Vaire Computing is building Near-Zero Energy Chips through reversible computing to unlock the future of computing.
Why are you dedicating your time to Vaire and reversible computing right now?
In 2018, I began thinking about the future of AI and realized its growth depends heavily on computing power. Historically, companies like Intel and NVIDIA have driven the field forward by developing more powerful chips. However, I predicted that by 2028, computing power would hit a limit, just as AI would demand even more processing capabilities. This intersection of demand and hardware stagnation led me to seek solutions. Knowing that software alone is not going to solve this with the breakthrough needing to come from hardware, I looked more into reversible computing - a concept explored decades ago. As a result In 2021, I founded Vaire, assembling a team that includes one of ARM’s founding engineers and other world-class experts. I strongly believe this could revolutionize computing in the coming years.
What exactly is reversible computing, and are we interacting with it today?
Reversible computing is rooted in efficiency, much like the industrial revolution's focus on using energy more effectively. Mechanically, reversibility means reusing energy instead of wasting it as heat, like how an internal combustion engine operates without coming to a full stop. The same principle applies in computing, though traditional transistors weren't originally designed for computation. Current computing systems waste energy as heat during processing. Our approach is to recycle that energy within the chip, preventing it from being lost. This architecture allows us to reuse the charge for subsequent computations, offering immense improvements in efficiency. By redesigning just 1% of the system and keeping the rest intact, we can make the transition to reversible computing commercially viable with potentially 4,000x more efficiency than today's AI chips.
Moore's Law is mentioned frequently in discussions around semiconductors. Your company has said that it will be obsolete soon. Could you explain that?
Moore’s Law originally referred to the doubling of transistors on a chip, but over the past 20 years, that pace has slowed. Companies like Intel now redefine it to include other metrics, such as performance in data centers. At this point, Moore’s Law is becoming increasingly irrelevant because the industry struggles to increase computing power without drastically raising energy and water consumption. The exponential growth in AI and computing demands will soon outstrip current capabilities. Today, a single data center can use as much energy as 700,000 homes, which is unsustainable. Our aim is to decouple computing growth from resource consumption, and reversible computing could be the key. It allows for exponential growth in computing power without the corresponding rise in energy use.
Do you think reversible computing will become the only viable option in the next five years, or will there be competing technologies?
In the long term, we believe reversible computing is the only viable solution. Some companies are already adopting similar concepts, like quantum computing, but the broader adoption of reversible computing will dominate over the next few decades. Once classical computing reaches its limits, the industry will have to shift toward reversibility to maintain growth. Our focus on near-zero energy chips represents a specific application of reversible computing. The challenge has been how to transition commercially without overhauling everything. Our breakthrough was figuring out how to rebuild the foundation without needing to redesign the entire system, making it feasible for widespread adoption.Why is your company based in London, as opposed to more traditional tech hubs like the Bay Area?
Our R&D is not based in London; it is in Cambridge (UK), the Bay Area, and Seattle. London serves as the hub for our finance, commercial operations, and administration. Cambridge was an ideal choice for R&D due to its deep tech expertise and Dr Hannah Earley was based there. We also maintain offices in Seattle and Sunnyvale for their proximity to key talent and industry partners. Interestingly, our Cambridge office once housed Sinclair Research, the creators of the ZX Spectrum. While we initially chose it for its affordability, we later discovered its historical significance in computing innovation. Being in these key locations allows us access to the talent and resources needed for advancing our work.
Are you collaborating with hyperscalers or other companies at this stage, or is it more focused on R&D?
While we are still in a research-intensive phase, hyperscalers have been instrumental to our progress. Early on, we had discussions with several major hyperscalers, and they showed significant interest in our technology. These companies, which require vast numbers of chips, recognize the potential impact of reversible computing and have been supportive in our journey.
Although I can’t go into too much detail, these partnerships have been critical in confirming that there’s a market for our chips. Unlike quantum computing, which may take years and billions of dollars to commercialize, our technology is on a faster track to market, and the hyperscalers share this vision.
You must have an exceptional team to tackle such a significant challenge. How do you see the competition in your space?
One of the benefits of starting a chip company is that competition is relatively limited compared to sectors like B2B SaaS or AI. In those fields, there are thousands of companies vying for talent and market share, but in the AI chip space, there are only about 50 companies, with NVIDIA being the dominant player. We have been fortunate to hire world-class experts because, at the early stages of a technological shift, the demand for specialized talent is lower. For example, in 2010, few companies were hiring AI engineers, and in 1997, cybersecurity experts weren’t in high demand. By entering early in a field like reversible computing, we can attract top talent and train the next generation of innovators before the larger players come into the picture.
What is your personal thesis on staying ahead of the curve in technology?
My approach has always been about anticipating the next big shift. In 1997, it was cybersecurity; in 2010, it was AI. Now, I see reversible computing as the next frontier. Traditional chip design discharges energy, limiting the potential for 3D chip development. But with reversible computing, we could stack chips into cubic structures without overheating. Memory chips have already made advances in this direction by stacking thousands of layers. We want to achieve the same with compute chips, turning a massive data center into something you could hold in your hand. Today, such a chip would reach temperatures over 4,000 degrees Celsius, but with reversible computing, we believe this is a problem we can solve.
