When Do High Bond Yields Crush the Stock Market? Four Transmission Channels Unpacked
nashnova research
The U.S. 10-year Treasury yield has climbed to 4.85%, and multiple institutions warn that a decisive break above 5% could trigger an equity sell-off — yet the AI construction boom is blunting the traditional rate-transmission mechanism. The key variable is time.
How close are yields to "breaking" equities?
The 10-year U.S. Treasury yield has reached 4.85%, just basis points from its 2023 high and the 5% round number.
UBS Holt's Michel Lerner notes that crises triggered by bond vigilantes — investors who dump government debt to force fiscal discipline — have historically been almost impossible to contain.
TS Lombard's Freya Beamish and Davide Oneglia argue the 10-year yield needs to reach at least 5% to compensate for the current macro environment; the ongoing compression of the U.S. equity risk premium is unsustainable.
This means → the market is not debating *whether* stocks will fall, but when yields get high enough to flip capital flows.
What are the four transmission channels?
Channel 1: Higher rates raise borrowing costs → corporate profits decline → stock prices fall.
Channel 2: A higher discount rate — the rate used to convert future cash into today's value — shrinks the present value of future earnings → growth stocks are hit first.
Channel 3: Bonds become more attractive → capital migrates from equities to fixed income, draining the stock market.
Channel 4: High rates force governments to cut deficits → historically, deficit contraction has tracked falling corporate profits.
Why hasn't the classic "rate hikes kill the economy" script worked this time?
Channel 1 has traditionally run through housing: rates rise, mortgages get expensive, home-buying drops, and construction, renovation, and transport cool in sequence. Economist Edward Leamer summed it up: "Housing is the business cycle."
Housing is indeed weak right now: new-home sales are tepid, existing-home volumes sit at historic lows, and residential investment as a share of GDP keeps sliding.
Yet the broader economy is still humming. In plain terms = under the old playbook, a cold housing market should slow the whole economy. This time it hasn't.
Why has AI construction filled the gap?
Barclays' Ajay Rajadhyaksha points out that U.S. data-center construction spending reached $85 billion in 2026, nearly double the $45 billion of two years prior; 74 new facilities have broken ground this year across 28 states.
Hyperscalers — the mega cloud platforms like Microsoft, Amazon, and Google — employ roofers, electricians, plumbers, steelworkers, and concrete crews at scale, absorbing exactly the labor pool freed up by the housing downturn.
This means → AI construction has physically replaced housing investment as the economy's cyclical engine, temporarily disabling the traditional rate-transmission mechanism.
How long can this "rate immunity" last?
Rajadhyaksha notes that hyperscalers see AI as an existential race and are nearly price-insensitive — that is the root cause of the transmission breakdown.
But data centers are not large-scale employers once built; when compute-construction demand plateaus, traditional cyclical forces will reassert themselves.
The more critical variable: when AI return-on-investment starts to govern further spending decisions. This means → the moment companies begin asking "how much does each dollar of AI spend earn back," the discount rate becomes lethal again.
Who feels the pressure first?
Rajadhyaksha suggests that for financially or technically weaker AI participants — such as Oracle — the pressure to demonstrate returns may already be materializing.
Ruchir Sharma, writing in the *Financial Times*, notes that across three hundred years of history, every major bubble ended with a sharp rise in core corporate financing costs.
In plain terms = high yields and high stock prices coexisting is not a perpetual-motion machine — it runs on AI companies' willingness to spend without counting. Once that impulse cools, all four transmission channels activate at once.
市场有风险，内容仅供研究参考，不构成投资建议。