MIT paper states fewer than one-in-ten firms see return on AI investment
The globally renowned Masachusetts Institute of Technology has published a new paper, which has taken the hype around artificial intelligence to task in a major way. Amid the research’s findings that only 5% of companies investing in the technology actually make a profit from it.
2025 seems to be the year in which sentiment around artificial intelligence begins to turn – both among researchers and investors. Various organisations have excitedly published research over the past three years, breathlessly anticipating a colossal AI-generated boom. But while that has continued in certain quarters – Accenture recently claimed AI could be used by the UK to put its productivity on par with the US – specific landmark figures (such as McKinsey & Company’s claim AI would lead to an annual productivity boost worth $4.4 trillion) have been phased out from headline findings.
That’s not all, though. While the consulting sector – which has broadly hitched its wagon to AI, expecting to be able to sell implementation services as it did with other digital hypes, like the metaverse and cryptocurrency – might still broadly be pushing the idea that large language models will come good in the end, other noted experts have been more inclined to point out the gaps between its alleged promise, and the reality it delivers.
MIT study
The Masachusetts Institute of Technology (MIT) was famously associated with ‘Artificial Intelligence, Scientific Discovery, and Product Innovation’, a headline-making study published in December as a pre-print by an MIT graduate student in economics, Aidan Toner-Rodgers. The paper claimed that AI would accelerate the speed of science at meteoric rates. For example, figures which were focused on in reports from The Wall Street Journal, Nature and The Atlantic alleged that AI-assisted researchers “discover 44% more materials, resulting in a 39% increase in patent filings and a 17% rise in downstream product innovation.”
But MIT later – even more famously – disavowed the paper. A press release from the institution admitted that following an internal investigation, it had “no confidence in the provenance, reliability or validity of the data and has no confidence in the veracity of the research contained in the paper." While no official reason for the change was cited then, The Wall Street Journal reported that an unnamed computer scientist "with experience in materials science" had raised questions about how the AI in the study actually worked, and "how a lab he wasn’t aware of had experienced gains in innovation." When MIT experts who had praised the paper were unable to get to the bottom of those questions, they took their concerns to senior management.
Now, MIT has followed this up with a report which appears to have shaken the foundation of the AI sector: investor sentiment. The majority of AI organisations do not turn a profit – with their overheads vastly outweighing their income – but until now, many have been able to rely on private equity and other technology giants like Microsoft and Meta shovelling capital into their furnaces. But after a report from MIT’s NANDA (Networked Agents and Decentralised AI) initiative found 95% of enterprise organisations have seen no return at all from their AI efforts, the market saw a sizeable hiccup.
Investor enthusiasm for AI wavered the day after, as chip-giant Nvidia’s shares dropped 3.5% and Palantir nearly 10%. And when OpenAI’s Sam Altman also warned of a potential bubble in AI, the Nasdaq also declined more than 1.2% the following morning.
MIT’s study did find that 5% of organisations had successfully integrated AI tools into production at scale. In those cases, firms are said to be “extracting millions in value,” with Aditya Challapally, the MIT researcher who led the study, telling Fortune that some large companies and younger startups are "excelling" with AI because "they pick one pain point, execute well, and partner smartly with companies who use their tools." That has led some startups led by young founders to see revenue "jump from zero to $20 million in a year".
No return at all
The report is based on 52 structured interviews with enterprise leaders and on analysis of more than 300 public AI initiatives and announcements, and a survey of 153 business professionals. Assuming the professionals were each from a different firm, just over 7 per every 153 businesses is hardly a glowing recommendation though – and while a small number may have made $20 million from AI investments, many more will feel they have incinerated larger amounts for technology that was never suited to their operations in the first place. As such, members of the C-suite are beginning to overcome the FOMO which led them to sink so much capital into AI in the first place.
Illustrating this, the MIT study quotes an unidentified COO at a mid-market manufacturing firm, "The hype on LinkedIn says everything has changed, but in our operations, nothing fundamental has shifted. We’re processing some contracts faster, but that’s all that has changed…”
Another anonymous CIO told the authors, “We’ve seen dozens of demos this year. Maybe one or two are genuinely useful. The rest are wrappers or science projects.”
This comes at a time when the scepticism of investors has already seen a number of key changes to spending at the world’s leading technology firms. In particular, Microsoft – which has bankrolled a lot of OpenAI’s expansions – abandoned data centre projects set to use 2 gigawatts of electricity in the US and Europe in the six months leading to March 2025, due to an oversupply relative to its current demand forecast. The tech giant’s withdrawal from new capacity leasing was said to be largely led by the decision not to support additional training workloads from ChatGPT maker OpenAI.
This pull-back from private capital may explain why many AI leaders are pivoting to state spenders instead. A recent leak in the UK suggested that the government might be considering paying OpenAI £2 billion pounds, to give the whole country access to ChatGPT’s premium model. With nation states less concerned with ‘profit’, and looking for digital tools which could excuse their cutting of genuinely useful social services – from the Citizens Advice Bureau, to health consultation services – AI leaders may also have identified this as the next source of the huge amounts of cash needed to keep them afloat.
