Cathie Wood's Case for a Bull Market Even With High Rates
ARK's latest letter argues AI-driven productivity could push growth up and inflation down while rates stay high. The real question for investors isn't where rates go, but who can turn falling AI costs into rising cash flow.
The letter's core claim isn't "rates must fall"
Growth-stock logic often gets reduced to a simple line: rates have to come down for tech stocks to rise. Cathie Wood's latest ARK letter flips that premise. ARK argues the current technology wave could produce a stronger productivity shock than the Industrial Revolution, one that lifts real growth even as it pushes prices lower.
That is where rates come in. ARK lays out a scenario in which real GDP growth accelerates past 7% and inflation falls below 2%, yet short-term rates stay in the 6% to 8% range, tracking nominal GDP growth, while long-term rates hold around 5% to 6%. Under that scenario, an inverted yield curve would not necessarily signal a recession.
There is some overlap with today's market. The US 10-year Treasury yield stood at 5.01% as of September 18, and the 10-year real yield, stripping out inflation, climbed to 2.64% on a weekly basis. That gives some support to the idea that rising long-term rates reflect strong real returns, not just inflation expectations.
If you remember one number, make it the pace of AI cost declines
<div class="grid"> <div class="card"><div class="metric">99%+</div><div class="label">ARK's estimate for the annual decline in AI inference costs</div></div> <div class="card"><div class="metric">25x</div><div class="label">ARK's estimate for the increase in AI inference token demand in 2025</div></div> <div class="card"><div class="metric">7x+</div><div class="label">The seven-month jump in Anthropic's annualized revenue, per ARK</div></div> </div>
The part of this letter investors should watch most closely isn't the sweeping GDP forecast, it's the cost curve. ARK argues that the inference costs powering services like ChatGPT, Claude, Gemini and Grok are falling more than 99% a year. At the same time, ARK estimates inference token demand rose 25-fold in 2025.
When prices are falling fast but usage is growing even faster, AI becomes a textbook case of "good deflation." Per-unit prices can drop while total revenue and usage explode. That is also why demand is spreading across semiconductors, cloud, data centers, power and networking, and software broadly.
ARK estimates Anthropic's annualized revenue rose from $9 billion to $65 billion over seven months, and could top $100 billion by year end. Anthropic is a private company, however, and these figures are ARK's estimates built on outside data, not audited disclosures. The direction is notable, but it should not be treated as confirmed results.
Why "an inverted curve means recession" may not hold either
ARK points to the late 19th century Industrial Revolution as precedent. Over roughly 50 years, the yield curve was inverted more than 60% of the time, yet productivity gains and technological innovation lifted the economy and stock market for a sustained stretch.
The logic: short-term rates stay elevated, tracking strong nominal growth and demand for capital, while long-term rates rise less because they reflect the lower inflation that technology produces. In that case, an inversion could reflect strong growth paired with low inflation, not an approaching downturn.
The historical parallel has limits, though. Today's financial system, central bank framework, government debt load and global capital flows look very different from the 19th century. Rather than concluding the yield curve has lost its信号 value, it's more useful to treat this as a hypothesis: the stronger the productivity shock, the more existing recession signals may need reinterpreting.
What matters more than "is AI right" is "who keeps the cash"
| ARK's claim | What it means for investors | What to verify |
|---|---|---|
| Technology deflation | Falling prices aren't automatically bad news. If usage grows faster than prices fall, revenue and profit can still climb. | AI usage volumes, inference unit costs, gross margins, customer ROI |
| High productivity | Companies that reinvest savings to grow market share may end up stronger than those that just cut labor and operating costs. | Revenue growth, operating margin, R&D, capex, free cash flow |
| Rates staying high for longer | "Growth stocks are always weak when rates are high" matters less than how a company is financed. | Net cash, interest expense, floating-rate debt, debt maturity schedule |
| Disruption of incumbent industries | Chasing AI winners isn't enough. Avoiding companies losing pricing power or customers to AI competitors matters just as much. | Pricing pressure, customer retention, rivals' pace of AI adoption |
The sharpest warning is about floating-rate debt
ARK argues the technology wave is more likely to raise credit demand than to push rates down, because autonomous driving, robotics, data centers and power infrastructure all require enormous capital. The riskiest position, in ARK's view, is a company carrying heavy floating-rate debt without benefiting from the innovation wave.
The letter notes roughly $16 trillion is tied up in private equity and private credit, and that US federal debt is approaching roughly $40 trillion. It sketches a scenario where a 2 percentage point rise in rates could add roughly $800 billion to the government's simple interest cost. The actual burden will depend on debt maturities and refinancing pace, but the message for investors is straightforward: in a high-rate era, the balance sheet matters as much as the growth rate.
Insight Times Editorial Desk
