AI, boom or bubble?

MCH market insights

On the topic of AI, I've lost count of the number of times it's been said this year that AI is in a bubble, and yet despite some sharp selloffs the market has so far managed to retain its resilience. 

It is true that the amount of money being thrown towards AI has been eye watering with little of any question about whether investors will see a return on investment, especially with all the headlines surrounding the deals being announced by OpenAI, which has seen the likes of Nvidia, Oracle, AMD, Broadcom, Amazon and Coreweave sign agreements of a combined sum total of $1.4trn.

This has inevitably raised questions as to how all of this will be financed, when all the companies involved appear to be playing a game of pass the parcel when it comes to cash investment in what looks like a high-tech variation of a Ponzi scheme. 

These deals also raise all manner of questions about how this cash will generate a longer-term return on investment, and that’s before we look at the considerations of the power requirements needed when it comes to sustaining these buildouts when the infrastructure eventually comes on line. 

That said we do appear to be starting to see some evidence that this ever growing spend is prompting a few doubts with some investors as to who are likely to be the winners and losers in all of this spending.

This year alone we’ve had large scale commitments of capex from the likes of Microsoft, Meta, Amazon and Alphabet, to the tune of $80bn, $71bn, $100bn and $92bn respectively, as they look to build out their AI offerings of CoPilot, Meta AI, Alexa, and Gemini respectively in order to compete with OpenAI’s ChatGPT. 

It was particularly noteworthy that the recent announcement from Meta that they are scaling back their investment on the Metaverse and AI by 30%, could be an early acknowledgement that appetite for AI spending could be approaching a peak, although in the case of Meta their Reality Labs has already lost $73bn over the past few years. 

Could this be a welcome shift that raises the question that companies are now starting to think about how and when all of this spending will start to see a (ROI) return on investment.

In the first half of this year spending announcements had begun to take on the look of a virility contest, however as we head into 2026, that narrative appears to be shifting. 

With questions now being posed about ROI, along with a focus on margins we may start to get a more realistic picture of who the winners and losers are likely to be, with the losers likely to be punished heavily. 

We've already had a flavour to some extent this year that signing big AI deals is no guarantee of share price outperformance with the likes of Coreweave shares which IPO’d earlier this year at $40 a share, surging to over $180 a share during the summer, and which are now back around $60.

Oracle shares have also seen a significant sell-off over doubts about their own AI build out on doubts about how their infrastructure plans are being funded, as well as a sharp rise in their debt pile. These concerns were amplified after one of their major investors Blue Owl Capital refused to back a $10bn deal for a facility based in Michigan.

This reticence appears to be down to concerns over the debt terms relating to the deal, with Oracle looking to build a $300bn site as part of its agreement with OpenAI to provide the company with 4.5GW of computing power over the next 5 years.

This concern over debt and lease terms contrasts with the likes of Amazon and Microsoft who have much greater capacity to fund their own capex from their own revenue streams, and as such are able to get better terms, with Oracle choosing to fund their own build plans by tapping debt markets.

As we head into 2026, expect more focus on how companies are financing these build out plans, in the context of their revenue and cash flow levels, as well as the levels of debt being raised, and when to expect some sort of return for the amounts of capital being used.