AI Capex Turns Into a Rates Problem for Equities

A clean beat beneath a flat headline

The headline durable-goods number looked muted. The Census Bureau reported August durable-goods new orders at $338.6 billion, essentially unchanged on the month.¹ But the mix mattered more than the headline.

Orders for nondefense capital goods excluding aircraft — the category investors watch as a proxy for underlying business equipment spending — rose 1.6%, driving the upside surprise highlighted by Reuters and fixed-income analysts.¹³

That distinction matters because aircraft orders are volatile and often obscure the business-spending signal. Haver Analytics noted that core capital-goods shipments rose 0.6%, a sign that stronger orders are translating into current-quarter activity rather than remaining only in the pipeline.⁸ First Trust also pointed to the implications for third-quarter business investment, emphasizing the role of data-center-related capital spending in the durable-goods internals.⁹

In other words, the August report was not just a statistical beat. It reinforced the view that the AI infrastructure buildout is providing a measurable offset to higher borrowing costs. Data centers, power equipment, electrical components, cooling systems and related supply chains are helping sustain manufacturing demand at a point in the cycle when many investors expected restrictive policy to bite more visibly.

Why the bond market cared

Treasury investors responded less to the flat headline than to the growth signal in the core data. Reuters reported that yields extended their recent run higher after the durable-goods release, with long maturities setting fresh multi-decade highs.⁴ A separate Reuters report framed the move as part of a broader global bond selloff driven by resilient growth, inflation concerns and rising expectations for Fed rates.⁵

The official rates backdrop was already elevated. The Federal Reserve’s H.15 release showed the 10-year Treasury constant-maturity yield at 5.18% and the 30-year at 5.47% for September 24, before the September 25 market reports of further pressure at the long end.² Market commentary the same day put the 10-year near 5.20% and the 30-year close to 5.5%, levels that directly compete with equity expected returns.¹¹¹²

This is the central tension: AI-led capex improves the growth outlook, but stronger growth reduces the urgency for policy easing and can lift the term premium investors demand to hold long bonds. If AI investment is large enough to keep the economy running above the speed implied by restrictive policy, the rates market will price less relief from the Fed and more compensation for duration risk.

The equity valuation trade-off

For equities, the August capex data cut both ways.

The positive read is that AI infrastructure spending remains broad enough to support revenue for semiconductor, networking, electrical equipment, power, engineering and industrial automation firms. It also helps explain why equity indexes have shown resilience despite 5% Treasury yields. Schwab’s market update noted that stocks were still tracking a weekly rise even as the 10-year yield hovered near 19-year highs.⁷

The negative read is valuation math. When long-dated Treasury yields move toward 5.5%, the discount rate applied to future cash flows rises. That matters most for long-duration equities, including many AI beneficiaries whose valuations depend on profit growth years into the future. The same capex cycle that validates the AI revenue story can erode the multiple investors are willing to pay for it.

That is why the market debate is shifting from whether AI is real to whether AI is large enough to overpower the rates drag. MarketScreener’s same-day commentary captured that framing directly, describing the contest between AI-led equity momentum and 5%-plus Treasury yields.¹²

The answer may vary sharply by sector. Companies selling into the physical AI buildout may continue to see earnings upgrades, while broader growth stocks face a tougher hurdle as risk-free yields rise.

AI spending is becoming macro-relevant

Independent analysis of the Census data also supports the idea that AI-related spending is no longer confined to a narrow technology narrative. Wolf Street linked the 1.6% monthly rise and 14.1% year-over-year increase in core capital-goods orders to AI data-center supply chains and argued that the boom is feeding demand — and potentially inflation pressure — across industrial categories.¹⁰

That matters for rates because the Fed is not setting policy on AI enthusiasm. It is responding to realized activity, inflation pressure and financial conditions. If AI capex keeps equipment spending firm, supports corporate hiring and tightens resource constraints in power and construction markets, policymakers may see less need to ease even if other parts of the economy slow.

Reuters also noted risks from rising oil prices, interest rates and long-term Treasury yields, underscoring the feedback loop now facing markets.³ Strong activity keeps yields elevated. Elevated yields threaten valuations and financing conditions. But AI investment may be strong enough to delay the slowdown that would normally bring yields back down.

The investor takeaway

The August durable-goods report strengthens the case that AI infrastructure is cushioning the US economy from high rates. But that is not an unambiguous bullish signal for equities. The same investment surge that supports earnings is also helping keep Fed expectations and long-end Treasury yields elevated.

For global investors, the key question is whether AI-driven earnings revisions can outrun the repricing of discount rates. If the answer is yes, leadership can remain concentrated in firms with direct exposure to the infrastructure buildout. If the answer is no, 5%-plus long bonds will continue to pressure equity multiples, especially in the most duration-sensitive parts of the market.

August’s capex surprise suggests the AI boom is real in the economic data. The bond market’s response shows it is also real enough to make the equity valuation debate harder.