bigrus

Seed VCs warn AI startups will need ‘revenue quality’ in 2025

Harry November 5, 2024

Fundraising in 2025 will continue to be a “tale of two cities,” VC Renata Quintini, co-founder of early-stage VC Renegade Partners, said on stage at TechCrunch Disrupt last week.

“Some companies that show promise to go after large, fast-growing markets will receive a lot of funding and momentum.” But on the other hand, Quintini warned that “companies that need to build real businesses and productive businesses” will have a hard time finding cash.

He talks about the tight fundraising market startups face in the era of high interest rates. in 2023 Approximately 3,200 startups died They easily raised money during their party time in 2021. In 2024, venture capitalists turned to funding AI companies, and many other early-stage startups struggled to raise money and more deaths occurred. Fintech was particularly brutal In 2024.

Your best chance of getting venture funding come 2025 will be to have solid business fundamentals. (Unless you say, a world-renowned artificial intelligence scientist.) This means selling a product or service at a profitable price point that serves a large customer base.

But wait! Greylock partner VC Corinne Riley warned on stage that this would be about more than just reaching paying customers.

Riley says there is “no set milestone” in terms of sales or growth that will turn venture capitalists’ heads next year. “There’s no significant number that says if you can pull this off, you’ll raise a successful Series A. What we’re really looking for is not the quantity of ARR, but the quality of ARR (annual recurring revenue).”

In other words, once customers get on board, do they stay? Do they tend to increase their spending on the venture over time? The startup may actually have fewer customers and less revenue than its competitors, but if the customers it signed up for remain, investors will write checks.

“We look for a quality customer base that you can repeat over and over again when you have more capital,” Riley says.

“This is what VCs mean when they say ‘you need to build moats,'” VC Elizabeth Yin, co-founder of Hustle Fund, said on stage. This is one way of explaining how to lock in customers so they don’t choose to leave. “The more you do unique things that others can’t do, the more it helps you.”

Riley offered as an example a Greylock portfolio company called Braintrust, which helps developers build AI applications and evaluate their performance. Greylock was convinced to lead the $5 million seed deal because it found initial customers that were “industry acclaimed,” founder Ankur Goyal said. He explained that Zapier, Coda, Airtable, and Instacart are his customers, “people at the forefront of product development with AI.”

If well-connected key customers are happy with the offer, they bring in other such customers and the cycle continues. It is stated that Braintrust has also acquired other well-known technology names such as Stripe and Notion. Andreessen Horowitz’s Martin Casado in October Led $36 million Series A, With Greylock in attendance.

This type of “customer quality” has always been important to investors; therefore naming customers is important. But in 2025, in the wake of the AI gold rush, this metric will become even more critical because a large portion of the revenue generated by many AI startups will become one-time revenue.

In another panel Disrupt, someone who bounces back after a down tour, VC Elliott Robinson, partner at Bessemer Venture Partners, explained the situation. By the beginning of 2024, the board of nearly every major company is biting its nails on AI, allocating huge research budgets to its CIOs to buy and discover products.

“We’ve now lived 18 and 24 months buying AI products and seen companies go from zero to four (million dollars in revenue) and from four to 20 (million dollars) in revenue,” he said. “Now the question is; What will be renewed? Because the CIO budget is starting to dry up.”

CIOs will continue to purchase only products that make a measurable difference. So all this revenue and all these customers of so many AI startups (and potentially customers in near-AI fields like application monitoring) doesn’t mean that any particular startup is a good bet for the future.

Or as Quintini describes it, “What you’re trying to do at the end of the day is build something that comes together. Second, you will either work at something that you can execute faster than others, or you will do something that other people cannot imitate.