Nature has always been the most ruthless optimizer on the planet. Billions of years of evolution have produced systems — brains, forests, flocks of birds — that accomplish astonishing feats of computation and coordination on a vanishingly small amount of energy. Meanwhile, the AI industry is building gas-fired power plants to keep its models running.

We think that's the wrong direction. And we believe Refiant AI has found a better one — by going back to nature for the answer. That's why VoLo Earth is proud to be leading Refiant's Seed round.

Every major AI model today is built on the same fundamental architecture, and it has a scaling problem: the more context you give it, the more energy it consumes — quadratically. At production scale, today's frontier models process roughly 150–370 tokens per watt-hour of electricity. Google recently announced plans to build a gas turbine plant with carbon capture just to power its AI operations. When the most energy-intensive generation technology available is being bolted onto one of the most energy-hungry technologies ever invented, something structural needs to change.

As Refiant puts it: Adding more size stops feeling like progress and starts looking like a costly stand-in for optimisation. The industry has been chasing efficiency at the margins — better chips, smarter memory management — but none of it changes the underlying dynamic.

Refiant's founders asked a question that most AI researchers don't: how does nature do so much with so little? The smarter path, they concluded, draws on a principle visible across nature and cognition alike: focus resources where they matter most. A flock turns as one because each bird responds only to the signals closest to it. Human thinking works in a similar way — we narrow the field and move toward what seems most relevant, rather than searching every memory each time we need an answer. The human brain runs continuous, sophisticated reasoning on about 20 watts — the same as an LED light bulb — while modern AI systems require megawatts for far narrower tasks. Refiant has translated this principle — honed across billions of years of natural evolution — into a new kind of AI architecture. Instead of treating every piece of information as equally important at all times, the way today's models do, Refiant's system routes computation selectively, the way nature does. The result is that as the model handles more information, its energy consumption grows only slightly rather than ballooning. That's a fundamentally different curve, not a marginal improvement on the existing one.

Bold claims require rigorous validation. So we got hands-on: we orchestrated independent benchmarking with a senior engineer at a major cloud provider and our quantitative partners at Voloridge, who provided measurement and evaluation guidelines.

The results were striking. Refiant achieved a 10-100x better energy efficiency than three leading models, without sacrificing accuracy — a real-world demonstration that the nature-inspired approach works in practice, not just in theory. When we stress-tested the model by dramatically expanding the amount of information it had to hold in context — from 128,000 tokens to 1 million — its energy draw increased by less than 8%, versus the 4–5x increase you'd expect from a standard model. It also ran on a fraction of the hardware: 3–6 GB of memory, compared to the 64–80 GB required by comparable incumbent models. That's the difference between a laptop and a data center rack.

The best breakthroughs often come from people who look outside their own field for answers. Refiant's founding team fits that profile.

Viroshan Naicker, CEO, is a mathematician who studies how large-scale complexity emerges from minimal building blocks — and a published poet. He frames the company's thesis this way: people and LLMs are the only things that produce poetry, yet people do it with thousands of times less energy. That's not a metaphor. It's the research question. We found him to be exceptionally sharp and unusually able to bridge deep theory with clear commercial judgment.

Mathew Haswell, COO, brings experience scaling technology companies across multiple sectors and a track record of disciplined, capital-efficient execution — exactly what a company with a broad technological canvas needs to stay focused.

Sid Gutta, CTO, is a builder. He ran our eight-hour benchmark session with calm, methodical precision — pausing only when we'd kept him talking so long he needed water. That kind of composed technical leadership matters when you're betting on a team to deliver. The three have worked together across prior ventures, which meaningfully reduces early execution risk.

At VoLo Earth, we back capital-efficient solutions that accelerate decarbonization and generate compelling returns. Refiant sits squarely in that frame — and the timing is right.

The AI compute market is large and growing fast, with demand currently outpacing available infrastructure. Refiant's approach allows it to undercut incumbent pricing while running on commodity hardware. No specialized chips to manufacture. No new data centers to build. The competitive moat is the architecture itself, with a clear go-to-market path through partners' existing infrastructure and sales channels. The carbon impact is tangible. A 10x+ improvement in energy efficiency, applied at even modest scale across global AI workloads, represents a material reduction in data center electricity consumption — and in the new gas generation capacity being rushed online to support it.

We've long believed that the most powerful solutions to our energy and climate challenges won't come from fighting nature — they'll come from learning from it. Refiant is proof of that conviction: a team that looked at the elegance of the natural world, asked the right questions, and built something the industry didn't think was possible.

That's the kind of company we exist to back.

VoLo Earth Ventures is an early-stage venture capital firm backing capital-efficient solutions across the energy, mobility, building, and industrial sectors. Led by former RMI leaders and seasoned quantitative investors, the firm combines rigorous techno-economic analysis with hands-on portfolio engagement to accelerate decarbonization while generating superior returns. With more than 75 years of combined operating experience in the energy industry, VoLo Earth invests with clarity and conviction in the sector’s complexity — screening and winning high-quality deals, creating value from the first founder conversation, and driving portfolio companies toward successful exits. For more information, visit www.voloearth.com