Inside a Silicon Valley AI venture with DVC

On September 30, Raison held a closed webinar with the founders and managing partners of DVC — Nick and Marina Davydov. In the article, we summarized the webinar: we explained how DVC and DVC AI Fund I are structured, examined the economics of the AI industry, and showed how it affects vertical selection.

How DVC is structured

DVC is a venture fund built around a community of more than 200 investors and founders who jointly source deals, invest, and help portfolio companies. The roles of founders and investors can change: the co-founder of Higgsfield AI started as an LP (limited partner, investor) in DVC, and the CTO of Perplexity AI became an LP after DVC’s investment.

Over five years, DVC has completed more than 160 deals. According to the fund, eight portfolio companies have become unicorns with valuations above $1 billion. Three of them, Perplexity, Etched, and Thinking Machines, reached decacorn status with valuations above $10 billion.

DVC has automated its routine work. An AI pipeline tracks startups across the market, builds company profiles, checks founders against court and bankruptcy records, collects feedback from community members, and recommends whom to contact for additional information.

DVC AI Fund I

AI startups receive funding through DVC AI Fund I. The fund invests primarily at early stages, when company valuations are lower than in later rounds. Despite the risks typical of early-stage venture capital, this, according to Nick, preserves greater upside potential for the position. Nick also noted that DVC considers exiting a company when its valuation reaches approximately $30–50 billion.

When selecting startups, the team looks at revenue growth and its dynamics. These data help assess the company’s further potential.

The AI portfolio includes major companies in which DVC was among the first investors: Higgsfield AI, Etched, and Perplexity AI. Less than two years after launch, Higgsfield AI reached $1 billion ARR (annual recurring revenue, annual recurring revenue) and is growing profitably. According to Nick at the webinar, DVC became the company’s first investor and participated in every round.

Etched develops chips for running trained AI models, while Perplexity is an AI-powered search service. Both companies have decacorn status.

The economics of the AI industry

Nick divides the AI industry into five layers: energy, chips, cloud, foundation models, and applications. The first three belong to AI infrastructure.

Right now, for every dollar a user pays in the applications layer, $12 is invested in infrastructure. Right now, that makes sense. But models are getting better and cheaper, and compute capacity is increasing. At some point, these lines will cross.

Co-founder and Managing Partner of DVC

Nick Davydov

Nick sees the main risks in the cloud segment. Companies assume a GPU (graphics processing unit) lifespan of six years, and their margins depend on that. If the lifespan drops to five years and financing becomes more expensive, some companies may become unprofitable. In Nick’s view, the correction may begin here.

In AI, costs rise together with revenue: the more customers, the higher the compute expenses. Nick compared this economics to industrial economics. At the same time, a sustainable business model has not yet emerged: companies are testing subscriptions, token-based pricing (tokens are units of text processed by the model), and pay-per-task pricing.

How DVC selects verticals

This economics leads to the selection approach: DVC analyzes each vertical separately. During the webinar, Nick broke down this approach using U.S. healthcare as an example.

First, the fund team determines who pays: private insurers, government programs Medicare and Medicaid, or patients themselves. Then it looks at what is being paid for and to whom, which companies already sell software, and what functions exist within the industry. After that, the team determines which stage AI is at: assisting a human (co-pilot) or operating autonomously (autopilot). DVC looks for verticals that are moving from the first stage to the second.

Nick believes that even strong AI will not remove constraints in value chains. New drugs will get stuck in clinical trials, then at the U.S. regulator FDA, and then in distribution. That is why DVC chooses companies with a durable market advantage: technology, data, sales channels, or another barrier that is difficult to replicate.

The team then looks for the strongest founders and checks them through feedback from people in the industry. If there are no suitable founders in the market, the fund waits. DVC’s first check is usually small. When a company confirms growth, the fund increases its investment.

Results and timing

According to Nick, DVC has already returned one-third of the capital to investors in its first fund. Distributions began earlier than the team had planned. The fund is designed for 10 years, but Nick expects investors to be paid back sooner.

During the first four years, the fund invests in companies. Over this period, it should enter deals at different points in the market cycle. In venture capital, results depend on the number of deals, so a small portfolio will not deliver the desired outcome.

Nick does not predict when DVC will sell its stakes in companies because the market is unpredictable. Conditions for exits remain favorable for now. In his view, the next market cycle will arrive in a couple of years.

Participation through Raison

Raison invests in DVC AI Fund I and is one of the fund’s LPs. We previously covered the cooperation between Raison and DVC.

If you would like to learn about the participation terms, leave a request in the Raison bot or contact your personal manager.