Start

The Trillion Dollar Question

Somewhere in the Nevada desert, a datacenter the size of a small town is drawing enough electricity to power Birmingham. It is one of dozens being built across America right now, each one a cathedral of silicon and cooling water, erected in the name of artificial intelligence. The price tag for this construction programme? North of 1.6 trillion dollars, committed between 2023 and 2026 by a handful of technology companies who believe they are building the infrastructure of the future.

The Guardian published a striking piece this weekend that laid out the AI boom in six charts. I’d encourage you to read it, because the numbers deserve a moment. The S&P 500 has risen nearly 80% in five years, driven overwhelmingly by a small cluster of AI-adjacent stocks. Just 41 companies now account for almost half the index’s total market value. Datacenter investment alone was responsible for 92% of American GDP growth in the first half of 2025. SpaceX is seeking a valuation of 1.77 trillion dollars. Anthropic, the company behind Claude, has filed for an IPO.

I find those figures extraordinary. I also find them, if I’m honest, quite frightening.

The question I keep coming back to is this: where is the money coming back? To earn even a modest 10% return on 1.6 trillion dollars of infrastructure, the industry needs to generate hundreds of billions in new annual revenue. It is not generating anything close. Surveys suggest that up to 95% of early enterprise adopters report modest or low financial returns on their AI investments. Jim Bianco of Bianco Research put it bluntly: “We are building a massive financial edifice on a single technological premise, and if the cash flows do not materialise quickly, the correction will be historic.”

If you have been around long enough, you will have seen this before. The railways of the nineteenth century transformed the world, but not before they bankrupted half the companies that built them. The dotcom fibre-optic boom of the late 1990s laid the cables that eventually gave us Netflix and Zoom, but the firms that laid them mostly went bust first. The infrastructure outlives the hype cycle. The investors who funded the hype often do not.

Closer to home, British AI companies have attracted 8.3 billion pounds in funding, and 75% of UK financial services firms report using AI in some form. But only 31% report a positive return on investment. That gap between spending and measurable outcomes cost UK businesses an estimated 78 billion pounds in unrealised value in 2025 alone. I have sat in enough boardrooms to know that those numbers will eventually prompt some uncomfortable conversations.

So what does any of this mean for mortgage professionals?

It means we should pay attention without losing our heads. Over the past two years I have watched vendors pitch AI-powered CRMs, AI sourcing tools, AI compliance platforms, and AI customer journeys. Some of them are genuinely useful. Some of them are a chatbot bolted onto a database with a premium price tag. If you have sat through one of these pitches, you know the difference.

And that difference will matter, because we are entering a period where technology spending in financial services will face serious scrutiny. If the broader AI market corrects, and it may well do, the ripple effects will reach every firm that signed up for an AI subscription without asking hard questions about what it actually delivers. The firms that will come through are the ones that can point to specific, measurable improvements: faster processing times, better compliance evidence, higher client satisfaction, more cases completed per adviser. The firms that will struggle are the ones that bought the story rather than the product.

What I keep coming back to in my own work is something more fundamental than any of this. The value of a mortgage adviser does not live in the technology they use. It lives in the conversation they have with a client who is nervous about their first purchase, or confused about remortgaging, or trying to work out how to help their children onto the ladder while planning for their own retirement. That conversation requires knowledge, empathy, and judgment. I have not yet seen a model that can replicate it.

Technology should make that conversation easier to have. It should reduce the admin burden so the adviser has more time to sit with the client. It should surface the right products faster. It should create an audit trail that keeps the regulator satisfied without burying the adviser in paperwork. When technology does those things well, it earns its place. When it does not, it is overhead dressed up as innovation.

The trillion-dollar bet being placed in the Nevada desert may well pay off in the long run. I suspect the infrastructure will find its purpose eventually, as it always has. But between now and then, there will be a reckoning for businesses that spent heavily on promise and lightly on proof. In mortgage and financial services, the winners will be the firms that kept their eyes on the client, used technology where it genuinely helped, and did not mistake a vendor’s slide deck for a strategy.

I think about this a lot. The AI boom is real, and I would not bet against it over a twenty-year horizon. But what I would ask is whether we are building on rock or sand. For those of us whose job is to advise real people on the biggest financial decision of their lives, I think the answer comes down to something technology cannot manufacture. Build on what you know works. The relationship. The advice. The trust. Let the technology earn its place around it.

Research notes