
Nov 16, 2026 · 51 min
Election politics put AI infrastructure’s next phase under pressure
The Blue Wave Meets AI: What the Midterms Mean for Data Centers | The Real Eisman Playbook Ep 84
The episode tests whether AI’s enormous infrastructure buildout can deliver durable returns as politics, financing costs, and model competition reshape the investment case.
- 1Political resistance and higher financing costs could slow data-center construction while increasing the value of existing hyperscaler capacity.
- 2AI models may commoditize, but enterprise control layers, proprietary data, and scarce technical talent could preserve durable advantages.
- 3A healthier AI cycle may depend on cash-flow-funded growth, separating resilient platforms from leveraged infrastructure and exposed software businesses.
Don't miss
The guests’ debate over whether interchangeable AI models will commoditize the market—or whether talent, enterprise control layers, and proprietary data create lasting moats.
The brief
Steve Eisman, Dan Ives, and Gil Lurie revisit AI’s investment case after the election, focusing on whether political resistance and capital demands can slow the infrastructure boom.
The guests argue that slower data-center construction could make existing hyperscaler capacity more valuable, while debt-funded neoclouds face a harsher test from borrowing costs and politics.
Google’s search moat, Oracle’s leveraged AI expansion, and Intel’s broken turnaround become case studies in how AI can expose different kinds of strategic and financial weakness.
The sharpest disagreement concerns durable moats: models may become interchangeable, but enterprise control layers, proprietary data, and scarce technical talent could still protect leading platforms.
Apple emerges as an unusually insulated participant, able to distribute AI through its installed base and services ecosystem without funding frontier-model development.
The closing framework favors a slower, cash-flow-funded AI cycle that separates sustainable demand from infrastructure bets whose returns depend on constant capital spending.
What was said on this episode
35 statements · 16 positive · 13 negative · 2 mixed · 4 neutral
AI capital expenditure likely peaks in 2027 and its effects appear by 2029.
“CapEx 27 is probably a peak year for CapEx. You probably wouldn't see that until 2029.”
Listen at 3:20
U.S. data-center construction will become harder and more expensive.
“it's going to be harder and more expensive to build data centers in the U.S.”
Listen at 4:55
Political resistance may prevent 10%–15% of planned data centers from being built.
“with incrementally, let's call it 10% to 15% of data centers now not being built”
Listen at 5:42
Hyperscalers benefit through revenue growth or cash-flow acceleration regardless of AI demand.
“If there's more AI, the revenue continues to accelerate. If there's less AI, their cash flow accelerates.”
Listen at 7:05
Existing hyperscaler data centers become more valuable when new facilities cannot be built.
“the installed base of data centers. That just became more valuable because we can't build other data centers.”
Listen at 7:39
Google Search’s abrupt growth slowdown may indicate a structural problem.
“it indicates that there may be something broken”
Listen at 9:55
Advertisers may shift 10%–20% of search budgets to ChatGPT or OpenAI.
“all the advertisers will switch 10%, 20% of their budget to ChatGPT, to OpenAI.”
Listen at 10:40
Alphabet stock could rise significantly if AI advertising succeeds 30%–40%.
“If they just show even 30%, 40% success rates, I think it's a stock that could go significantly higher.”
Listen at 13:06
Oracle made a $300 billion OpenAI bet while its business suffered significantly.
“They bet hugely on OpenAI with the $300 billion, and the business has suffered significantly.”
Listen at 15:42
Intel is fundamentally broken unless chip yields support margins of at least 50%.
“if you can't make chips at a high enough yield to get at least a 50% margin, you're a broken company.”
Listen at 19:35
Intel’s severe execution problems make a turnaround difficult.
“It's very hard to fix a company that's broken this badly.”
Listen at 20:10
AI models will become commoditized over time.
“models are going to commoditize over time.”
Listen at 22:07
Anthropic could reach $1 trillion in revenue within eight to ten years.
“if they could do $500 billion, $700 billion in revenue, get to a trillion over the next eight to 10 years”
Listen at 23:01
A scarce group of roughly 100 experts forms leading AI labs’ core moat.
“Their moat is the team. There's only 100 people in the world, not an exaggeration, 100 people in the world that are smart enough to move AI forward.”
Listen at 23:38
OpenAI and Anthropic will lose share of the overall AI market.
“Open AI's and Anthropics' slice of the pie is going to shrink.”
Listen at 27:14
OpenAI and Anthropic will not supply the entire AI market.
“they will not be able to provide all of AI”
Listen at 27:25
Investors should buy Anthropic only if total AI growth exceeds its share loss.
“you have to be very confident. that the pie is growing faster than their slices shrinking in order to buy into this.”
Listen at 29:10
Traditional software companies can eventually monetize their installed bases after a painful AI transition.
“I view it as much more of a painful transition period where they will get on the other side. They'll be able to monetize their install base.”
Listen at 31:02
Adobe’s enterprise value comes from workflows, governance, and collaboration rather than code alone.
“Adobe doesn't sell us this code that we use internally. They sell us workflows and data governance”
Listen at 33:19
Some neoclouds will fail, while CoreWeave and Nebius are likely leaders.
“Not all of them are going to survive. Some will. I think CoreWeave and Nebbias will ultimately be ones that lead.”
Listen at 35:42
CoreWeave generates low-single-digit returns while borrowing at roughly 9%, creating negative carry.
“CoreWeave is an enterprise that generates low single-digit returns and borrows at 9%. It's a negative carry.”
Listen at 38:35
CoreWeave equity holders may receive no value.
“The equity holders own nothing.”
Listen at 39:22
Neoclouds may face problems when compute prices stop rising rapidly.
“It's when the price of compute starts even leveling off that you may see issues.”
Listen at 40:43
One-fifth of the world will access AI through Apple devices in coming years.
“20% of the world will access AI over the coming years to an Apple device.”
Listen at 42:08
Apple receives 30% of ChatGPT subscription revenue downloaded through its ecosystem.
“when I download ChadGPT and do a $20 subscription, guess what? They get paid 30% of that.”
Listen at 42:58
Agentic AI adoption will take several years to become normalized.
“we're still a few years from this becoming, I think, normalized.”
Listen at 44:55
Regulatory scrutiny will slow agentic AI adoption.
“the political backdrop, what we see today, you're going to have a much higher bar relative to regulatory. And I think that's something that would be negative. That was slow roll this.”
Listen at 45:04
Agentic AI can create consumer value by reducing transaction friction.
“Reducing friction creates value. It creates surplus.”
Listen at 45:17
Chip supply and demand may not reach equilibrium until late 2028 or early 2029.
“we probably don't have equilibrium from a chip perspective to late 2028, early 2029.”
Listen at 46:38
AI capital expenditure likely peaks in 2027, with free cash flow rebounding significantly in 2028.
“CapEx is probably peak year 2027. Free cash flow is going to start to significantly rebound 2028.”
Listen at 47:13
Technology stocks may rise significantly through year-end and the following six to nine months.
“tech stocks are going to see a pretty... significant run, not just in the year end, but over the next six, nine months”
Listen at 47:40
AI is entering a healthier, sustainable growth scenario.
“we're headed into a Goldilocks scenario for AI”
Listen at 47:52
AI data-center growth may be 25% next year rather than 70%.
“if we grow 25% next year instead of 70%”
Listen at 48:07
Leading chip companies would be substantially undervalued under sustained moderate growth.
“those companies are worth a lot more than they are now.”
Listen at 49:36
Responsible, debt-light AI infrastructure growth would be highly beneficial.
“If it ends up growing in a responsible way, without debt, with data centers that communities want, that's a fantastic thing.”
Listen at 50:38
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.