The rapid advancement of artificial intelligence is often discussed in terms of technological breakthroughs and innovative applications. However, Bruce Kahn, PhD, senior portfolio manager at Shelton Capital Management and lecturer at Columbia University argues that the real challenge lies in building the physical infrastructure necessary to support this growth.
In an interview with the Investing News Network Kahn highlighted the critical role of infrastructure in the AI boom, emphasizing that the technology’s expansion is ultimately constrained by the physical systems required to power and cool data centers.
AI’s Revenue Reality Check
Kahn expressed skepticism about the immediate revenue generation potential of AI, contrasting it with the more tangible demand from the cryptocurrency industry. “We have yet to see any real revenue generation from (AI),”. he noted, pointing out that much of the recent earnings growth in the S&P 500 was driven by one-time gains rather than sustainable AI revenue.
For instance, Alphabet reported a roughly US$98 billion gain tied to its stake in SpaceX while Amazon saw a similar US$53 billion gain from its investment in Anthropic. Even with these gains, the underlying earnings growth was concentrated in companies selling into the AI buildout, such as chipmakers, rather than proof of substantial end-user revenue.
The spending on AI infrastructure is outpacing genuine revenue. Meta‘s results showed US$31.1 billion in quarterly capital expenditures, leaving the company with just US$784 million in free cash flow. Despite having real AI revenue streams, such as its Advantage+ ad tools the spending shows no signs of slowing, with companies like Alphabet and Meta raising their capital expenditure guidance for 2026.
The Infrastructure Bottleneck
Kahn emphasized that the mismatch between capital deployment and physical grid buildout is a current risk, not a future one. “We’re already at the bottleneck,” he stated, explaining that the demand for compute and data center capacity is already beyond what can be powered and cooled.
He expects frontier AI model providers to raise subscription prices to ration scarce compute and bandwidth, potentially leading to degraded system performance for users. Kahn also suggested that the bottleneck is part of why AI companies are moving to go public now, timing their IPOs to capture value while compute scarcity supports high valuations.
On the efforts of hyperscalers to sidestep the grid through off-grid solutions, Kahn sees these as short-term stopgaps rather than lasting fixes. He believes the long-term answer lies in large, centralized power, which he expects to be “more nuclear than renewables” at the scale required by AI and broader electrification.
Investment Opportunities in AI Infrastructure
Kahn’s investment approach focuses on physical, structural supply-and-demand trends that hold regardless of which AI company wins or whether the AI revenue story fully materializes. He avoids trading hyperscaler valuations or making calls on macro variables, preferring to invest in grid equipment, materials, and infrastructure—the “picks and shovels” of the buildout.
He cited digital water metering as an example, noting that water utilities must transition from analog to digital systems, creating a steady demand for related equipment. Kahn also highlighted specific sectors like geothermal and wind power, arguing that political efforts to block major wind projects signal their economic significance rather than a reason to abandon them.
His portfolio is built to be indifferent to which of these individual stories turns out to be right, focusing instead on the industrial and utility-side businesses that will be built out either way.
