Investment in generative AI startups, which are developing AI-powered products that generate text, voice, video and more, isn't slowing down, but it's consolidating into earlier-stage ventures and their numbers are shrinking.
According to Crunchbase data obtained by TechCrunch, 225 startups raised $12.3 billion in venture capital in the first half of 2023 (January through July 16). If this trend continues, generative AI companies are on track to raise as much as or more than the roughly $21.8 billion they raised in 2023.
Here's a breakdown of the first half of 2024 totals:
198 Angel/Seed Investments: $500M 39 Early Stage Investments: $8.7B 18 Late Stage Investments: $3.1B
Early-stage startups were the clear winners, including Elon Musk's xAI ($6 billion raised in May), China's Moonshot AI ($1 billion raised in February), Mistral AI ($502.6 billion raised in June), Glean ($203.2 million raised in February), and Cognition ($175 million raised in April). According to Chris Metinko, an analyst and senior reporter at Crunchbase, investors seem to be betting on larger startups that they believe have a higher chance of success and “killing” those they are less sure of in the early stages.
“Some venture capitalists expect that potential legal and regulatory dilemmas that AI companies may face both in the U.S. and abroad could lead to a slowdown in funding for AI,” Metinko told TechCrunch. “Others point to the fact that when the mobile revolution happened over a decade ago, the biggest winners in terms of the underlying infrastructure layer ended up being established technology companies.”
Metinko points out that the fate of many generative AI businesses is unclear, even for the best-funded companies.
Generative AI models are typically trained on data such as images and text taken from public web pages, and the companies argue that fair use protects them from legal action if that data is found to be copyrighted. But it's still not clear whether courts will ultimately rule in favor of generative AI companies, which is perhaps why some companies are starting to enter into licensing agreements with copyright holders.
Regardless of the court's outcome, quality training data is becoming harder and more expensive to come by as startups eat up the web's supply and more creators block crawlers from scraping their data. (One analysis predicts that the market for AI training data will grow from $2.5 billion to $30 billion within a decade.) The process of training a model isn't getting easier or cheaper, either: A recent Stanford University report estimated that OpenAI's GPT-4 cost $78 million to train, compared with Google Gemini's price of $191 million.
Given the large upfront investment required to build flagship models, it's no surprise that so few generative AI startups, even big names like OpenAI and Anthropic, are profitable. OpenAI, which has reportedly generated about $3.4 billion in revenue, could lose $5 billion this year, according to The Information.
Investors in generative AI seem to be in for the long haul. In particular, big tech investors like Google, Amazon, and Nvidia view generative AI as a strategic bet. But will the bubble burst soon? That seems like a real possibility if generative AI startups can't overcome the existential challenges they face.