For as long as there have been computer chips, the heat they throw off has been treated as the price of computing. Earley's position is that it is a consequence of engineering choices, not a property of the physics. Her company, Vaire Computing, builds chips that recover the energy normally lost as heat — an approach known as reversible computing — and last year it reported the result the field had been waiting on for decades: a chip with Vaire's resonator returned more energy than it lost, even after accounting for the energy needed to run the resonator itself.
Source: technologyreview.com
The mechanism starts with erasure. A conventional chip discards information it no longer needs as it computes, and dissipates energy as heat in the process. Earley's analogy is driving across a city and braking at every intersection: the car sheds its speed, then burns fuel again to get it back. Reversible computing keeps the momentum. The circuit holds on to the data from intermediate steps, which means the computation can be run backwards and some of the energy recovered.
The idea is more than 50 years old. What was missing was hardware — existing transistors and circuits could not do it in practice. Earley redesigned that layer herself, and built a resonator: a microscopic on-chip component that stores the recovered energy for reuse. She describes it as, in essence, an elaborate pendulum.
Igor Markov, a researcher in electronic design automation and a former professor at the University of Michigan in Ann Arbor, calls it an interesting piece of work while placing it where it belongs on the maturity curve. The technology is early. To win support from the industry and get to commercialization, he says, Vaire will need a series of increasingly realistic and convincing demonstrations.
Earley arrived at chip design earlier than most people arrive at anything. She started programming at around nine, beginning with high-level web programming and moving through other languages, including Perl and Java, working across levels of abstraction until she reached the transistors. She went on to a doctorate at the University of Cambridge under the computational biologist Gos Micklem, initially studying how materials such as DNA could be used to compute. A few months in, Micklem sent her Michael Frank's 1999 dissertation, one of the first substantial works on reversible computing.
She read it, was skeptical, read it again, and spent several weeks turning it over. Gradually she became convinced that the relationship between information, energy and heat could change computers permanently. It redirected her thesis entirely: she studied the physical limits of computation and wrote a program that converts ordinary programs into reversible ones. At some point Micklem stopped putting his name on her papers, on the grounds that he could not have delivered a full talk on them by himself — the work was hers.
She defended in 2021, met the technology entrepreneur and investor Rodolfo Rosini the same year, and the two founded Vaire. Since then the company has raised more than $12 million, hired Michael Frank — the author of the dissertation that started it — as a senior scientist, and begun turning the theory into physical hardware.
The work itself has been closer to grinding than to inspiration. In the winter of 2022, Earley spent several weeks in the basement apartment of her now-wife in Grinnell, Iowa, with the wind chill outside running to about −40 °F, filling and refilling a whiteboard with designs for the key section of circuitry that reversible logic depends on. When it finally resolved — by then the couple had escaped the cold for Las Vegas — it registered as gradual relief rather than a flash of insight. What she remembers feeling is that she did understand what was going on, and had not taken on something beyond her.
My read: the physics result and the business are further apart than the announcement makes them sound. Net-positive energy recovery on a test chip is a genuine milestone for a field that spent 50 years on paper, and it is also exactly the kind of milestone that arrives long before anyone can buy anything. The company is five years old with $12 million raised — a rounding error against what it costs to put a new logic style into a fab — and Markov's phrasing is the tell. He does not doubt the result. He wants a series of demonstrations, each more realistic than the last, which is another way of saying that one clean lab number persuades nobody who has to commit a production line.
The numbers nobody has supplied are the ones a buyer would ask for first. How much energy the chip actually returns, on what workload, at what clock speed, on what process node, and what a system built from these parts does per watt against a conventional one. Recovering more energy than you lose is a threshold. Beating an existing chip is a benchmark, and those are different claims.
Earley's next problem contains its own contradiction. She wants to fit a fundamentally different chip into familiar devices and existing manufacturing systems — and she also believes the future is a full redesign of computers around reversibility rather than further improvement of the chips we have, with every element of the machine rethought from that angle. Those two ambitions pull in opposite directions, and the industry has historically funded only the first.