The issue economists and financial analysts always face when attempting to assess the impact of higher interest rates on equity valuations in general or artificial intelligence in specific is that evaluations must be made on the assumption that “all other things remain equal,” which, of course, is rarely the case. Markets are dynamic, so, by definition, all else does not remain equal.
Consider the expected impact of price-earnings ratios and higher real interest rates.
The cyclically adjusted price-earnings ratio (CAPE) being at a historically-high level suggests lower returns from equity markets over the next decade or so and a possible "reversion to mean" for AI equity values.
But the high CAPE ratio does not mean artificial intelligence public equities are necessarily overvalued or in a “bubble.” It does mean that AI firms must grow into their valuations by continued high earnings growth.
Observation | What CAPE suggests |
U.S. equities are expensive relative to historical earnings | Strongly supported |
Long-term expected real returns are probably lower than historical averages | Historically supported |
AI stocks specifically are overvalued | Not established by CAPE |
AI is a speculative bubble | CAPE cannot establish this |
AI earnings expectations are unusually important to market valuation | Yes |
A substantial earnings disappointment could have disproportionate market consequences | Yes |
Higher interest rates would be especially consequential | Yes |
AI productivity could eventually justify some of today's valuation | Possible |
Today's valuation requires unusually strong future earnings growth | Yes |
Is today's high CAPE a denominator problem or a numerator problem? The denominator problem happens if investors have simply bid up stocks too far relative to sustainable earnings.
The numerator problem happens if today's earnings substantially understate the future earnings power of the companies because AI is about to raise productivity and profits dramatically. Take your pick.,
The second possibility is the "this time is different" argument, an obvious red flag for most of us. Historically, that argument has sometimes been correct economically while still being wrong financially.
In the shorter term, investors will have to evaluate the likely impact of higher interest rates (the “real” rate being the thing that matters) on the AI ecosystem.
The conventional wisdom is that higher real interest rates (nominal rate minus inflation) tend to slow economic growth. That should mean the AI ecosystem also grows more slowly, as higher financing costs reduce investment and therefore potentially future revenue growth.
Also, the present value of distant future cash flows falls when the discount rate rises. This can be particularly severe for AI companies because much of their expected value is based on future rather than current earnings.
Part of AI ecosystem | Likely effect of higher rates | Why |
Hyperscaler AI capex | Moderate slowdown | Higher cost of capital raises the hurdle rate for marginal data centers, GPUs and power projects. But Microsoft, Alphabet, Amazon and Meta have enormous cash flows and strategic reasons to keep investing. |
Frontier-model companies | Moderate-to-large pressure | Companies with large compute bills and limited current profits become more dependent on external capital. Higher rates make investors demand a clearer path to monetization. |
AI startups | Large pressure | Valuations depend heavily on discounted future cash flows and access to venture capital. Higher rates particularly hurt companies whose revenues are distant or speculative. |
Data-center developers | Moderate-to-large pressure | These are extraordinarily capital-intensive, long-duration projects. Financing costs directly affect project economics. |
GPU/accelerator suppliers | Initially modest; eventually meaningful | Existing compute shortages and contractual commitments can insulate near-term demand. But slower infrastructure deployment eventually feeds back into equipment orders. |
Power/electrical infrastructure | Moderate pressure | Projects with long lead times and large upfront investment become harder to finance, although genuine power scarcity can preserve demand for some projects. |
Cloud AI services | Mixed | Higher rates can restrain customers' IT budgets, but AI can also be justified as a way to reduce labor costs or increase productivity. |
Enterprise AI software | Mixed/slightly negative initially | Discretionary experiments are vulnerable, while applications with measurable ROI may actually become more attractive if companies are under pressure to improve productivity. |
AI applications with little capex | Relatively resilient | Higher rates don't materially change their marginal cost of development or deployment. |
AI infrastructure with long-term contracts | More resilient | Contracted cash flows can support project financing even when the general cost of capital rises. |
Existing profitable technology companies | Relatively resilient | High cash generation makes them less dependent on external financing. |
Also, it matters greatly “why” interest rates are climbing. If rates rise because the economy is strong and inflation remains persistent, that is a very different scenario from rates rising because of a recession. And, at the moment, U.S. gross domestic product is still growing, according to the Federal Reserve.
Higher rates “should” put pressure on some suppliers in the AI value chain, such as high-performance-compute-as-a-service suppliers. But some will argue that better capital investment discipline will result.
Higher rates might even help suppliers of used graphics processor units, as such units might retain their value better.
Effect of higher rates | Likely impact on AI |
Higher cost of debt | Negative for leveraged AI infrastructure |
Higher required return on new projects | Negative for marginal data centers and GPU deployments |
Lower equity valuations | Negative, particularly for startups dependent on new funding |
More expensive private credit | Negative for neoclouds and infrastructure developers |
Pressure on hyperscaler free cash flow | Negative, potentially causing capex discipline |
Higher discount rate applied to future AI profits | Negative for valuations |
Stronger incentive to monetize existing GPUs | Mixed/positive—could increase utilization |
Higher hurdle rate for speculative projects | Potentially positive for industry discipline |
Cash-rich hyperscalers' ability to self-finance | Mitigates the effect |
As usual, higher borrowing costs will be a negative for startups and smaller firms more reliant on borrowed money.
AI layer | Rate sensitivity |
Nvidia/AMD-type highly profitable chip suppliers | Low–moderate |
Hyperscalers | Moderate |
Large profitable AI software companies | Moderate |
AI infrastructure developers | High |
Neoclouds | Very high |
Frontier-model startups | High |
Early-stage AI startups | Very high |
Still, ceteris paribus (“all other things being equal”) rarely describes events in the real world. All other things will not remain equal. Higher interest rates “should” slow AI investment. But competitive pressures within the industry, evidence of value and financial impact, capital availability and all sorts of other potential macroeconomic influences are likely to have an effect as well.
So interest rate increases might not have the slowing impact one might otherwise expect. Nor might the impact of a high CAPE necessarily cause the bursting of the "AI bubble." There are simply too many moving parts.
