He signed the letter. Someone at the LP meeting slid it across the table. The room changed pitch in that single moment.
I’ve watched venture cycles bend careers and capital before, and you should notice this one. You’ll want to read the signals here the way an investor reads a balance sheet: small details matter.
At a May limited-partner call someone passed around a letter — Ashton Kutcher is leaving Sound Ventures
The actor and early-stage investor who co-founded Sound Ventures with Guy Oseary in 2015 is stepping away to start a new firm with Morgan Beller. Sound Ventures, which manages close to $2 billion (€1.84 billion), has been an early backer of fintech winners like Brex and Affirm and enterprise plays such as GitLab.
But Kutcher’s public arc at the firm tilted toward the AI winners—more than $800 million (€736 million) has been allocated into Anthropic, OpenAI and World Labs, and a leaked filing suggested Sound holds a 0.15% stake in OpenAI valued at about $1.3 billion (€1.2 billion).
In a letter shared with limited partners, Kutcher wrote he plans to invest “infrastructure, energy and deep tech.” The split, according to that same letter, came down to different visions: timing, capital structures and operational needs that no longer matched.
Why is Ashton Kutcher leaving Sound Ventures?
You can chalk this up to a classic founder-versus-fund divergence. Kutcher wants to back earlier-stage projects that require patient capital—quantum, nuclear, aerospace and robotics—while Sound has been leaning into later-stage AI and software winners. That misalignment turns into hard choices when the money needed is measured in megawatts and square feet, not lines of code.
On a tour of a hyperscale data center the air smelled of coolant — the new focus is physical infrastructure
Kutcher’s stated aim is infrastructure, energy and deep tech: physical layers that keep advanced models alive. If frontier models are the flashy engines, the new bets are the fuel tanks, power plants and fiber that make continuous operation possible.
Investors are already thinking like this: build the pipelines that feed AI, rather than only funding the labs that race to build the fastest model. It’s a safer ledger move—those assets have utility beyond a single AI cycle. Think of it like a city building new highways; even if one car maker flops, the road still carries traffic.
Morgan Beller brings crypto and platform experience from NFX and Meta’s Libra effort, signaling this won’t be a purely hardware play. Expect a portfolio that spans real-world engineering and software that orchestrates it.
What will Kutcher’s new firm invest in?
Short answer: long-horizon, capital-intensive projects that support compute at scale. That includes data centers, power generation and grid integration, advanced manufacturing, robotics and possibly quantum stacks. The team will need different fund structures and longer timetables than the software bets Sound favored.
At late-night industry dinners the whisper is no longer only about models — capital is creeping toward supply chains and power
This move is a signal, not a shout. When a visible investor swaps model bets for metal and megawatts, it nudges others to re-evaluate where risk lives in the AI boom. If the lab frenzy cools, physical assets—racks, substations, reactors—still hold value.
Large tech players are already hedging: Microsoft and Meta are framing data-center strategies around AI demand, and even SpaceX has talked about repurposing infrastructure. If you believe models need constant scale, then the real money may flow into what keeps scale affordable and reliable. It’s as if a dam were being built for an incoming tide—one side plans for the water, the other for the force behind it.
How will this affect AI funding?
Expect a slow rotation rather than an abrupt reallocation. Labs and algorithm teams will keep attracting talent and capital, but more dollars will be earmarked for deployment—power, cooling, networking, and specialized fabs. Funds that can stomach long development cycles and regulatory complexity will get first access to those assets.
Guy Oseary and the remaining Sound team still control a strong franchise of software and model bets, and Kutcher’s move is less an indictment than a wager on the next infrastructure layer beneath AI. I’ll be watching the fund structures and who signs up as LPs—those names tell you whether this is a niche play or the start of a broader shift.
If capital follows Kutcher’s hunch, do you think the next decade of AI will be defined by smarter models or by the physical systems that power them?