By Sophie Lindqvist

The S&P 500 will trade 12% below its current level by the end of 2027, according to Capital Economics — a forecast that would leave the index 17% below its early-October peak. The call rests on a single, contestable premise: that AI investment has outrun the revenue needed to sustain it.

In a note on Friday, Chief Economic Adviser Vicky Redwood argued AI retains genuine potential to lift productivity, but she is bearish on how much revenue the technology can actually generate for external customers. That distinction — potential versus monetizable demand — is where the entire boom-or-bust debate now turns.

What does the spending actually amount to?

The numbers behind the boom are, on their face, extraordinary:

• A handful of hyperscalers (the largest cloud and computing providers) are expected to spend $800 billion this year, double the level of a year ago.
• Estimated investment reaches $1 trillion or more in 2027.
• Many forecasts assume annual capital expenditures stay above $1 trillion for several years.

Redwood's arithmetic stress-tests that assumption. Annual AI-related revenues earned from external customers would have to grow tenfold by 2030 to justify the spending. That is not impossible — but the note is careful to note what it would imply: a far more severe and rapid disruption to the economy than anything currently visible, including major displacement of existing businesses and labor.

In other words, the bull case for sustained capex requires an economic transformation on a timeline no serious observer has yet documented. The revenue assumptions are doing heavy lifting; the observable disruption to date is not.

What does this change for the economy?

Redwood's baseline is a slowdown, not a collapse. She expects AI investment to keep rising next year but at a slower pace, then fall outright in 2028 by around 10%.

The GDP arithmetic is modest by her own estimate. Higher AI investment added 0.3 percentage point to U.S. GDP growth last year, so a capex slowdown should trim roughly that much from the recent growth rate. A 10% capex drop might shave about 0.1% off GDP.

That is a slight haircut — especially compared with more AI-dependent economies in Asia. Redwood projects Taiwan's real GDP growth next year at around 3 percentage points lower, and South Korea's about 1 point lower, with a further drag in 2028. The exposure, as ever, sits with the suppliers, not the spenders.

Equity markets, she argues, will price the turn before it shows up in the real economy. "We expect considerably larger falls in tech-heavy indices," Redwood said. She also flagged a financing risk that sits outside the AI story itself: "A tightening of financial conditions driven by factors such as higher energy prices or term premia (the extra yield investors demand for holding long-dated bonds) could also make financing investment hard."

Lower stock prices then feed back into the economy through the wealth effect — the link between portfolio values and consumer spending. The market's rally over the past year lifted GDP by around 0.4%, Redwood calculated, so a selloff of comparable scale should produce an economic drag of similar magnitude. Roughly symmetric, in her framing; asymmetric risks are not addressed.

Is this a dotcom rerun?

Redwood's own answer is no, though her caveat deserves attention. "The bottom line for the U.S. economy is that we expect slower GDP growth in 2028, but not a major downturn," she wrote. "Admittedly, recessions have sometimes followed previous capex booms and busts, including in the early 2000s after the dotcom bust. However, a fall in capital spending alone has generally not been sufficient to cause a recession."

That last claim is doing quiet work. It says capital-spending busts have not historically caused recessions on their own — it does not say they cannot, nor does it model what happens if the equity selloff, the wealth effect and tighter financial conditions arrive together. Readers should note the forecast treats these channels additively rather than interactively.

There are structural reasons, in her view, why spending will not fall off a cliff. Capex could remain relatively high even after declining because hyperscalers must keep spending to replace depreciating equipment, exploit economies of scale, and stay ahead in what she calls the "winner-takes-most" AI race. They could also spread any investment declines over several years rather than take them all at once.

What survives the correction?

The productivity story, Redwood insists, outlives the capex cycle. AI will keep working its way into the economy, helped by infrastructure already built and new investment still on the way, albeit at a slower rate. She estimates AI will raise U.S. productivity by about 1% per year for several years — a meaningful cumulative gain if it materializes, though one that remains an estimate rather than a measured trend.

She does not expect an abrupt collapse in AI investment or a severe downturn when the boom ends. "Rather, we expect investment to adjust to a more sustainable pace, with sufficient capacity still being added to support further growth in AI usage," Redwood concluded.

The forecast, then, is a controlled deceleration: stocks down double digits, GDP growth trimmed, tech indices hit hardest. The promise of a soft landing is explicit. What the note does not say is what happens if the tenfold revenue expansion fails to appear and the hyperscalers' spreadsheets — not their strategists — force the adjustment instead.