Cathie Wood: Token Growth of 25 Times Will Boost Real GDP
ARK Invest founder Cathie Wood cited Elon Musk's statement that "the AI rift has begun" and stated that the reasoning Token volume is growing 25 times a year and is still expanding exponentially, potentially pushing the real GDP growth rate into double digits; many clients and companies do not understand this magnitude of numbers.
She added an arithmetic note: most investors hearing "25" in growth will automatically think of 25%, which has historically been considered good; 25 times is on a different scale, where 100 becomes not 125, but 2500. She then mentioned that revenues from leading AI labs are seeing 5 to 10 times growth within six months to a year, and more mature companies on the correct side of AI changes are seeing revenue growth rates accelerating from 25% back to over 30% to 40%. She is starting to see more "deer in the headlights."
Federal Reserve Chairman Kevin Warsh referred to AI as a historical hinge point in his Jackson Hole speech, citing that the annualized sales of Tokens from two leading labs have exceeded $100 billion, increasing by over 500% from a year ago, and the Fed is beginning to view AI as a potential new factor of production. ARK analysts estimate that Anthropic's annual recurring revenue exceeds $74 billion, and OpenAI exceeds $41 billion, totaling over $115 billion, close to Microsoft's productivity business of about $150 billion, with data sourced from third-party tracking rather than audited reports.
Wood previously revised global real growth expectations up to about 7% and stated that the convergence of five platforms: robotics, energy storage, AI, blockchain, and multi-omics sequencing will further elevate long-term growth rates. Musk has publicly discussed growth potentially being several times the old benchmarks. Tokens are reasoning and generation units sold by labs based on usage, with rising sales corresponding to expansions in electricity, chips, and data center capital expenditures.
ARK's portfolio writes this type of nonlinear growth as a condition that must cover valuation multiple compression: internally assuming a market multiple of 18 to 19 times EBITDA in the fifth year, requiring revenue and profit expansion to outpace multiple declines. Cyclical segments like memory are excluded by her due to their high level of commoditization.
In market mechanisms, buyers treat Token usage as a theme fund for new economic flows, while sellers are institutional clients still pricing tech stocks using linear macro models. The event was triggered by Musk's rift metaphor, with funds continuing to flow into labs for pre-purchased computing power, cloud vendor capital expenditures, and ARK's heavily weighted innovation portfolio. Beneficiaries are labs and chip chains that can directly convert usage into bills, while those pressured are the pricing frameworks that interpret 25 times as 25%.
ABAB AI Insight
Wood has been writing macro forecasts into ARK product descriptions for a decade: first establishing a growth rate that is ridiculed by the mainstream, then waiting for the revenue curves of one or two labs to validate it. Around 2014, she used mobile, genetics, and blockchain for the first round of nonlinear stories; this round has shifted the unit from user numbers to Tokens. The correction of 25 times versus 25% is not a teaching moment, but a redefinition of which compounding table clients should use. ARK's holding discipline—compressing back to market multiples in the fifth year—acknowledges that thematic premiums will drop and can only be filled by explosive revenue growth.
Money flows from ETF subscriptions into the public market's growth stocks, then seeps into labs through cloud contracts and equity investments. Labs convert Token prices multiplied by usage into recurring revenue, then lock in megawatts from Nvidia, Oracle, and CoreWeave. The motivation is: if clients interpret 25 times as 25%, the fund's net value will be seen as a bubble; it is essential to first change the clients' magnitude language for subscriptions to follow suit. The strategy is to integrate the "new factor of production" acknowledged verbally by the Fed Chairman into the same narrative, reducing the policy discount on thematic investments.
A similar path can be seen in the 1990s when internet user numbers were treated as a GDP independent variable, and after 2007, smartphone activations were treated as a consumption independent variable. Microsoft Office took decades to reach the $100 billion level, while two labs are approaching this range in just a few years using third-party metrics. The industry is in a phase of usage expansion, and pricing power has not yet shifted from labs to the enterprise software distribution layer; the 30% to 40% re-acceleration of mature companies is the first time old ledgers are forced to change their slope.
The essence is that technological substitution rewrites growth accounting. Tokens break cognitive labor into billable increments, which occur faster than the statistical frequency of employment and capital stock, so actual output may lag in official quarterly numbers. The mechanism is: indexing usage raises leading indicators for electricity and chips, with GDP as a posterior aggregate following behind; whoever uses 25 times instead of 25% as the denominator will leverage their stock duration ahead of others.
ABAB News · Cognitive Law
- Misinterpreting multiples as percentages is the costliest hearing error.
- Once new factors of production become billable, old growth tables will become obsolete.
- Revenue must first rise vertically before cognition can follow suit.