Vaire Computing Inc: Pioneering Reversible Computing for the AI Era
Vaire Computing Inc (and its subsidiary Vaire Computing Ltd) is a deep-tech semiconductor company specializing in reversible computing. Headquartered in London, United Kingdom, with additional offices in Seattle and Bellevue, Washington, the company was founded in 2021 by serial entrepreneur Rodolfo Rosini and Dr. Hannah Earley, a leading researcher in reversible computing from the University of Cambridge.
Vaire?s primary mission is to break the thermodynamic limits of traditional silicon chips. By utilizing a "reversible" architectural paradigm, the company develops hardware that recycles electrical energy instead of dissipating it as heat. This technology is specifically designed to meet the massive energy demands of Large Language Models (LLMs) and Generative AI, where thermal constraints and energy consumption have become the primary bottlenecks to performance scaling.
Continue…
Core Technology: Reversible and Adiabatic Computing
Vaire Computing operates at the intersection of information theory and thermodynamics. Traditional "irreversible" chips (CMOS) waste energy every time information is overwritten or erased?a physical reality known as the Landauer Limit. Vaire overcomes this through two primary technological pillars:
- Reversible Logic: In a Vaire circuit, every computational step is logically reversible. By ensuring that no information is lost during a calculation, the chip does not "discard" bits as heat. Instead, the input information is preserved and can be used to "uncompute" the state, allowing the energy used for the initial calculation to be recovered.
- Adiabatic Resonator Design: To physically recover and recycle energy, Vaire employs an adiabatic switching regime. Using custom-designed on-chip resonators, the energy that charges the gates of the transistors is captured at the end of a clock cycle and re-delivered for the next cycle. This process minimizes the potential difference during switching, nearly eliminating resistive heat loss.
Products and Services
Vaire Computing is a hardware-first company currently moving from the proof-of-concept phase to commercial-grade silicon. Its offerings include:
1. Near-Zero Energy AI Accelerators
Vaire is developing specialized chips tuned for AI/ML Inference. These chips are designed to provide a radical improvement in "performance-per-watt," enabling:
* Massive Throughput: Increasing token generation speeds for LLMs within existing power and cooling envelopes.
* Scalable Data Centers: Reducing the water and electricity requirements for cooling massive AI clusters.
* Always-On Edge Intelligence: Bringing sophisticated Agentic AI to mobile and edge devices that lack the battery capacity for traditional processors.
2. The "Ice River" Test Chip
In late 2025, Vaire demonstrated its first functional test silicon, codenamed Ice River.
* Energy Recovery: The chip showcased a world-first demonstration of energy recovery via an adiabatic resonator in 22nm planar CMOS.
* Logic Integration: The silicon integrates both the energy-recycling resonator and reversible logic structures (such as shift registers and adders).
* Standard Manufacturing: Importantly, Vaire produces its chips using standard CMOS fabrication processes, ensuring the technology is compatible with existing global foundry infrastructure.
3. Hardware-Embedded "Reversible-ization"
Recognizing that reversible programming is a paradigm shift for software engineers, Vaire embeds the complexity of reversibility directly into the hardware layer.
* Conventional Interfaces: Future commercial chips will feature standard interfaces, allowing them to accelerate existing code without requiring developers to write reversible algorithms from scratch.
* Universal Compatibility: While initially focused on AI accelerators, the company?s IP is designed to eventually support general-purpose computing (Arm/RISC-V) architectures.
Strategic Partnerships and Leadership
Vaire Computing is the only venture-backed startup globally focused on the commercialization of reversible computing. The company is supported by a team of industry veterans, including Andrew Sloss (a founding engineer at Arm) and Steve Pawlowski (former lead of the Reversible Computing team at Intel Labs).
- Intel Ignite: Vaire was selected as part of the Intel Ignite London cohort, gaining access to Intel?s deep technical expertise and foundry resources.
- Cadence Design Systems: The company utilizes the Cadence tool suite (including Virtuoso Studio RF and Clarity 3D Solver) to design the complex electromagnetic components required for their energy-recovery coils.
- Funding: As of 2026, the company has raised approximately $9.5 million in funding from investors including 7percent Ventures, Seedcamp, Eden Ventures, and Ginkgo Bioworks founder Tom Knight.
By decoupling computing power from energy consumption, Vaire Computing is positioning itself as the infrastructure backbone for a future where AI growth is no longer limited by the capacity of the power grid.
