Bruce Kahn on AI’s Real Bottleneck: “It’s Not the Chips, It’s the Wires”
The biggest constraint for artificial intelligence (AI) may have little to do with technology. The real challenge is building enough physical infrastructure to power its expansion.
That’s the view of Bruce Kahn, PhD, senior portfolio manager at Shelton Capital Management and a lecturer at Columbia University. He told the Investing News Network where he sees the top investment opportunities in the AI boom.
From grid capacity and corporate power deals to federal policy, Kahn pointed to a common thread throughout the conversation: AI may be advancing at unprecedented speed, but its growth ultimately depends on the physical infrastructure needed to support it.
The reality of AI revenue
Kahn believes that a meaningful share of the new data center capacity being built right now may end up serving cryptocurrency operations rather than AI workloads.
He argued that the crypto industry represents a more tangible source of compute demand, because it has immediate, proven revenue behind it. “We have yet to see any real revenue generation from (AI),” he said.
That skepticism lines up with numbers seen during this earnings season.
S&P 500 (INDEXSP:.INX) profits jumped 52.6 percent in the second quarter, the biggest year-on-year gain since 2021, but a meaningful share of that wasn’t operating profit.
Alphabet (NASDAQ:GOOGL) posted a roughly US$98 billion gain tied to its stake in SpaceX (NASDAQ:SPCX), and Amazon (NASDAQ:AMZN) saw a similar US$53 billion gain tied in large part to its stake in Anthropic.
Without those one-time gains, earnings growth was still a strong 33.8 percent, but much of what’s left is concentrated in companies selling into the AI buildout, like chipmakers, rather than proof that end users are paying enough to justify it.
The spending is also outrunning even genuine revenue.
Meta Platforms’ (NASDAQ:META) results show US$31.1 billion in quarterly capital expenditures, leaving the company with just US$784 million in free cashflow, even though it does have real, working AI revenue to point to, like its Advantage+ ad tools, which are now running at a US$75 billion annual pace.
And the spending shows no sign of slowing; Alphabet raised its 2026 capex guidance to US$195 billion to US$205 billion and Meta increased its amount to US$130 billion to US$145 billion.
Microsoft (NASDAQ:MSFT), which dropped the US$37 billion AI revenue run-rate figure it touted in April without explanation, told investors to expect capital spending to keep growing into next year during its latest earnings call.
Additionally, a large share of Microsoft’s headline backlog growth traces to a single customer, OpenAI, which by most accounts is dependent on continued outside fundraising to make good on those commitments, meaning a large portion of the AI buildout’s demand is tied to companies promising to pay with money it hasn’t fully raised.
Feeding into that is a broader pattern analysts have…
