It’s safe to say that VCs struck while the iron was hot this year where it concerned generative AI.
While venture capital investments overall fell compared to last year thanks to macroeconomic challenges and other related factors, startups in the generative AI space — and AI more broadly — did quite well.
Funding for AI-related startups surpassed $68.7 billion in 2023, according to PitchBook, with generative AI vendors like OpenAI, Stability AI and Anthropic accounting for a substantial portion of that figure. And it appears that the sector will likely close the year with substantially higher investments than the past couple of years.
But could the top-level numbers be misleading?
A report on AI investment in Q3 by PitchBook, released Tuesday, found that “mega-deals” (i.e., multi-hundred-million-dollar investments from big-name backers) vastly inflated deal totals this year.
For example, just a few months ago, Amazon pledged to invest up to $4 billion in Anthropic, the company developing the AI-powered chatbot Claude. OpenAI secured a $10 billion investment from Microsoft (albeit not all at once and partly in the form of cloud compute credits). Inflection AI, a firm creating what it describes as more “personal” AI assistants, raised $1.3 billion in a funding round led by Microsoft. The list goes on.
In Q3, VC funding inclusive of mega-deals totaled around $22.1 billion. But after subtracting the tech-giant-led tranches secured by generative AI startups, the total is closer to $15.1 billion for the sector.
That’s a big discrepancy, but it isn’t exactly unanticipated.
Earlier this year, PitchBook reported that the number of VC deals involving generative AI companies declined 27% in Q3, while aggregate deal value increased by 39% to $6.1 billion, buoyed by massive investments from Big Tech companies. The aggregate value would’ve been lower, too, had Big Tech giants not thrown their hats in the ring.
Could it be that VCs, initially swept up by the generative AI hype, are coming down to earth and/or growing wary of the tech’s risks? Perhaps.
The case of Stability AI, the startup behind the open source image-generating model Stable Diffusion, is instructive. After raising more than $100 million in VC capital within the last 14 months, the company is now said to be positioning itself for an acquisition at the behest of its investors, including Coatue, which has grown concerned about its financial position.
Running a cutting-edge AI venture is expensive, it turns out. Not only do data scientists command sky-high salaries, but the computing power necessary to train and run AI models is exceptionally costly. Case in point: One of Microsoft’s premiere AI products, the code-generating GitHub Copilot, has reportedly been losing the company up to $80 per user, per month due to model inferencing costs. And it costs OpenAI an estimated $700,000 a day to operate ChatGPT.
Lawsuits over generative AI tech might be giving investors pause, too — especially those concerning IP disputes. Because of the way generative AI models are architected, they sometimes spit out mirror copies of the millions to billions of examples of text, images and so on that they’re trained on, and not all of those examples are in the public domain.
Microsoft, GitHub and OpenAI are currently being sued in a class action lawsuit that accuses them of violating copyright law by allowing Copilot to regurgitate licensed code snippets without providing credit. Elsewhere, authors in California and New York have sued OpenAI for alleged IP theft of their works.
A number of tech giants have promised to defend customers against copyright lawsuits implicating their generative AI technologies. But many startups, including Stability AI, haven’t, and likely won’t given the substantial legal expenses such protection would entail.
Now, this isn’t to suggest it’s all doom and gloom in AI startup land.
The PitchBook report out today suggests that exit activity is showing signs of recovery in 2023, with 108 AI exits so far worth a combined $9.1 billion. MosaicML, which Databricks acquired for $1.3 billion, and InstaDeep, bought by vaccine developer BioNTech for $680.9 million, were the biggest winners here by far.
AI modernization via mergers and acquisitions appears to be a reviving source of AI exits, PitchBook says, adding that more “mega-exits” over $1 billion are expected next year.
PitchBook also notes that all AI verticals, except for consumer AI and transportation, raised more than $1 billion for the time period the report tracked — a first since the broader VC funding downturn.
Will the momentum endure? That’s the billion-dollar question. But PitchBook, for what it’s worth, asserts that the trend shows “health for late-stage funding in specific domains.”