The Strategic Split at Sound Ventures
Ashton Kutcher is out at Sound Ventures. After eleven years building the firm alongside Guy Oseary, Kutcher is stepping away to launch a new, early-stage venture fund with Morgan Beller.
Let's be clear about one thing: this isn't a post-mortem cleanup. Sound Ventures isn't bleeding, and this isn't a sign of institutional rot. Quite the contrary. The firm has put up numbers that most Valley shops would kill for. They were early in OpenAI, Anthropic, and Fei-Fei Li’s World Labs. They backed Brex. They backed Gusto. Those are massive, generation-defining home runs.
But venture capital is a game of alignment, and when you disagree on where the next epoch of value will be created, you split.
Sound has increasingly leaned toward mature, late-stage growth bets. They want established, de-risked winners. Kutcher, however, wants the early-stage wild west of AI infrastructure, deep tech, and energy. It's a classic GP divergence. Instead of fighting it or muddying the firm’s strategy, they're parting ways amicably. Kutcher remains an adviser to Sound, and Oseary—along with Sound general partner Effie Epstein—will advise the new firm. It's clean. But the underlying strategy shift is what we need to talk about.
The Battle for the AI Underbelly
For the past three years, the venture capital play has been simple to the point of laziness: back the massive foundation model labs. If you got into the early rounds of OpenAI or Anthropic, you won.
But the math is shifting.
The foundation model layer is capital-intensive, crowded, and increasingly looks like a utility business. Building a model costs billions in compute. Kutcher’s new fund isn't chasing the bots. They are chasing the pipes.
The focus is the infrastructure, energy, and deep tech that power these labs. We are talking about silicon, power grids, cooling systems, and specialized hardware. AI models eat power. They require megawatts of electricity and massive data centers. If you control the energy bottleneck or the infrastructure layer, you control the margins. This shift is already fueling massive valuations for companies building specialized open-source AI neoclouds.
As someone who watched these cycles at Sequoia, I know how fast the platform premium decays. The software layer gets thin when compute is the only moat. That's why early-stage money is migrating toward the physical and vertical constraints. You can write about SaaS all day, but if the local grid can’t power the cluster, your AI agents go offline. That's the real thesis here.
Morgan Beller and the Search for Hard Science
To pull this off, Kutcher isn't doing it alone. He's partnering with Morgan Beller, and that name carries major weight in the Valley.
Beller was most recently a general partner at NFX, the seed-focused outfit. Before that, she was the co-creator of Meta's cryptocurrency project, Libra. And she spent three years at Andreessen Horowitz.
Beller understands massive networks and hard regulatory and technical headwinds. The intersection of cryptocurrency, decentralized systems, and AI infrastructure is where she thrives.
Kutcher and Beller aren’t looking for another software wrapper. They’re looking for hard science and engineering breakthroughs. Deep tech. This isn’t a quick line-of-code play. You can't mock up a deep tech prototype in a weekend. It requires serious capital, PhDs, and years of R&D. But when it works, the IP is a fortress.
It's a stark contrast to the trend of consolidation and service roll-ups. While some firms are playing the cash-flow game through venture roll-ups, Kutcher and Beller want to fund the technology that makes those systems possible. They are entering the market at the seed level, looking for the weird, risky engineering teams that other firms might pass on.
The Power Grid and the Compute Bottleneck
Let’s talk about the physical reality of the AI boom.
Every prediction about AI agency assumes we have infinite energy. We don't.
We are running out of grid capacity in northern Virginia, Oregon, and the Dublin suburbs. The energy bottleneck is the single greatest threat to frontier AI. Kutcher's move is a bet that the next generation of founders won't build software. They will build small modular reactors, geothermal grid connections, and custom ASIC chips that cut power draw by 90%.
This is why the early-stage pivot is so crucial. At the growth stage, you're buying into established architectures. At the seed stage, you can fund the radical departures. Kutcher has been investing in technology for a long time. Stanford finance professor Ilya Strebulaev, who compiles data on top VCs, noted that Kutcher and his fund consistently rank among the top unicorn investors. It is an impressive track record.
He’s also known Sam Altman since Altman founded Loopt in the mid-2000s. He has seen the entire arc of Altman’s career, from mobile check-ins to OpenAI. That relationship gave him a front-row seat to the scaling laws. And those scaling laws are now saying one thing loud and clear: we need more power.
Finding a Moat in a World of Slop
The timing of this split is perfect. The market is tired of AI slop.
People are beginning to see the limits of pure software wrapping. If you want to build lasting value, you have to build at the physical or hardware level, or inside highly specialized verticals. (Firms are discussing this shift at events like StrictlyVC Los Angeles).
Kutcher and Beller’s new firm is aiming directly at this reality. They want the stuff that is hard to copy. It's risky. Deep tech investments have long dev cycles and high failure rates. But the payoffs are systemic.
If you own the energy solution, you own the lab.
We don't know the name of the new fund yet. It doesn't matter. The playbook is clear: find the hard problems, fund them early, and let the late-stage firms bid up the valuations later. While Sound Ventures continues to write expansion checks for the giants, Kutcher is getting back in the mud of early-stage tech.
VC is circular that way. First you fund the platforms. Then you fund the infrastructure. Then you search for the next platforms.
The cycle is starting again.