Early stage venture capitalists have expressed caution about rushing to invest in the wave of artificial intelligence (AI) start-ups launching in the wake of ChatGPT’s release last November, warning it is not yet clear which applications are unique and offer long-term value.

AI has dodged the worst of the global slump in venture capital (VC) funding, with Crunchbase data showing the sector attracted $25bn in funding over the first half of 2023 — down from the $29bn funded over the same period last year. That compares to a 49% drop in VC funding across all sectors. 

However several seed-stage VCs speaking at the start-up event TechChill in Milan on September 27 said while there is money to be made in AI today, they are rejecting the herd mentality to avoid bad investment decisions.

“We are seeing a burst of entrepreneur activity, but not a lot of it is very high quality,” said Gil Dibner, founder of Angular Ventures, during a panel discussion at the event. “In fact, most of it is not particularly interesting or unique.”

Founders building AI products and services based on third-party large language models make up 90% of the AI pitches he sees. These businesses are not based on the founder’s proprietary technology and can be replicated by others. “That's pretty scary from a VC point of view,” said Mr Dibner, because many of these start-ups are looking to launch the same product which has no competitive edge over their peers.

Other investors at the event stressed that the AI revolution is just beginning and it is not yet clear which applications are novel and needed long-term. Alessandra Mazzilli, vice president at Berlin-based Cavalry Ventures, tells fDi: “There are definitely a lot of [AI] products that don’t have a lot of defensibility”, a reference to a business’s ability to maintain its competitive edge. “Those are the products that scare me,” she added. 

For an AI start-up to be interesting, it must have a clear return on investment, long-run defensibility and be mission-critical — meaning it is essential to business operations. “Probably 90% of the opportunities in generative AI right now don't have these characteristics,” says Ms Mazzilli.

More reading on AI:

Advertisement

• Mind the tech gap: the AI divide in Europe
• AI in FDI | Investment promotion to get "harder", argues Cinde's investment director
• AI in FDI | Economic development to shift from promotion to consultancy

Waiting for the dust to settle

Those in the industry liken today’s surge in creation and funding of AI start-ups to the gold rush. It has prompted VCs to focus on business models that provide the infrastructure for the AI era — known as ‘picks and shovels’ — which will be needed to support various AI applications.

However Mr Dibner, who has been looking at AI start-ups for a decade, said during his panel that his firm invested in one such start-up this year only to see 10 businesses coming out with identical products within a matter of weeks. “We had a board meeting [after that] and I said, ‘we need to slow way down ... There's no advantage to trying to be first in the wrong thing.”

Another reason they are holding back is to wait for the best entrepreneurs. “From a VC perspective, we are operating in very uncertain times,” Mr Dibner also said. “I think … a lot of smart founders are waiting on the sidelines because until the dust settles, it’s not clear what people should build.”

During this period of uncertainty, Cavalry Ventures is approaching AI pitches by placing a premium on mission critical products, meaning those which customers will stick with during a downturn. “The way we translate that into generative AI is offerings that are able to substitute at least one [full-time employee],” said Ms Mazzilli, contrasting it with products that improve workers’ efficiency. “That is when it becomes interesting with generative AI from a VC perspective.” 

While these innovations will make some jobs obsolete, she said “this is what technology is all about, unfortunately”, stressing that it frees up workers for more rewarding activities.