There is a particular kind of funding round that exists mostly to make other founders feel bad, and Viktor’s new one is squarely in that genre. The Warsaw-and-Munich startup raised €64.7 million in a Series A led by Accel. Its product launched publicly in February 2026. By the time the round was announced in May, it had reached a $15 million revenue run rate, in roughly ten weeks, and was being used by more than 2,000 organizations.

Ten weeks. To put that in context: most startups have not finished arguing about their logo in ten weeks.

Viktor is an “AI coworker.” It lives inside Slack and Microsoft Teams, connects to the tools a company already uses, and, this is the part the company keeps insisting on, takes responsibility for outcomes rather than just answering questions. You message it like a colleague and it goes off and builds the report, the dashboard, the campaign, the internal tool, the code. Founder Fryderyk Wiatrowski, formerly of Meta, frames it bluntly: “We’re seeing teams treat Viktor like a hire, and that is the shift we want to bring to every company.”

The cap table is doing a lot of the talking

Normally I’d tell you to ignore the angel list in a funding announcement, because it’s usually three famous names and a lot of “and others.” Viktor’s list is not that. The round drew Stewart Butterfield and Cal Henderson, the co-founders of Slack, which is notable, given that Viktor’s entire premise is living inside Slack. It drew Guillermo Rauch of Vercel, Victor Riparbelli of Synthesia, the Framer co-founders, Deel’s Alex Bouaziz, Lenny Rachitsky, Harry Stebbings, plus executives from DeepMind, Figma and Lovable. The earlier backers already included Nat Friedman and Daniel Gross.

When the people who built the platform your product runs on decide to put their own money in, that’s either a very strong signal or a very expensive way to keep an eye on you. Probably both. Either way, it’s the kind of list you can’t fake, and in a market drowning in AI-agent demos that work beautifully exactly once, “people who would obviously know better are buying in” is a real data point.

The honest question about any AI agent isn’t whether it’s impressive in a demo. It’s whether anyone’s still paying for it in month four. Viktor’s whole pitch is that the answer is yes, and it has the run rate to wave around.

What it actually claims to do

The specifics are where these stories usually get either convincing or queasy, so here are Viktor’s, lightly salted. The company says Viktor can operate autonomously for weeks rather than minutes, holding context across thousands of emails, documents and tools. Most customers wire in more than 30 applications. And the case studies it’s chosen to publicize are pleasingly weird: it says a health-and-lifestyle group used Viktor to cut $4.5 million from a construction budget; that one founder stood up the infrastructure for a $1-million-a-year content agency, with no staff, in nine days; and that a landscaping business, a landscaping business, ran 63 automations across operations, HR, collections and email in two weeks.

I have no way to verify the landscaping company’s automation count, and neither do you, and that’s the appropriate level of skepticism to hold. But the shape of the claim is the interesting bit. Viktor isn’t selling a co-pilot that makes your existing employees 20% faster. It’s selling headcount, a thing you “hire” that does the work of a person, billed like software. If that framing holds up past the novelty period, it reprices a lot of jobs. If it doesn’t, it’s an extremely well-funded autocomplete.

The European footnote that isn’t a footnote

It’s easy to read “Accel-led, $15M run rate, famous angels” and assume Bay Area. It isn’t. Viktor was founded in 2023 by two former Meta engineers, Wiatrowski and Peter Albert, and is run out of Warsaw and Munich. The new money will fund a global rollout and, yes, a New York office, the now-standard move where a European company that found product-market fit at home goes to plant a flag in the market it’s selling into.

But note the direction of travel. The company didn’t relocate to raise. Accel came to Warsaw. The Slack founders wired money to a Polish-German startup building on top of their platform. For a continent whose defining tech anxiety has always been “our best companies leave,” a ten-week run rate and a marquee Series A that stayed put is the kind of counter-evidence that’s getting harder to wave away.

The risk is the obvious one: run rates that arrive in ten weeks can leave in twelve, and “AI employee” is a promise that gets audited the moment the novelty wears off and the renewal invoice lands. Viktor has bought itself a long runway and a cap table full of people motivated to make it work. Now it has to do the genuinely hard thing in this category, which is not launching. It’s still being indispensable in a year.

Synthesia, the London company that makes AI avatars read your corporate training script so a human doesn’t have to, has raised $200 million and is now worth $4 billion. A year ago it was worth $2.1 billion. So the valuation roughly doubled, which in the current AI environment is the sort of thing that happens on a Tuesday and barely makes the newsletter.

What makes it worth your time is the second half of the announcement, the half that isn’t about raising money at all. Alongside the Series E, led by existing investor GV, with Kleiner Perkins, Accel, NEA, Nvidia’s venture arm and others piling back in, Synthesia is running an employee secondary sale, in partnership with Nasdaq, that lets early team members turn their shares into cash. And it’s doing it in an unusually clean way. Pay attention to the plumbing here, because the plumbing is the story.

Why this is a big deal and a small deal at the same time

Employee secondaries are not new. People who join a startup early get paid partly in equity, the company stays private for ten years, and at some point those people would like to, you know, buy a house. So they sell some shares to a willing investor. This happens all the time, usually off to the side, often at a price that has nothing much to do with the company’s official valuation, sometimes below it, sometimes above, occasionally in a way that quietly annoys everyone else on the cap table.

Synthesia’s version is tidier. It is running the sale through Nasdaq acting as a private-markets facilitator, not as a public exchange, to be clear, Synthesia is emphatically not going public, and every employee sale is pegged to the same $4 billion valuation as the Series E. One price. Same price the new money paid. The company keeps a measure of control over who buys in, two new investors (Matt Miller’s Evantic and the famously cagey Hedosophia) come onto the register, and the staff get liquidity without the usual back-alley pricing drama.

The valuation round is what gets the headline. The secondary is the part that tells you how long these companies now expect to stay private, and how much machinery they’ll build so their employees don’t have to wait for an exit that may never come.

The real signal: the window reopened, and they built an off-ramp anyway

Step back and the structure tells you something the press release won’t. Companies build elaborate, blessed, fairly-priced secondary programs precisely when they have given up on the old liquidity schedule, the one where you IPO around year seven and everyone’s options turn into money. Synthesia was founded in 2017. It is now nine years old, has more than 500 employees, an HQ in London and offices in Amsterdam, Copenhagen, Munich, New York and Zurich, and it is signaling that it intends to stay private for a good while longer.

Synthesia’s own head of policy, Alexandru Voica, basically said as much: as UK private companies stay private longer, structured cross-border employee liquidity “may become increasingly common.” Translation: the IPO is no longer the assumed finish line, so we are building our own off-ramp and putting Nasdaq’s name on it for credibility. For a generation of European startup employees who watched the IPO window slam shut in 2022 and stay unreliable for years, that’s arguably better news than the valuation. The window has since cracked back open, Klarna listed in 2025 and a long European pipeline is forming behind it through 2026, but an open window is a moment and a blessed, fairly-priced secondary is a standing option, which is the one Synthesia just built. A $4 billion paper number you can’t sell is a story you tell at parties. A $4 billion number you can sell some of, at the marked price, is a mortgage.

And, fine, the business is real

The reason any of this works is that Synthesia, unlike a lot of companies attaching nine zeros to themselves this year, makes money in a way you can explain to a skeptical relative. It crossed $100 million in annual recurring revenue in April 2025. Its customers, Bosch, Merck, SAP, are exactly the kind of large, unsexy enterprises that have enormous training budgets and no desire to fly a film crew to forty offices to shoot the same compliance video in nine languages. Synthesia’s avatars do it for a fraction of the cost, which is the entire pitch and, refreshingly, a true one.

The next act is the one every software company is now contractually obligated to announce: agents. Synthesia is building AI agents that let employees “interact with company knowledge” by asking questions and role-playing scenarios, and CEO Victor Riparbelli is making them a “core strategic focus.” Whether that becomes a real second product or a slide that ages badly is a 2027 question. The avatars, at least, already work and already bill.

So here is the whole thing in one breath: a London company most people outside enterprise software have never heard of doubled to $4 billion, didn’t need to go public to make its employees rich, and used Nasdaq as a tasteful intermediary to prove the point. It’s a small structural innovation dressed up as a giant funding round. The giant funding round will be forgotten by spring. The structure, staying private, paying your people anyway, at a real price, is the part everyone else is going to copy.