The AI Bubble That No One Is Talking About..

Everyone is talking about the AI boom. I want to talk about what’s hiding inside it.

Because buried beneath the extraordinary headline numbers which can’t be ignored, the record investment, the soaring valuations, the breathless vendor announcements, there is a reality building that nobody in the industry seems willing to name directly.

I will.

$1.63 Trillion of Committed Spend. And Counting.

Bob Evans at Cloud Wars has been tracking something remarkable, he is an incredible source of the commercial reality of the AI world, kudos to him.

Having worked in Data & AI for a while now however, on the coalface of real world application of AI to tackle business pains, there is an inflection point that doesn’t seem to be trending towards a happy place.

Microsoft, Google Cloud, AWS and Oracle, have accumulated a combined RPO and backlog of $1.63 trillion. That is contracted revenue not yet recognised. Money that companies have committed to spend on cloud and AI services, sitting on the books of the world’s largest technology companies, waiting to become real.

To put that in perspective: Evans notes this figure is almost certain to approach $2 trillion next year. Oracle’s RPO grew 359% year-on-year. Google Cloud’s backlog grew 82%. Microsoft’s sits at $392 billion alone.

These are not projections or estimates. These are contractual commitments. Real money, already signed.

The question nobody is asking loudly enough is: what happens if the businesses that signed those contracts can’t actually use what they’ve committed to?

The Adoption Gap No One Wants to Admit

Here is what we know about AI adoption inside the enterprise right now.

Microsoft has 450 million commercial Microsoft 365 seats. Copilot, its flagship AI product, has penetrated approximately 3% of them. Roughly 15 million paying users out of a potential base of 450 million.

The rest, just sitting there, collecting dust.

Microsoft’s own CEO has acknowledged that Copilot integrations “don’t really work” as intended. Carnegie Mellon researchers found that AI agents fail to complete real-world office tasks 70% of the time. Microsoft has quietly slashed its Copilot sales targets by up to 50%.

Gartner’s most recent data is equally concerning: only 6% of enterprises have successfully moved generative AI projects beyond the pilot phase. McKinsey finds that only 39% of organisations report any EBIT impact from AI. And Gartner predicts that over 40% of agentic AI projects will be cancelled by the end of 2027 due to escalating costs, unclear business value and inadequate risk controls.

So on one side of the ledger: $1.63 trillion in committed spend, growing at extraordinary rates. Bookings, commitments, hopefully aligned to the possibilities of value & ROI.

But, single-digit adoption means revenue isn’t being recognised, value is left in the margins, but the cost is growing.

This is not a technology problem.

The Tactical Death Spiral

Here is where it gets particularly uncomfortable.

When a business commits to Google or Microsoft for, and those commitments are substantial, often multi-year, often nine figures, the assumption underlying that decision is that AI will generate value. That it will grow revenue, cut costs, or both. That the investment will pay for itself.

But when adoption lags, when the tools don’t embed in workflows, when employees don’t use what’s been bought, when pilots don’t convert to production, the business is still obligated for the spend. The contract doesn’t care that your change management programme stalled. The invoice arrives regardless.

What happens next is predictable, and I see it constantly. Leadership teams, under pressure to justify the commitment they’ve already made, start making tactical decisions about AI application. Not strategic ones. The question stops being “where will AI create the most commercial value?” and starts being “how do we use enough of this to feel like we’re not wasting the contract?” or “what can we direct through a market place to draw down on our commit”?

That is a fundamentally different question. And it produces fundamentally different outcomes.

You get AI deployed for its own sake rather than for a defined commercial purpose. You get activity without accountability. You get projects that look busy but don’t move the needle.

The committed spend backlog grows. The adoption gap widens. The tactical decisions multiply. And everyone, the businesses, the vendors, the boards watching from above, ends up dissatisfied.

The Vendor’s Problem Too

This matters for the hyper scalers as much as for their customers.

Microsoft, Google, Oracle and AWS need to recognise revenue from that $1.63 trillion backlog. Recognition happens when customers consume the services. But if adoption is slower than anticipated, and the data strongly suggests it is, revenue recognition lags. The backlog grows faster than it converts.

That creates its own pressure. Vendors respond by pushing adoption harder, adding more features, dropping prices, bundling more products into existing subscriptions. Which creates more noise, more complexity, more decisions for already-overwhelmed technology teams.

None of this is sustainable. A $1.63 trillion backlog built on the assumption of AI adoption, in a world where AI adoption is demonstrably lagging, is a tension that has to resolve itself one way or another. Either adoption accelerates dramatically, which requires businesses to get far more disciplined and strategic about how they deploy AI , or the commitments start getting renegotiated, the backlogs start shrinking, and the AI investment narrative takes a very different turn.

Both of those outcomes are uncomfortable, and the discussions between vendor and customer (although this has been rehearsed historically) isn’t a happy one.

The Root Cause Is Strategic, Not Technical

I want to be clear: I am not making an argument about AI. It’s been my livelihood for the last 10 years, and is able to genuinely transform an organisation.

The technology is real. The potential is real.

But the current dynamic, commit first, figure out the value later, is the root cause of everything I have described above. And it is avoidable.

The businesses that will navigate this well are not the ones with the largest AI budgets or the most advanced technology partners. They are the ones that started with a clear commercial question: which specific problems, in which specific parts of our business, will AI solve, and what will solving them be worth?

That question, asked and answered rigorously, before the contract is signed and before the first pilot is launched, is the difference between strategic AI investment and obligated spend that nobody can justify.

It is the difference between AI as a competitive advantage and AI as a line item on a balance sheet that the CFO is quietly starting to question.

At FernAI, that is the question we help leadership teams answer. Not “what AI can we use?” but “where will AI create value, and how do we prove it?”

Top down and bottom up meeting to create a beautiful relationship.

Because the bubble I am describing is not inevitable. It is a consequence of a process that starts in the wrong place.

Start with value. Everything else follows.

If this has struck a chord, if your organisation is sitting on committed AI spend and not yet sure how to turn it into demonstrable commercial return, I would welcome a conversation.

Book a call with FernAI: fernai.co.uk/contact

FernAI helps organisations move from AI activity to AI accountability. Working with C-suite leadership teams to identify, prioritise and measure the AI use-cases that deliver genuine commercial return.