Episode summary
US Has Under-Invested in Required AI Infrastructure, Says Michael Green
Bloomberg Surveillance
· · 8 min · Guests: Michael Green
Summary
Michael Green argues that the AI boom is running into a physical bottleneck, not a software one. He says the U.S. has underinvested in power and grid infrastructure for years, so the surge in data centers, model training, and industrial AI use is colliding with rising electricity prices and an electricity network that cannot easily balance demand across regions.
He describes the capital needs as far larger than typical corporate budgets, saying firms are shifting from equity financing toward debt financing and that the spending could reach trillions. Green also frames the moment as a broader industrial reallocation: the economy is moving from feeding humans to feeding machines, with electricity becoming the gating factor for AI growth. He cites Meta’s $135 billion capex as evidence that one company alone can’t solve the scale problem.
On the business side, Green says AI is still likely to show up first as a search replacement and an advertising product rather than as a mass consumer subscription. He expects companies like Dow Chemical to use AI, but thinks the biggest margin gains will depend on whether firms can pass costs through to consumers or face competition that limits pricing power.
Key takeaways
- Green says the U.S. has underinvested in electricity and grid infrastructure, making power the gating factor for AI data-center expansion.
- He argues AI capex is shifting from equity to debt financing and could amount to trillions, not just tens of billions.
- Green expects ad-supported AI and search replacement to matter more than $200 consumer subscriptions like Gemini Pro.
- He says companies such as Dow Chemical will adopt AI, but margin gains depend on whether they can pass savings or costs through to customers.
Notable quotes
“I think the reality is we're hearing more and more of this from individual companies that they're encountering the reality of the physical world”
“the U.S. has underinvested in the infrastructure that is required for the projections for air”
“we are shifting from a world where we're basically focused on feeding humans to feeding machines”
“We're talking trillions potentially here. This is ultimately going to be one of the largest CapEx expenditures we've ever seen”
“the vast majority of consumers are going to be signing up for $200 Pro editions of Gemini or WeChat Gbtc really seems increasingly fantastical to me”
How the episode opens
Well, I am skeptical about it. And unfortunately, I think the reality is we're hearing more and more of this from individual companies that they're encountering the reality of the physical world as compared to the largely digital world that we tend to think about when we talk about the cloud. It's not, you know, an imaginary space.
It's actually a heavy industrial complex that requires massive amounts of data centers. And what we have really experienced in the United States since roughly 2000 has been the outsourcing of much of our manufacturing activity, which has freed up capacity in the electricity production. We've moved aluminum, steel, etc., things that are very energy intensive offshore, and now suddenly we've exhausted that capacity.
And so it means that Americans and their households are starting to experience rising electricity prices. This is creating the same challenges for data centers and others. And it's a simple reality that the U.S.
has underinvested in the infrastructure that is required for the projections for air. So we see in that vein, we saw Mehta just last night talk about a huge CapEx of $135 billion. And they're just one company.
That's just one company. So where and I guess what we hear in terms of the development and the of I it's one of the if not the gating issue is going to be the access to power. How is that something that can be solved over a reasonable period of time?
I think it's really challenging, unfortunately, because you are talking about putting physical infrastructure in place. And the bigger issue that you ultimately face in this is that the overall electricity network, right, our grid is effectively incapable of conducting the balancing that is required to smooth that across wide regions. And so what you've now heard from President Trump and others is that they are going to require these companies to effectively source their own electricity.
From the episode’s captions: roughly the first two minutes, lightly cleaned. Keep watching on YouTube →
Topics
- ai infrastructure underinvestment
- ai credit/issuance boom
- high-grade corporate bond sales
- tmt debt issuance forecasts
- equating machine labor to human labor
Companies and tickers mentioned
- IBM
- META
- MSFT
- JPM
- Simplify Asset Management
- International Business Machines
- Meta Platforms
- Microsoft
- JPMorgan Chase & Co.
Watch the episode
Watch “US Has Under-Invested in Required AI Infrastructure, Says Michael Green” on YouTube
Summary, takeaways and quote selection by Octagon. Quotes are short excerpts; the full conversation is in the episode.
More from Bloomberg Surveillance
-
Stocks Rise as Inflation Data Ease Fed-Hike Worry | Bloomberg Surveillance -
Bloomberg Surveillance TV: November 13th, 2026 | Bloomberg Surveillance -
Bloomberg Surveillance TV: November 12th, 2026 | Bloomberg Surveillance -
Jensen Huang Ousted by Nvidia Board as CEO | Bloomberg Surveillance -
Bloomberg Surveillance TV: November 11th, 2026 | Bloomberg Surveillance -
AI Stocks Rebound as Fed Hike Bets Return | Bloomberg Surveillance
Bloomberg Surveillance, this episode and its artwork belong to Bloomberg. Octagon is not affiliated with the show; this page is an independent summary.
