Unconventional AI

Unconventional AI is rethinking the foundations of a computer to bring biology-scale efficiency to artificial intelligence.

[Unconventional AI @unconvAI]

Introducing our first model, Un-0!

We trained an image generator powered by a backbone of coupled oscillators in place of a more traditional conventional neural network.

[Unconventional AI @unconvAI]

AI efficiency depends on how information moves through a machine.

Our CEO @NaveenGRao spoke with @furrier at CASPA about rethinking the separation of memory and compute, and using physics directly for computation.

https://t.co/dpYPwDgNkf

[Unconventional AI @unconvAI]

“We’re energy first in our mindset.”

On @twentyminutevc, our CEO @NaveenGRao explains what that means: everything we do has to give us an energy advantage.

Watch his conversation with @HarryStebbings:
https://t.co/UFXApFc2vs

[Unconventional AI @unconvAI]

Unconventional AI has reached a definitive agreement to acquire @NeuReality_AI , whose team and technology will play a central role in advancing our mission to make AI compute radically more efficient.

Unconventional is building a 4-dimensional substrate targeting a 1,000x improvement in AI compute efficiency over conventional architectures. Reaching that target requires co-design across device physics, silicon architecture, and the models themselves. @NeuReality_AI brings seven years of work on purpose-built AI inference silicon and infrastructure, plus an engineering team spanning system architecture, silicon design, and AI systems software. Together that materially compresses our timeline to production systems, and it establishes our engineering presence in Israel.

We're excited to welcome @MosheTanach , @tzshmueli , @YossiKasus , @LiorKhermosh , @iritfast83 , and the NeuReality team to Unconventional. Together we'll build a next substrate for intelligence, and the systems to put it into production.

Learn more: https://t.co/PottTJDMAF

[Unconventional AI @unconvAI]

"I get a high off creating a new way of doing something that changes how that category works."

Learn more about our cofounder and CEO @naveengrao in @Microsoft Signal magazine: https://t.co/k20MQ2YO34

[Unconventional AI @unconvAI]

Our cofounder and CEO @NaveenGRao, with @KristenScholer on @NYSE Floor Talk at the @theallinpod Summit:

The human brain has about 16 billion neurons in the cortex, and only about 16 billion bits are moved every second in about a centimeter. An AI chip with HBM is moving on the order of 30 trillion bits every second over about the same distance, and that's only traffic on and off memory.

"A lot of ink has been spilled about this whole memory wall. We're actually circumventing the problem entirely by rethinking it."

https://t.co/RXWsb9lVP3

[Unconventional AI @unconvAI]

In January, we gave ourselves 2 to 3 years to figure out the science behind building a new compute substrate to cut the power consumption of AI systems by 1,000x. We got there in 9 months, by answering these three questions:

Can you compute with dynamics?
Yes. The brain doesn't have floating point operations or instruction sets. It forms neural pathways when you learn and traverses them when you perform a task. Our systems work the same way: training sets the weights, a token perturbs the system, and the final state is the answer.

Is it power efficient?
Yes. Our first chip taped out on June 1st and is back in the lab. The power numbers are orders of magnitude below traditional inference systems.

Does it scale?
Yes, and this was the big one. A fully connected system needs trillions of wires, and nobody can physically build that. We found that cutting connections to roughly 2 to 3% improves output quality and learning and allows you to scale.

Watch @theCUBE where our CFO @aesfahani and @furrier discuss the research that brought us to the tape-out: https://t.co/2UhDbzYRCy

[Unconventional AI @unconvAI]

Full episode, 1 hour 9 minutes: https://t.co/DCGyZP07tP

[Unconventional AI @unconvAI]

"Computing is not some construct that humans came up with. Computing is everywhere, it's all around us. Our bodies are built out of computation."

Cofounder and CEO @NaveenGRao on @thecompute100

[Unconventional AI @unconvAI]

On getting a chip designed in six months without a dedicated team: "That's not a story about needing fewer engineers. It's a story about how many more things you can try."

Our growth is limited by how fast we can hire. We're hiring: https://t.co/npFbI3SgQu

[Unconventional AI @unconvAI]

"Every technology disruption, every one over history, has created more jobs than it destroys."

@NaveenGRao and @PGelsinger in @FortuneMagazine on why the AI transition will be no different, and why the real constraint is people.

https://t.co/dynNQMFm5V

[Unconventional AI @unconvAI]

More than 11,000 Americans turn 65 every day. The AI buildout is short ~350,000 construction workers this year. Electrician wages are rising 2-4x faster than wages overall. The constraint on data center expansion isn't chips or capital, it's people who can wire a building.

[Unconventional AI @unconvAI]

At the @theallinpod Summit, our cofounder and CEO @NaveenGRao shared new details on our first test chip tape out:

"This is the first physical dynamical computer ever built. We did it in five months."

@chamath joins him to discuss the path to product.

https://t.co/BiwQLXnADJ

[Naveen Rao @NaveenGRao]

I made some public disclosures about our hardware at the @theallinpod conf this week and wanted to share the progress here. Earlier this summer we got hardware back! The execution was madness…we went from no team in Jan to a tape out in 5 months. AI enabled MUCH tighter loops of research and our execution speed shows the results.

This is the first large-scale demonstration of a causal, physical dynamical system to do real compute. 🚀

We released the Un-0 model (see link) in June and it runs on this physical system; the images below come from the actual hardware. What’s more, this chip requires <900 nJ to generate an image; this is many orders of magnitude less energy than conventional machines.

https://t.co/zYU0ezXJUq

[Unconventional AI @unconvAI]

Unconventional AI was founded on a suspicion: that more parameters and more attention is one path among several, and that the alternatives stay underexplored not because they are less promising but because they are harder to fund.

The Unconventional Grant exists to fund those alternatives. We said plainly that we would rather back a well-posed risk than a safe increment.

Today we are announcing our first cohort. Five projects, each receiving $100,000:

- Thierry Tambe (Stanford) is treating data retention as an architected resource, so processors stop spending area and energy preserving information long after the computation stops needing it.

- William Gilpin (UT Austin) is testing whether the slowdown in reasoning models is transient chaos, the same phenomenon that governs turbulent flows, and looking for the fractal traps in latent space that delay convergence.

- Logan Wright (Yale) is asking what neural scaling laws become when the neural network is a physical analog machine rather than a digital simulation of one, noise, drift, and energy budgets included.

- Yoon Kim (MIT) is pushing looped Transformers past the tens of loops where gains have appeared to saturate, into the hundreds and thousands, with primitives designed to keep parameters on-chip.

- Wanyu Lei, Daniel Kunin, and Friedrich T. Sommer (UC Berkeley) are deriving recurrent architectures directly from the mathematics of state updates, on the premise that efficient state tracking is not one capability among many but a foundational design principle.

To the five teams: congratulations. Each team describes its own project, in its own words, in the full announcement. https://t.co/rb3BU3KpIG

[Unconventional AI @unconvAI]

We removed 93.25% of the connections in our Un-0 image model, fully expecting to pay for it in quality. But it got better.

FID 7.15 on ImageNet 64x64, roughly 1.9 ahead of the dense baseline at matched size. Same family of model, a fraction of the couplings, a better score.
Here is why that is not as strange as it sounds. Un-0 is a coupled oscillator model, and in the dense version every oscillator talks to every other one. That sounds like a strength, but it means the whole system can fall into catastrophic synchronization: everything locks into step, gradients go flat, and learning stalls.

Sparser connectivity leaves room for coherent and incoherent activity to coexist. The dynamics stay alive, and the model keeps learning.
Connectivity turns out to be a control knob, not a dial you turn up until it stops.

Learn more here: https://t.co/N4lIri8UQQ

[Unconventional AI @unconvAI]

Building a new memory means a twenty-year grind to drive the error rate toward zero. AI just made that grind optional.

Because neural networks are remarkably tolerant of noise, we no longer have to chase punishing reliability targets. Training can settle into a good solution even when the hardware underneath is imperfect (which is exactly what the loss landscape below is showing).
That single shift lets us stop optimizing for zero errors and start optimizing for what matters now: more bits per cell, higher density, lower energy. The universal memory never showed up, but AI may be the workload that tells each technology exactly which job it was born to do.

Read the full breakdown by Giacomo Pedretti (MTS, AI Hardware Architectures) and Srenik Mehta (VP of Engineering): https://t.co/ukj4GnTC7b

[Unconventional AI @unconvAI]

What if your AI chip didn't calculate the answer, but physically settled into it?

That's the premise behind dynamical system hardware, and there's never been a standard way to program it. In an upcoming International Symposium on Computer Architecture (@ISCAConfOrg) 2026 paper, Unconventional AI MTS Prof. Ang Li and co-authors introduce DS-ISA: a minimalist 9-instruction architecture that bridges digital processors and continuous-time analog physics.

It's a proof-of-concept, but it's the abstraction layer the field needs to build real compilers and software stacks for dynamics-based AI hardware.

Full deep dive: https://t.co/WxfjrlgURm