NORMAL COMPUTING RAISES $50M LED BY SAMSUNG CATALYST TO ACCELERATE SILICON DESIGN AND SOLVE AI HARDWARE ENERGY CRISIS

Funding advances AI-native semiconductor design and physics-based computing architectures targeting dramatically greater intelligence per dollar and per watt

NEW YORK, March 25, 2026 /PRNewswire/ -- Normal Computing today announced $50 million in strategic funding led by Samsung Catalyst Fund, bringing total funding to more than $85 million.

New investors also include Galvanize, Brevan Howard Macro Venture Fund, and ArcTern Ventures, alongside existing investors Celesta Capital, Drive Capital, Eric Schmidt's First Spark Ventures, and Micron Ventures.

Normal was founded to tackle two converging crises: the escalating complexity of silicon design and the energy requirements of AI infrastructure. The company builds AI for the semiconductor industry and is developing a new class of computing hardware, using its software to design its own hardware IP.

"Datacenters are expected to hit an energy wall around 2030. Finding more energy is not a durable strategy. Meeting growing 'intelligence per dollar per Watt' demands a fundamentally novel architecture, and a unification in design methodology. Normal EDA exists to accelerate custom silicon to market by 2x today, and, over time, to enable 1000x gains in efficiency with our platform," said Faris Sbahi, CEO of Normal Computing.

Today, Normal is partnered with more than half of the top ten by revenue semiconductor companies and has differentiated itself by earning the industry's trust to drive real production silicon. Normal EDA, the company's purpose-built AI platform for semiconductors, aims to both radically accelerate chip design and also realize architectures that could drastically shift AI scaling economics.

"Our industry is facing unprecedented transformation propelled by AI, and the complexity of leading-edge semiconductor chip design has increased dramatically," said Dede Goldschmidt, SVP & Managing Director, Head of Samsung Catalyst Fund. "Normal has assembled a strong team with broad experience across AI and semiconductors. The platform they have built has the potential to offer faster time-to-market for the most demanding chip design customers in the ecosystem. We are very excited to support them and lead this financing round."

AI to Scale Custom Silicon

AI has begun to transform the various disciplines of knowledge work, including software engineering, data analysis, and graphic design. Meanwhile, hardened engineering tools in sectors like the semiconductor industry have undergone minimal changes, stretching over recent decades. Today, the industry faces the intense pressure from talent shortages, historically low first-silicon success rates and high respin costs, and the need to meet the energy and performance requirements of the AI scaling laws that its own chips power. Normal promises to close the gap, compressing silicon timelines from years to months without sacrificing the highly collaborative, zero-tolerance-for-failure methodology that hardware requires.

Normal EDA applies auto-formalization, a frontier area of AI research which combines LLMs with formal logic, to enable AI and human engineers to solve the hardest problems in silicon design. The AI platform builds an understanding of a team's intent and goals by integrating with their specs and workflow, and uses that context to design, optimize, and prove the correctness of the silicon while continually learning. Normal spans DV/RTL automation today as well as more frontier end-to-end and unconventional silicon design workflows, which will self-improve alongside software and kernels in the future. Normal is an active supporter of open EDA standards and industry-wide benchmarking, and serves as a founding member of the Silicon Integration Initiative (Si2) LLM Benchmarking Coalition.

Physics-based ASICs

In parallel, Normal is advancing its Carnot hardware program, using Normal EDA to develop its own silicon IP. In August 2025, Normal completed the tape-out of CN101 - the world's first thermodynamic computing chip, targeting multi-modal diffusion GenAI model inference. This is the first step on a roadmap targeting up to 1000x gains in energy efficiency and dramatically lower latency.

Where conventional GPUs expend energy by suppressing the inherent randomness of physical systems, Normal's physics-based ASICs work with those dynamics to compute more efficiently, unlocking a fundamentally new architecture for AI workloads like image and video generation.

The work is funded in part through the Advanced Research + Invention Agency (ARIA), as part of its Scaling Compute Programme.

"Most of the advanced computing landscape is oriented toward the next incremental improvement. ARIA exists to fund the ideas where the potential impact is not marginal but transformational, even when the technical risk is high. Normal's team has taken a fundamentally unconventional approach and delivered working silicon in CN101. That is an exceptionally rare outcome for work this ambitious and we are excited to witness this next phase of the journey," said Dr. Suraj Bramhavar, Programme Director of ARIA's Scaling Compute programme.

About Normal Computing

Normal builds mathematical and physical intelligence to accelerate hardware innovation with the world's leading semiconductor companies. Founded in 2022, the company is pioneering advances in AI methodology and physics-based computing to unlock a step change in how chips are designed and built. The team includes leading researchers and engineers from Meta, Palantir, Graphcore, Apple, Broadcom, NVIDIA, Google Brain, Google X, and national laboratories, alongside exited founders and medalists spanning AI, chip design, physics, and functional programming. Normal is headquartered in New York, with offices in San Francisco, London, and Copenhagen, and future expansion in Korea.

Learn more at normalcomputing.com

Media contact: [email protected]

We are hiring: careers.normalcomputing.com

SOURCE Normal Computing