Is the AI super-cycle just clearing its throat ?

Just clearing its throat?

I walked into 2026 thinking we were at the end of the beginning phase of the artificial intelligence trade. It turns out we were not even through the warm-up lap. The market is no longer debating whether AI matters. The debate is whether the payoff curve bends steeply enough to justify the most aggressive capital cycle since the post-war build-out of the interstate highway system.

The beginning phase is just starting. The game is accelerating. And the tape is now circling one central question on its probability pad. Is this a compounding flywheel or a capital sinkhole?

Start with the physical footprint. The data centre boom has moved from a thematic story to a macro variable. These are no longer industrial sheds with blinking lights. They are balance sheet devourers. Single campuses now carry price tags north of $20 billion. Construction pipelines are tilting around them. Power grids are being stress tested. Transformers are becoming strategic assets.

Electricity is the new oil of this cycle. Goldman estimates power demand tied to AI growing at 17 percent annually through 2028. That is not incremental. That is structural. The constraint is no longer imagination. It is amperage. Grid connection queues are stretching. Tier one hubs such as London, Dublin, Amsterdam and Frankfurt are congested. Capital is migrating to secondary markets across Europe as hyperscalers chase electrons the way shale drillers once chased acreage.

The United States remains the epicentre. Europe is playing catch-up with pipelines up sharply year on year. Yet here is the nuance the market is digesting. Data centres are a powerful construction impulse, but they do not immunize the broader sector from high rates, labour scarcity and stubborn input costs. This is a capital-intensive build-out layered on top of a fragile financing backdrop. If liquidity tightens, the cranes slow.

Which leads directly to the second axis of risk. Revenues versus investment.

By some estimates, nearly $2 trillion has already been deployed in this AI arms race. Big tech capex is set to exceed $630 billion this year alone. For context, the Apollo programme that put a man on the moon cost less in real terms. We are funding the equivalent of a digital moonshot every twelve months.

For now, demand for advanced chips is described as sky-high. The order books are full. But free cash flow growth at the hyperscaler level is beginning to feel the strain. Debt-funded capex is rising. The market is making a forward bet that monetization will bend sharply higher. If revenues outrun investment, the equity multiple expansion holds. If they lag, the market will start asking harder questions about return on invested capital.

This is where positioning matters. AI stocks have driven the bulk of index level performance since ChatGPT detonated onto the scene. The S & P has effectively become a referendum on silicon productivity. When three quarters of index returns are concentrated in one theme the index stops being diversified and starts being directional.

That does not mean the trade is wrong. It means it is crowded. And crowded trades do not collapse because the story fails. They wobble when the marginal buyer hesitates.

Now layer in the third dimension. Healthcare.

This is where the narrative shifts from capex spectacle to societal utility. The global healthcare system is short millions of workers. Billions lack access to essential services. Here AI moves from hype to tool. Diagnostic models that scan X rays and CT images. Retinal analysis that flags diabetic complications early. ECG pattern recognition that surfaces cardiac risk. These systems do not replace clinicians. They compress time. And in medicine time is margin.

If AI meaningfully lifts productivity in healthcare the economic dividend is profound. Fewer administrative hours. Faster triage. Earlier detection. That is not just revenue. That is cost deflation in one of the most inflation-prone sectors on earth.

The market understands this intuitively. It is why the narrative has broadened from chips to applications. From infrastructure to outcomes. The first wave was shovel sales. The second wave is operating leverage.

So where does that leave us?

We are no longer pricing a technology story. We are pricing an ecosystem. Power grids, bond markets, labour markets, hospital workflows and sovereign balance sheets are now tethered to the AI cycle. This is not a sector rotation. It is a capital reallocation across the real economy.

The key question for 2026 is not whether AI is transformative. It is whether cash flow transformation keeps pace with capital formation. If revenues begin to outrun the spend, this becomes the defining growth engine of the decade. If debt rises faster than monetization, the market will begin to circle not the opportunity but the vulnerability.

For now the tape is not circling the drain. It is circling a launch pad. But launch pads are unforgiving places. They reward execution and punish miscalculation.

The AI supercycle has only just cleared its throat. The next move will determine whether it sings in compounding harmony or strains under the weight of its own ambition.