The AI Bubble And What It Means for The Workplace

When Softbank sells its entire Nvidia holding, JP Morgan's Jamie Dimon warns that tech stocks are overheated, and Jeff Bezos admits there's an AI bubble, perhaps we should pay attention. That said, Big Tech firms are collectively spending nearly $400 billion on AI infrastructure this year alone, 30% of European workers now use AI daily, and US workplace AI usage rose from 40% to 45% between Q2 and Q3 2025. At the same time, Irish workers seem reluctant — only 10% use AI tools daily versus 14% globally. Adoption rates tell a mixed story. So are we witnessing genuine workplace transformation? Or is this a cycle of speculation that will inevitably collapse, leaving ordinary working people out of work and out of pocket?

Bubbles Too Easily Forgotten

Remember the last speculative bubble? If you were around when it popped in 2008/9, you'll know what comes with a collapse. It was a bizarre time. All of a sudden, the roads were empty and the airports were full. Construction equipment filled auction yards while jobs disappeared like fairy dust. Some people handed keys to their houses back to the banks and left. Others who couldn't cope were found on the end of ropes. It took until early 2010 for the penny to drop for me. One by one, I let my staff go and closed the doors. I regard myself as one of the fortunate ones.

We were collectively consumed by the hype machine, convinced that the party would never end. This time, I get the same uncomfortable feeling. Goldman Sachs estimates AI capital expenditure will reach $390 billion this year. AI stocks have contributed 75% of the S&P 500's returns and 90% of capital spending since ChatGPT launched in 2022. The heads of JP Morgan, Morgan Stanley, Goldman Sachs, and The Bank of England have all warned of an impending correction. When beneficiaries of a boom start hedging their bets, maybe we should take heed.

The Circular Investment Problem

Part of what's fuelling AI concerns is the tight circular nature of AI investments. Microsoft's valuation has been rising on belief that OpenAI will become profitable despite being deep in the hole — while simultaneously hedging with Anthropic. Nvidia's revenue boom stems from the broader scramble to build AI infrastructure. Oracle has borrowed up to its oxters ($300 billion according to Bloomberg) betting that data centre leasing will double the value of the company. The big players are investing in the smaller players, who are handing it back to the bigger players, and everyone is excited about everyone's growth. When you look at who's funding who, all this growth looks like financial incest.

Compare this gap between debt and profitability to the dot-com bubble. The top tech companies of the time traded at 70 times forward earnings in 2000. Today's AI giants sit around 26 times earnings — not cheap, but maybe not the same inflated valuation. The difference, AI evangelists argue, is that today's Big Tech generates real revenue from real customers.

Profitability and productivity aren't the same thing, though. AI might be profitable for Nvidia at the moment without necessarily being productive for organisations buying it. Research from MIT earlier this year found that 95% of businesses implementing AI reported zero measurable value. That should raise eyebrows. An Upwork study found AI agents from OpenAI, Google DeepMind, and Anthropic failed to complete many straightforward workplace tasks autonomously.

What's Happening at Work

From what I see training business people and educators on AI, whether or not stocks are overvalued is somewhat irrelevant at the ground level. I work at the intersection of people and technology, and I see an appetite for learning generative AI — to improve workflows and create efficiencies. And I believe Gen AI is capable of delivering on that if people know how to use it, if they can develop a particular structural way of thinking.

The hype machine suggests that you just type a one-shot prompt into ChatGPT or Claude and your problem is solved. But that's not how it works. To get the best from AI tools requires strategic thought and the ability to see where the gaps are. The AI needs humans to give it structure and direction — otherwise it gives you inaccuracies at best, gibberish at worst.

Even if this bubble bursts, the genie isn't going back in the bottle. Much like the internet in the 2000s, this technology isn't going away. These tools are good, if we can accept that people are the key to digital transformation, not the technology itself. The tech is agnostic — it's how organisations test, plan, train, and implement that will ultimately count. The businesses that get this right will thrive regardless of what the markets do.

The danger isn't artificial intelligence itself. The danger is treating socio-technical transformation as merely a financial opportunity.

Even if this bubble bursts, the genie isn't going back in the bottle. The businesses that get this right will thrive regardless of what the markets do.

Where Do We Go From Here?

If investment flows into learning and development, up-skilling workers and digitally transforming workplaces, then a correction might hurt investors without devastating workplaces. But if spending is primarily speculative, chasing short-term returns through cost-cutting automation, the bursting bubble could leave organisations with expensive technology and a workforce that's been deskilled or displaced.

McKinsey estimates AI automation could unlock $2.9 trillion in US economic value by 2030. But value for whom? Distributed how? The danger isn't artificial intelligence itself. The danger is treating socio-technical transformation as merely a financial opportunity.

Bubbles will come and go, but what work means to human beings continues to matter beyond trend and hacks. Productivity for productivity's sake is a farce. It allows organisations to think of people as machines, objects to be manipulated and, perhaps, even replaced altogether. Whether this bubble bursts or deflates gradually, the real work is building systems that treat humans as more than resources to be optimised for shareholder benefit. If your work doesn't give you that, build your own thing.