The AI revolution is generating some investor ‘hallucinations’ too

Stay informed with free updates

Simply sign up to the Big tech myFT Digest — delivered directly to your inbox.

The writer is founder of Sifted, an FT-backed site about European start-ups

A brassy new venture capital investor, going by the ugly acronym Mang, has been making a lot of noise in Silicon Valley. Microsoft, Amazon, Nvidia and Google, four of the five biggest US companies by market value, have become the dominant investors in the hottest artificial intelligence start-ups. In doing so, they have been increasingly muscling out the traditional Sand Hill Road VC merchants, who like to think they know best how to fund the future.

Of the fearsome foursome, Microsoft has made the biggest bets, investing $13bn in OpenAI and also backing Inflection AI. Amazon and Google have both invested heavily in Anthropic, founded by former OpenAI members. And Nvidia has backed Inflection AI, Databricks and Cohere and co-invested with Google in Hugging Face and Runway.

Apoorv Agrawal, a partner at investment firm Altimeter, calculates that the Mangs participated in data and AI start-up investment deals worth $23bn last year — about 30 per cent of the total. However, as Agrawal notes, for the Mangs these investments are as much strategic as financial.

This financial flexing seems to be rattling some old-school VCs. Benchmark’s voluble partner Bill Gurley last week took to X to suggest that the Mangs had found a new way to “goose [their] revenues”. “Expect a massive mess in the end,” he predicted.

As Gurley’s comments suggest, these investment deals are far murkier than they first appear. Although its stated intent is to accelerate innovation and competition, Microsoft’s tie-up with OpenAI has already attracted the scrutiny of regulators as to whether it constitutes a merger. This further concentration of corporate power risks creating a seemingly impregnable intellectual capital and computing complex in AI.

In one sense, there is nothing new in big companies investing in start-ups. Indeed, VC businesses have been doing so for years. In the past, tech companies including Cisco, Intel and Palantir have been particularly active at that game. In 2009, Google launched Google Ventures as a standalone VC. Renamed GV, it now boasts more than 400 companies in its portfolio and has some $8bn in assets under management.

However, the latest slew of Mang investments are about more than pure financial calculation and do not always involve direct equity participation. The intent is also to access technological expertise, acquire customers and stymie competitors. Microsoft has embedded OpenAI’s technology into its own software services. As big cloud computing vendors, Microsoft, Amazon and Google have also thrown in computing credits instead of cash in some cases. For its deals, Nvidia, the chip company, has been offering scarce graphics processing units, essential for running the latest AI models.

In an earlier era, the big tech companies might simply have bought these start-ups outright. As Agrawal notes, they are hardly short of cash: the Mangs collectively generated $276bn of operating profits in 2023 and spent $108bn on capital expenditure. However, Washington’s antitrust advocates want to restrain big tech bloat.

That suggests these enormous investments may yet be seen as a regulatory dodge. The big tech companies are in effect locking in their customers while juicing their own revenues. Regulators are already investigating whether some cloud computing companies are operating like Hotel California, imposing punitive “egress” fees on customers to prevent them from checking out.

Early VC investors in Mang-entangled AI start-ups may welcome the additional muscle — and valuations — the big tech companies provide. But they may also question how they are going to exit their investments given they are so tied in to the Mangs. Or whether they are going to be sucked dry. 

For the moment, the excitement around generative AI is still enticing leading researchers to quit big tech companies to set up on their own, knowing they can easily attract VC cash. One example is France’s Mistral, founded by three former employees of Google and Meta, which was the talk of Davos last week. And hyped-up VCs are still frantically chasing AI start-ups aiming to apply the technology to different fields. “We’ve yet to invent a great technology without going through a hype cycle,” Albert Wenger, managing partner at Union Square Ventures, tells me. 

The Mangs are set to dominate the commanding heights of the AI economy but investors believe there are still opportunities aplenty in the sunny lowlands. Only when the bubble bursts will we know how far they have all been hallucinating.