ProBackend
AI Business
3 weeks ago6 min read

How AI Startups Make Money: Vantora's $100M Bet on Customer-Owned Physical AI Ventures

How AI Startups Make Money: Vantora's $100M Bet on Customer-Owned Physical AI Ventures

Vantora, the rebranded UP.Labs, raised $100 million from Silversmith Capital Partners — its first outside investment — and is shifting to a "proprietary M&A pipeline" where corporate partners invest in and can acquire the AI startups built for them. A grounded look at how that venture-lab model attempts to answer the question of how AI startups make money, what the reporting actually supports, and what remains unproven.

A Startup Factory Rebrands Around Physical AI

UP.Labs, which launched in 2022, has rebranded as Vantora. TechCrunch reports that the firm — described by the publication as "not quite an incubator, accelerator program, or venture firm" — builds startups designed to solve problems for corporate customers, and that it announced a $100 million investment from Silversmith Capital Partners alongside the name change. Porsche was its first corporate partner at launch; since then the company has struck deals with Alaska Airlines, J.B. Hunt, Wabash, and TDG, the parent of Ashley Furniture, and per TechCrunch it is now working with additional, unnamed customers in industrial manufacturing and the oil and gas sector.

The substantive change behind the rebrand is a narrowing of focus. Founder and CEO John Kuolt told TechCrunch that Vantora will now build startups solely for its corporate customers rather than for the broader market — and that corporate partners get the option to fold those startups into their core businesses. That is the news hook. The interesting part, for anyone tracking AI venture financings and investor strategy, is what the rebrand reveals about how a new class of venture-lab business tries to answer an old question.

How Do AI Startups Make Money? The Venture-Lab Answer

The recurring question in coverage of technology startups is how companies riding the current AI wave generate revenue at all — model providers sell tokens, tooling vendors sell seats, application startups sell workflows. A venture lab plays a different hand. Per TechCrunch's reporting of Vantora's model, the corporate partner invests in each venture the lab builds and serves as its first customer. Revenue, in other words, starts contracted rather than discovered: the lab's balance sheet risk is shared with a named enterprise backer before the startup ever pitches a broader market.

Two caveats belong here, because the source doesn't settle the economics. First, "first customer" is not the same as "only customer," and the reporting doesn't say what fraction of any venture's eventual revenue comes from the sponsoring partner. Second, the option for a corporate partner to acquire the startup is exactly that, an option, described by Kuolt as part of a planned "proprietary M&A pipeline," not a realized exit that the article documents. Whether acquisition terms will reward the lab, its founder, and its investors is unreported. What is verifiable today is narrower: one $100 million round, from one named investor, funding a model in which the customer helps own the company serving it.

The Proprietary M&A Pipeline, As Its CEO Describes It

Kuolt's framing, quoted by TechCrunch, is that Vantora is moving toward a pipeline in which it builds startups for corporate partners, those partners invest and become first customers, and the partners can then "fold the startups into their core businesses, and essentially keep them to themselves."

The rationale Kuolt gives is about deal flow he says he was losing. "We were missing on the biggest value problems, which had the biggest upside because of that," Kuolt told TechCrunch. In the older model, ideas that were strategically sensitive to a partner got spiked because they couldn't be brought to the outside market. His example: "Imagine you're a Fortune 100 industrial company and you need to retrofit all of your hardware and machines for autonomy. You need to own that, it needs to be sovereign, and you can't rely on a third party to go do that for you. You need to own that intelligence layer. They're never going to let us go sell that to their competitors."

A concrete near-miss: Kuolt said the firm once proposed an AI idea for partner J.B. Hunt, was told "there is no way you can take this out to the world," and passed on it, a project the new proprietary structure would now allow Vantora to pursue. That is the pitch as its CEO tells it: exclusivity unlocks use cases the open-market version of the lab had to walk away from.

Why Physical AI Sits at the Center of the Thesis

According to Kuolt, the pivot to customer-owned ventures directly drove Vantora's increased focus on physical AI, including with existing customers. The logic in his telling is that the deepest, most guarded problems in industrial settings, retrofitting hardware for autonomy, owning a sovereign intelligence layer over factories and fleets, are precisely the problems a company will not let a vendor resell to rivals. Physical AI problems are, in Kuolt's framing, disproportionately the "you must own this" category, which makes them a natural fit for a build-for-one-customer model.

It's worth flagging this as reported rationale rather than settled market fact. TechCrunch does not report evidence that physical AI ventures are outperforming Vantora's earlier, outside-market startups, and the competitive field for enterprise physical-AI spend is not covered in the article. What the reporting supports is a strategic bet: that exclusivity is the product, and physical AI is where exclusivity commands the most value.

A First Outside Dollar, and the Up.Partners Question

The Silversmith Capital Partners round is notable for what Kuolt says it is not: he noted to TechCrunch that the $100 million is the company's first outside investment. In its early days, UP.Labs was tied, although never financially, per TechCrunch, to venture firm Up.Partners. Kuolt's description of that residual relationship is narrow and specific: Vantora still shares office space with the California-based VC, but it is its own entity.

That's the whole of the public record here, and it's worth not over-reading it. The reporting doesn't describe any legal or capital linkage between Vantora and Up.Partners, past or present, beyond Kuolt's statement that the tie was never financial and that the office is shared. The rebrand is a rebrand; nothing in the source characterizes it as formalizing a separation that was informally complete from day one.

What to Watch, Without Overclaiming

Read conservatively, this is a single financings story in the ongoing VC news cycle about venture-lab economics, one datapoint, not a category verdict. Still, several observable markers will show whether the model holds up:

  • Deal cadence. TechCrunch lists Porsche, Alaska Airlines, J.B. Hunt, Wabash, and TDG as partners, with new unnamed industrial and oil-and-gas customers. Whether that roster grows, and how many ventures each partner sponsors, is checkable over time.
  • The first fold-in. The "proprietary M&A pipeline" is described by Kuolt but not yet demonstrated in the reporting. The first disclosed acquisition of a Vantora-built startup by a corporate partner will be the moment the model's economics stop being theoretical.
  • J.B. Hunt follow-through. Kuolt's own example of a previously unpursuable idea is a promise to readers; whether that venture gets built is a rare pre-committed milestone.
  • Silversmith's follow-on behavior. The firm wrote one large, first-outside-dollar check. Whether it backs the approach again as rounds for individual ventures surface is its own signal, and the source doesn't let us claim more than that.

The Money Question, Still Open

How do AI startups make money? Vantora's answer, as its CEO tells it, is to make the customer a co-owner of the solution before the market ever sees it, trading breadth for depth, and open markets for one deep pocket per venture. It's a coherent response, and it is squarely the kind of structural innovation investor strategy watchers should track. But as of this reporting it remains a thesis backed by one $100 million commitment, a partner list, and a promised pipeline, not yet a demonstrated revenue engine. The honest verdict: the model's first test hasn't been published, and neither will the verdict be until the first startup is folded into the company that paid for it.

Source: TechCrunch, "A startup that builds other startups raised $100M and is all-in on physical AI," Kirsten Korosec, September 18, 2026.