Goldman Backs AI as Cloud Spend Heads to $1.2T by 2027

Goldman Sachs is staying bullish on AI-related equities even as long-dated government bond yields climb to their highest levels in roughly two decades. Timothy Moe, the bank's head of Asia-Pacific equity strategy, said hyperscale cloud companies are expected to invest around $800 billion this year, with that figure potentially rising to about $1.2 trillion by 2027. The thesis rests on two pillars: a spending signal that underpins Asia's AI hardware supply chain, and regional valuations near historic lows.

Context — why the AI trade is being tested by rising yields

Rising bond yields raise the discount rate applied to future corporate earnings. That mechanically pressures richly valued growth sectors, and AI-linked equities sit at the centre of that pressure. The report notes long-dated Treasury yields have climbed to their highest levels in around two decades this week.

The comparable the report supplies is the scale of cloud capital expenditure itself. Hyperscale spending of about $800 billion this year, against roughly $1.2 trillion by 2027, is the demand signal Goldman treats as decisive for Asian hardware makers. It is not a price level or a valuation target, but a capex trajectory that runs through the region's semiconductor and server supply chain.

What changed to trigger the note now is the bond sell-off. It has tested investor appetite for growth names, and Goldman is answering that test directly. Moe argued corporate earnings growth should buffer the high-rate environment rather than amplify it.

The geographic timing matters too. Taiwan and South Korea, the report notes, are shut for a holiday today, along with mainland China. That leaves the region's AI hardware markets without live price discovery while the yield story plays out elsewhere. Investors in those markets are effectively waiting on two variables at once: cloud spending plans holding up, and bond yields stabilising.

The note also lands against a broader shift in how major banks frame the AI trade. Morgan Stanley has a barbell call, and UBS holds a constructive stance, per the report. Together they add to a consensus that the cycle is not over.

Data — the numbers behind Goldman's AI call

The headline figures are the capex numbers. Around $800 billion of hyperscale cloud investment is expected this year, potentially rising to about $1.2 trillion by 2027. That is a rise of roughly 50% across the two-year window, and it is the single most important number for Asia's hardware supply chain per the report.

Valuation is the second number. Asian equities trade at roughly 10 times earnings overall, which Moe described as extremely low and near the bottom of their historical range. That compares with markets where valuations are already stretched, giving the region what the report frames as more room to absorb rate pressure.

The rate backdrop is the third. Long-dated Treasury yields have reached their highest levels in around two decades this week. The report does not specify a yield level, so the precise figure is not available here.

Metric	Figure
Hyperscale cloud capex, this year	~$800bn
Hyperscale cloud capex, 2027	~$1.2tn
Asian equities P/E	~10x
Long-dated Treasury yields	highest in ~2 decades

Before and after, in capex terms: the jump from roughly $800 billion to about $1.2 trillion by 2027 is what Goldman reads as a demand commitment, not a cyclical spike. It is the difference between a single-year budget and a multi-year build-out.

Analysis — which sectors carry the risk and the reward

Asia hosts much of the AI hardware supply chain, so the capex trajectory flows through foundries, server assemblers, memory makers and the component suppliers behind them. The report does not name individual tickers, so exposure is described at the sector level rather than the company level.

The valuation argument is the mechanism that matters. At roughly 10 times earnings, Asian equities start from a lower base than markets where multiples are already stretched. A lower starting multiple means the same rise in the discount rate removes less absolute value from the equity, which is why Goldman treats cheapness as a buffer rather than a signal of distress.

The counter-argument is straightforward and the report states it. The risk to the thesis is that bond yields keep climbing and eventually force the biggest spenders to rethink the pace of their investment. If capex plans slip, the demand signal that supports Asian hardware weakens at the same time as the discount rate rises. That is the scenario where both pillars fail together.

Near-term positioning reflects that tension. Moe expects markets to stay somewhat volatile into the US midterms in early November, with high energy prices and geopolitical risks adding pressure. After that period, he sees scope for a rally before year-end, driven by earnings growth and a recovery in valuations. The implied flow is investors buying dips rather than cutting exposure on rate-driven sell-offs.

Outlook — what to watch next

Three catalysts sit on the calendar. The US midterm elections in early November are the volatility window Moe identifies. A White House AI meeting on Tuesday is the near-term policy event. And a week without Chinese market participation, tied to the holiday closures in mainland China, Taiwan and South Korea, removes a major source of liquidity and price discovery.

On levels, the report names none for yields or for Asian equity indices, so no specific thresholds are set here. The observable conditions are directional instead: whether long-dated Treasury yields extend their climb or stabilise, and whether hyperscale capex guidance holds at the roughly $800 billion figure for this year.

A year-end rally, in Moe's framing, is conditional on earnings growth materialising and valuations recovering once the midterm window passes. It is a condition, not a forecast.

Frequently Asked Questions

What does Goldman's $1.2 trillion cloud spending forecast mean for AI hardware stocks?

It is a demand signal. Hyperscale cloud companies buying servers, memory and accelerators is what generates revenue for Asia's hardware supply chain. A trajectory from roughly $800 billion this year to about $1.2 trillion by 2027 implies multi-year order visibility rather than a single-year budget cycle. Goldman's argument is that this commitment outweighs the pressure higher bond yields place on growth-stock valuations.

Why do rising bond yields hurt AI stocks specifically?

Higher yields raise the discount rate applied to future earnings. Companies whose value sits mostly in distant cash flows, which describes many AI and growth names, lose more present value from that shift than companies with near-term earnings. The report notes long-dated Treasury yields have reached their highest levels in around two decades this week, which is what makes the discount-rate argument live.

Are Asian equities cheap enough to offset the rate pressure?

Goldman's Moe frames it that way. Asian equities trade at roughly 10 times earnings overall, which he described as extremely low and near the bottom of their historical range. A lower starting multiple means a rate-driven de-rating removes less absolute value. The report does not give a target multiple, and it flags that continued yield increases could force cloud spenders to slow investment.

Bottom Line

Goldman is betting that a $1.2 trillion cloud capex path and 10x Asian multiples outweigh the highest long-dated yields in two decades.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. CFD trading carries high risk of capital loss.