that demonstrating efficiency at the chip level is very different from delivering complete, production-grade systems. The path from tape-out to mass-market deployment typically takes several years and requires manufacturing scale that a $85M-funded startup doesn't yet have.

The UK's Advanced Research and Invention Agency (ARIA) is also a research partner, adding credibility from the academic and government research community.

What This Means for Developers

In the near term, developers building AI applications will not see direct access to Normal Computing hardware — the CN101 is still in early-stage validation, and volume production is not yet on a public timeline. However, the Normal EDA platform has near-term relevance: semiconductor companies using it to design their AI chips more efficiently could accelerate the timeline for next-generation GPU and ASIC availability across the industry.

For developers concerned about AI inference costs and energy sustainability — both increasingly important factors in evaluating cloud AI services — Normal Computing represents one of the most technically credible bets on fundamentally more efficient AI compute. If their thermodynamic approach scales as intended, it could eventually pressure GPU-based cloud providers to compete on energy cost per inference.

Keep an eye on Normal's CN101 product announcements: the next milestone after tape-out is silicon validation (testing the manufactured chip against the design). If CN101 validates at even a fraction of its target efficiency, the follow-up funding round and manufacturing partnerships will likely move quickly.

What's Next

Normal Computing has not announced a public timeline for CN101 commercial availability. CEO Faris Sbahi indicated in the funding announcement that the capital will be used to accelerate silicon design and scale the Normal EDA platform. Developers and researchers can follow progress at normalcomputing.com and watch for silicon validation results from the CN101 tape-out.

Sources

• Normal Computing Press Release (PRNewswire) — Official funding announcement with full investor list
• Fortune Exclusive — CEO Faris Sbahi quotes and company background
• SiliconANGLE — Technical analysis of the thermodynamic computing approach
• Normal Computing Blog — CN101 Tape-Out Announcement — Original August 2025 tape-out announcement
• Data Center Dynamics — Infrastructure perspective on thermodynamic compute
• Morningstar / PRNewswire — Full press release text with investor statements