By Hamza Faisal
Unconventional AI, the American hardware startup chasing far more energy-efficient AI chips, is buying Israeli chip startup NeuReality in a cash-and-stock deal, Reuters reported on Monday. The price was not disclosed. The acquisition adds a team that has already shipped finished AI chips to a company that, until now, has been primarily a research operation.
The deal marks a change of pace for Unconventional AI, which launched less than a year ago with a $4.5 billion valuation and raised $475 million in December. It is led by Naveen Rao, a serial chip entrepreneur who previously sold Nervana to Intel for more than $400 million and AI software startup MosaicML to Databricks for $1.3 billion. Its backers include venture firms Andreessen Horowitz and Sequoia, along with Amazon founder Jeff Bezos.
A different kind of computer
Mountain View, California-based Unconventional AI wants to build AI chips that are 1,000 times more energy efficient than current offerings, drawing on concepts from biological systems. Rao calls the approach a “dynamical systems” method: instead of following programmed, step-by-step logic the way conventional chips do, simple computing elements interact under simple rules and evolve over time to reach a result.
The idea matters because energy has become one of the binding constraints on AI. Training and running large models demands enormous amounts of power, and data-centre operators are racing to secure both chips and electricity. A chip that delivered comparable work for a fraction of the energy would change the economics of AI infrastructure, even if it took a different architectural path to get there.
AI tools sped up the roadmap
When Unconventional AI raised its $475 million war chest, Rao told Reuters, the company expected to spend three years on research and development and take five years to develop a product. Instead, by using AI tools to speed its own work, it has already developed its first test chip and is nearing completion of its second.
That acceleration is what prompted the acquisition. “Things have gone a lot faster than anticipated,” Rao told Reuters. “We need people who can go execute and build products and scale them. We built up the research bones and now we’ve got to build up the production bones.” In other words, the bottleneck has shifted from inventing the architecture to turning it into a manufacturable, shippable product.
What NeuReality brings
NeuReality, founded in 2019 by Moshe Tanach, formerly of Marvell Technology and Intel, together with Tzvika Shmueli and Yossi Kasus, had already developed several advanced AI chips before the deal. Its engineers bring experience in silicon design and data-centre systems, exactly the production discipline Unconventional AI says it now needs.
The agreement, announced on October 26, is expected to close in the fourth quarter of 2026, subject to customary conditions. For Rao, it is the latest in a run of bets that unconventional computing can break the power wall facing AI. Whether the dynamical-systems approach scales beyond test chips will now depend less on the idea itself, and more on the production team NeuReality brings with it.



