It turns out there’s more to AI than OpenAI.
Billions of venture capital dollars—bolstered by a solid secondaries market driven by unprecedented numbers of employee tender offers—have poured into top artificial intelligence startups. Those operators, among them Anthropic and Elon Musk’s xAI, may by now be startups in name only. They’ve amassed thousands of employees and millions of servers, their cash reserves rival publicly traded tech giants, and their fundraises are routinely oversubscribed.
That top-heavy sector concentration, however, may be just what AI-focused Rackhouse Venture Capital was built for. On Tuesday, the San Francisco-based firm closed its second early-stage venture vehicle on $45 million.
The firm’s first vehicle, which is still deploying capital, raised $22.3 million and held a final close in 2021.
Rackhouse founder Kevin Novak, formerly Uber’s inaugural head of data science, said that the investment imperative for Fund II is to do a due diligence dive below the frothy multiples and cash-burning tendencies of flashy AI names. It’s an apt tactic, considering Novak’s limited dry powder.
“The fund sizing, in general, has definitely precluded us from being involved in certain types of deals … at our size, we are not typically investing at the [large language model] layer of generative AI,” Novak said. “There’s a lot of open debate; a lot of reasonable people can disagree about how many LLMs the world needs, but what we all agree with is that these are companies that need hundreds of millions of dollars of VC money before they can fit to a product market-fit or prototype.”
Rackhouse, which backs AI startups in the Bay Area and elsewhere in the US, has a tall task, then: Deploying limited partner capital in smaller chunks to companies with growing pains that have a steady tech stack and definitive customer base.
That also means seeking out generalist operators with broader potential customer bases over constrained specialists battling it out in the generative AI weeds.
Rackhouse’s flagship fund stakes clocked in with an average post-money valuation of about $12 million, Novak said, otherwise known as “boringly on trend.”
“Our typical fund size, we need to be really focused on working with companies that have a complete business model,” he said. “There isn’t going to be a lot of blind alley or sort of, ‘Oh, we were right, but a year or two early. We need to wait for sales.’”
AI seed slowdown
What they won’t be waiting for is competition—which, for the moment, is sparse.
Generative AI deal value in pre-seed and seed rounds plunged 76% from its nadir in Q3 2023 to reach only $122.9 million in Q1 2024, according to PitchBook data.
Rackhouse has already tapped into LP interest in a couple of its early exits, including business software startup Spry and real estate technology services platform provider Onder.
“All the secular tailwinds … we definitely benefited from raising an AI fund last year,” Novak said.
Fixating on those off-the-beaten-path names and keeping valuations in-check will be key. On the scorching side of the limited partner infatuation with AI, Novak said he was approached by a family office weighing its first OpenAI investment.
Strings were attached.
The potential deal for OpenAI equity floated on the secondary market came wrapped in a picture-perfect, triple-wrapped special-purpose vehicle. Meaning three layers of fees.
“And the family office is like, ‘Well, it’s OpenAI,’” Novak said. “’Should we buy it anyway?’”
Featured image by Matt Anderson Photography/Getty Images
Learn about our editorial standards.
