No product, no problem: investors place big bets on AI neolabs

{"text":[[{"start":5.12,"text":"It would be a brave or foolhardy investor who set out today to build a new Walmart, Uber or Amazon. Yet in the world of AI models, venture capitalists are betting that there is still room for challengers."}],[{"start":17.12,"text":"Neolabs, as such AI hopefuls are known, are attracting tens of billions of dollars of investment, even though many of them lack products, markets or revenue. In the past two quarters alone, neolabs have raised $24bn, nearly five times as much as giants OpenAI and Anthropic did in the years up to the launch of ChatGPT, according to Radical Ventures."}],[{"start":39.32,"text":"Many boast supercharged valuations: month-old Emulate, founded by ex-Google DeepMind researchers, is on track to score a post-money valuation of nearly $4bn. Safe Superintelligence, established by OpenAI co-founder Ilya Sutskever in 2024 and yet to reveal anything resembling a product, raised money at a $32bn valuation last year."}],[{"start":null,"text":"

"}],[{"start":62.28,"text":"Setting up a neolab calls for two main ingredients: talented people and access to chips and servers — “compute” in tech speak. Both inputs are scarce and expensive. Research scientists, in particular, are costly to hire and retain. This is not the garage-dwelling college dropout of Silicon Valley’s yesteryear."}],[{"start":80.84,"text":"The result is that neolabs guzzle capital. Fundraising rounds are far more frequent than have been seen before: a median 7.6 months, calculates venture capital firm Chapter One, with a 3.5-times increase in price. Nonetheless, the capital isn’t so hard to come by. Reflection AI, a developer of open-weight AI models users can tailor and tweak, raised $105mn at a valuation of $555mn, only to pull in 25 times as much at a $25bn valuation barely a year later."}],[{"start":113.4,"text":"True, these numbers can hide some sleight of hand. It’s not unheard of for investors to pony up cash quietly at one valuation, then sign on to a more loudly touted fundraising at a higher valuation soon after. Founders can thus boast about the larger valuation implied by the second capital raise, while investors can say their average entry price was lower."}],[{"start":133.2,"text":"Like any venture capital bet, there’s an element of spray-and-pray. But unlike, say, supermarkets or rideshare, there’s every chance a neolab could elbow its way into the top tier. For one, many upstarts are focusing on specific applications, such as “world models” that build simulations of complex environments, or “physical AI” that powers autonomous machines."}],[{"start":154.84,"text":"Moreover, even for the biggest companies, customer habits are still in flux, and companies that use AI are proving keen to shop around for better tools at better prices. It’s too early to say which neolabs stand a chance of entering the AI pantheon, but for investors who have spread their bets — and also benefit from their spending, such as chipmaker Nvidia, which has backed more than half of the cohort — taking a punt on the hopefuls looks smart."}],[{"start":184.88,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1790315016_2811.mp3"}