Insider Brief
- Extropic signed a non-binding letter of intent with the U.S. Department of Commerce for up to $75 million in CHIPS R&D funding to accelerate development and domestic manufacturing of its thermodynamic computing hardware for AI.
- The funding would support development of Extropic’s Thermodynamic Sampling Units (TSUs), probabilistic semiconductor chips designed to improve the energy efficiency of generative AI, robotics, simulations, and other workloads by leveraging thermodynamic fluctuations in CMOS transistors.
- The proposed project includes scaling production of Extropic’s Z1 processor, developing rack-scale thermodynamic computing systems, and establishing U.S.-based manufacturing for a future Z1.5 chip using mature semiconductor process nodes.
PRESS RELEASE — Extropic, the American company pioneering thermodynamic computing hardware, today announced the signing of a letter of intent with the U.S. Department of Commerce for up to $75 million in funding through the CHIPS Research and Development Office (CRDO). The planned funding is expected to accelerate the development of Thermodynamic Sampling Units (TSUs), an entirely new category of semiconductor purpose-built for probabilistic AI workloads, and qualify a domestic manufacturing path for the technology.
TSUs use the natural thermodynamic fluctuations of standard CMOS transistors to sample directly from programmable probability distributions in a power-efficient manner, a capability useful for generative AI, simulations of biology, markets, and beyond. The same capability extends to probabilistic inference for physical intelligence and defense, where robots, autonomous systems, and sensor platforms must reason under uncertainty in real time on tight power budgets. On key generative-AI workloads, TSUs can achieve energy-efficiency gains of orders of magnitude over conventional GPUs.
“The current era of AI is one of exponential acceleration, and energy is its binding constraint. Brute-forcing the digital, deterministic paradigm to unfathomable scale cannot be the endgame,” said Guillaume Verdon, founder and CEO of Extropic. “TSUs harness the inherent randomness of nature to deliver more intelligence per watt. And because they run on mature nodes, we can build them here, in American fabs, without waiting in line for anyone else’s capacity. We are proud to partner with the Trump Administration, Secretary Lutnick, and Executive Director Frauenhofer to manufacture this new category of compute on American soil and help return the United States to compute dominance.”
Replaying the playbook that built Silicon Valley
Silicon Valley was built on public-private partnerships: the first integrated circuits found their earliest volume customers in federal programs, and the government acting as first customer and first investor is how the American semiconductor industry was built up. Six decades later, Extropic is reinventing how the transistor is used for the generative AI era, moving beyond deterministic digital logic into stochastic electronics and thermodynamic computation. Once again the federal government is acting as a key partner to bridge the gap from working prototype to production scale and accelerating the scaling and diffusion of this technology into the market. Thermodynamic computing will fundamentally change how we compute from the ground up, enabling new capabilities, unparalleled energy efficiency and greater access to AI at lower costs for all Americans.
Scaling American AI output
AI’s overall growth, accessibility, and costs are running up against the limits of the power grid: both power production and energy efficiency have become limiting factors for American compute. In the age of AI, a nation’s strength will be measured by its total intelligence production: the watts it generates, multiplied by the intelligence it extracts from every watt.
While America secures the watts through a new American nuclear renaissance, spurred by the President’s May 2025 executive orders on nuclear energy, power efficiency will determine whether artificial intelligence is ubiquitously and cost-efficiently accessible at scale. This investment in thermodynamic computing will push the frontier of how much intelligence per watt one can get from American silicon, unlocking a path to potentially orders of magnitude greater total intelligence output at a national scale. More intelligence output means more discovery, more industry, and more abundance, compounding directly into American economic strength and the prosperity of every American.
Building the first large-scale thermodynamic computing system
For Extropic, the bridge from prototype to production runs from the X0, the prototype silicon chip that took thermodynamic computing from zero to one, to the Z1, its first production-scale TSU. Under the letter of intent, the planned R&D funding would support bringing the first Z1 clusters online and demonstrating chip performance on generative-AI benchmarks. Much of the frontier of this work lies beyond the chip itself, in the systems around it: the infrastructure to control TSUs, connect them to one another, and integrate them with conventional forms of computing at rack scale.
The systems scaling work supported by this award will allow for Z1 clusters large enough to begin exploring thermodynamic AI algorithms at scales impractical to simulate on GPUs today. This will in turn enable the discovery of novel algorithms native to this hardware paradigm. Accelerating that algorithmic development is key to migrating generative AI workloads from traditional accelerators to the more power-efficient substrate that is the TSU.
Bringing the thermodynamic computing supply chain home
Extropic’s TSU chips are unique in that they extract more computation from orders of magnitude fewer transistors by operating them probabilistically, achieving dramatic power efficiency gains without requiring the smallest transistors from the cutting edge process nodes. As a result, TSUs can be manufactured on mature process nodes with substantial capacity available in the United States today, without competing for the predominantly offshore leading-edge wafers on which the world’s supply of digital accelerators depends.
The capstone of the project is a new thermodynamic chip: Z1.5, an iteration over Z1, but now fabricated at a U.S. foundry, establishing onshore production capability for thermodynamic computing for the first time. Extropic has already demonstrated the fab-portability of its probabilistic primitive IP, with two X0 variants taped out and successfully tested at two different fabs.
“With today’s compute supply chain investments, the Trump Administration is accelerating America’s innovation engine,” said Secretary of Commerce Howard Lutnick. “These strategic investments will enhance our country’s domestic capabilities, create high-paying jobs and keep America at the forefront of the semiconductor industry.”
Stochastic electronics represent the future of computing for AI, and they have different fabrication requirements than digital. This award will ensure U.S. foundries create the manufacturing know-how to begin optimizing the American supply chain for the next generation of computing. Onshoring that supply chain is essential to American sovereignty and national security as computing transitions to its next substrate.
“The CHIPS R&D incentives will support a breakthrough in compute architecture to power next-generation AI and scientific discoveries,” said Bill Frauenhofer, the Executive Director for Semiconductor Innovation and Investment at the Department of Commerce. “Accelerating R&D for probabilistic compute provides American industry the energy efficiency to scale complex AI workloads securely and rapidly.”
The company will share a comprehensive technical update on its full thermodynamic computing stack, spanning hardware, software, and API access, in the coming days.
The letter of intent is non-binding, and the planned funding remains subject to the negotiation of definitive agreements with the Department of Commerce and the achievement of project milestones.
