A year ago, Chris Konstantinos, CFA, and Rod Smyth of RiverFront Investment Group were watching two variables in the AI buildout: whether compute supply could scale fast enough, and whether real demand would follow. Today, the second question has answered itself decisively. In their October 13, 2026 Strategic View1, the strategists report that "adoption has exceeded our optimistic estimates... growing so fast that supply can't keep up." In their view, that demand shock has fueled a broad earnings surge across sectors, led by technology, driving equities higher even as interest rates have risen.
Three Pillars, One Year On
Last November, Konstantinos and Smyth framed a constructive case built on three pillars: accelerating infrastructure buildout, rapid adoption, and measurable productivity gains. Their analogy was the California Gold Rush, not the dot-com bubble. Rather than betting on AI model providers (the prospectors), they favored chip manufacturers and hyperscalers (the picks, shovels, and mining operations). "Almost a year on," they state, "our conviction on all of these topics remains intact." The skeptic's case (that AI is a solution looking for a problem and that infrastructure will outrun demand) "has not shown up in the data."
The Opposite of a Bubble
The most important distinction Konstantinos and Smyth draw is between today's AI cycle and prior infrastructure bubbles. Railroads, fiber optic cable, and housing all collapsed under oversupply. The AI infrastructure cycle is running in the opposite direction. The four largest hyperscalers guided to a combined $720-$745 billion in 2026 capital spending, roughly double the $350 billion pace of 2025. And they are still telling investors it is not enough.
Amazon states its roughly $220 billion plan will not provide sufficient capacity for 2026, and expects to remain constrained into 2027. Microsoft's Azure grew 43% last quarter; CFO Amy Hood states plainly: "Demand continues to surpass supply." Alphabet has raised its capex range twice since February, most recently to "deliver capacity faster to meet demand." TSMC expects 2026 revenue growth above 40% and acknowledges that "our packaging capacity is so tight that now it's limiting our customers' growth." Nvidia's data center revenue rose 117% year-over-year to $89 billion and guided higher still. The strategists observe: "a demand-constrained market looks like price cuts and inventory build...the opposite of what we're seeing today."
Faster Than the PC, Faster Than the Internet
Stanford University research published in 2026 confirms what RiverFront cited a year ago: generative AI reached 53% global adoption within three years of ChatGPT's launch. The PC took roughly 15 years to reach comparable penetration. The internet took 7. The report also finds that 88% of global organizations now use AI in at least one function. Konstantinos and Smyth note that workers are adopting AI faster than employers are formalizing it, which is typical of early-stage general-purpose technology. Their implication: the corporate enterprise wave may still have its largest chapter ahead.
The Productivity Case Holds
The Bureau of Labor Statistics reported non-farm business productivity growth of 2.2% year-over-year in Q2 2026, following 2.9% in Q1. Both readings sit above the roughly 1.5-2% pre-AI trend. Unit labor costs rose only 1.4% over the past year. The unemployment rate stands at 4.1%, with layoffs described as low and relatively flat. The strategists are measured: "It is too early to declare victory on that estimate, but so far the data is tracking it, not contradicting it."
What Could Change the View
RiverFront is not complacent. Konstantinos and Smyth are monitoring four potential trip wires: a divergence between reported earnings and free cash flow (the classic bubble signal, not yet visible); pricing power erosion as compute costs fall; debt-financed interlocking deals between chipmakers, hyperscalers, and AI labs flagged by the Bank for International Settlements as a financial-stability risk; and electricity and materials inflation, with the Fed having raised rates in September. "Humility and flexibility will be our friends as we navigate this rollercoaster of investor excitement and potential disappointment," they write.
As long as adoption is constrained by supply of chips, packaging, and power rather than by customer demand, the earnings runway for AI infrastructure companies remains long. Volatility is expected. For Konstantinos and Smyth, managing that volatility is preferable to missing the trend.
5 Key Takeaways for Advisors and Investors
- Supply, not demand, is the binding constraint. Hyperscalers are still building faster than capacity allows. This is structurally opposite to prior infrastructure bubbles.
- AI adoption speed is historically unprecedented. Reaching 53% global penetration in three years far outpaces the PC and the internet. The enterprise wave may have further to run.
- Productivity gains are measurable and above pre-AI trend. US non-farm business productivity is tracking above the pre-AI baseline, with modest labor cost growth. The economic case is holding.
- Infrastructure over models remains the preferred framework. Chip manufacturers and hyperscalers continue to expand margins even as compute costs fall. The picks-and-shovels framing is intact.
- Four risk indicators warrant ongoing discipline. Free cash flow divergence, compute commoditization, debt-financed AI structures, and rising rates and energy costs are the gauges Konstantinos and Smyth are watching most closely for signs that the constructive view warrants revision.
Footnote:
1 Konstantinos, Chris, CFA, and Rod Smyth. "The AI Boom Accelerates." RiverFront Investment Group, 13 Oct. 2026, https://www.riverfrontig.com/insights/the-ai-boom-accelerates/.
