Aug 21, 2026 · 1h 10m
AI Slows Wage Growth as Trade and Water Pressures Mount
AI Hits Wages, USMCA Under Pressure, Colorado River Crisis
The episode connects AI’s uncertain investment payoff and labor-market effects with fragile trade ties and a worsening Colorado River conflict.
- 1AI infrastructure requires long investment cycles, while returns remain uncertain and cheaper open-source models could challenge spending plans.
- 2Early evidence suggests AI exposure is slowing wage growth more than reducing employment, with productivity and business creation still unresolved.
- 3USMCA tensions and Colorado River shortages show how policy, legal rights, and economic interdependence complicate fast-moving crises.
Don't miss
Torsten Slok presents early research suggesting AI exposure is slowing wage growth more than reducing employment.
The brief
Sam Palmisano and David Westin examine the mismatch between rapidly changing AI models and the years required to build data centers, chips, and energy infrastructure.
Torsten Slok’s research finds early AI exposure affecting wage growth more than job counts, while Diane Gerson argues that employee involvement can make automation less demoralizing.
Chrystia Freeland describes a tentative U.S.-Canada agreement that may avert new tariffs but leaves major questions about steel, autos, workers, and USMCA unresolved.
The Colorado River report follows shrinking snowpack, declining reservoirs, and disputes among seven basin states as proposed cuts threaten agriculture, cities, and regional growth.
Water exchanges and costly efficiency projects offer flexibility, but the report’s larger conclusion is stark: the Southwest must price scarcity after decades of cheap abundance.
Palmisano closes by arguing that government should set standards and guardrails for AI before safety risks and autonomous systems outpace public expertise.
What was said on this episode
21 statements · 8 positive · 8 negative · 1 mixed · 4 neutral
Technology investments should exceed a 14–15% IRR hurdle before launch.
“if your IRR is 15% or 14%, that would be the hurdle rate the guys would have to come over before you launched.”
Listen at 4:14
AI data centers typically require three to five years before coming online.
“Data centers aren't changing fast. I mean, the estimates are three to five years before this stuff comes online.”
Listen at 4:43
Hyperscaler investment may squeeze other AI market participants out.
“Other guys that are trying to compete or even to participate in this market are going to, I think, be squeezed out.”
Listen at 5:25
Slower AI adoption would reduce the pace of capacity buildout.
“Just a simple thing. Your adoption curve, you assume to be seven to eight years. Let's make it 10 to 12. That slows. You're not going to build out the capacity as quickly.”
Listen at 6:04
Major technology cycles typically build infrastructure, scale, and then undergo a correction.
“the pattern's the same over time. It starts with... she calls the installation cycle uh we would call maybe the industrial i mean infrastructure build out today that's the base that that gets put in place and it takes several years to do that obviously to get it to scale and then normally what has happened over this history here is that it has corrected”
Listen at 8:29
Technology investment plans should include a downside case or Plan B.
“you have to have a downside case to go. We call it a plan B, quite honestly.”
Listen at 10:04
Off-balance-sheet AI investment estimates lack transparency.
“There's no transparency, as you probably know, in the off-balance sheet estimates.”
Listen at 10:54
Chinese open-source AI models require less capacity than U.S. proprietary models.
“the chinese are arguing with deep seek and those sorts of things and the latest ones have just come out that there's no reason why they can't have this open source capability available and it doesn't require their models don't require the same capacity as the us models”
Listen at 11:47
China will deploy large-scale, low-cost, high-quality AI models globally.
“they'll build a massive scale low cost High quality, which people say couldn't be high quality. High quality, take it to market, and then we know what happened everywhere else. And they're running the same play here.”
Listen at 12:38
AI automation is initially targeting high-attrition jobs.
“For certain jobs where there's high attrition, those are the jobs that are being automated the first with AI.”
Listen at 21:21
Paralegal roles will be automated before higher-level legal roles.
“it's the paralegals that are going to go first”
Listen at 23:02
AI-created products and innovation can increase company valuations through growth.
“But if we can create new products, if we can create new opportunities, if we can do breakthrough innovation, then, you know, then we'll have a higher valuation because we'll have growth.”
Listen at 24:42
Radiologists’ wages rose 42% faster than software engineers’ since 2021.
“radiologists, which everyone thought was going to go away. Actually, their wages have gone up faster, 42 percent faster than software engineers since 2021”
Listen at 26:41
Arizona communities are preparing for 760,000 acre-feet of water reductions.
“These reductions that we're talking about, 760,000 acre feet for Arizona, is something that our communities have been preparing for.”
Listen at 50:14
Government should facilitate AI opportunities rather than control them.
“fundamentally, the government needs to facilitate the opportunities of the new technology, not try to control the opportunities of the new technology.”
Listen at 1:00:05
Government should establish standards while allowing private-sector investment.
“they establish those standards right they had parties that They let the private sector invest.”
Listen at 1:00:37
AI needs guardrails but not necessarily heavy regulation.
“You need guardrails. I say guardrails. I don't know that you need heavy regulation”
Listen at 1:02:53
Government AI expertise currently lags behind the private sector.
“the government is well way way behind the private sector”
Listen at 1:03:53
The U.S. should recruit industry experts to develop a national AI strategy.
“I'd get some volunteers. I think they call them something like governmental subject experts or something like that. Get some volunteers, really smart people. who've been around this thing, who have industry knowledge, who know technology, who could come back with a strategy for the United States of America.”
Listen at 1:04:49
Unaddressed AI safety risks could become uncontrollable within minutes or seconds.
“if something's not done to address it, it could be out of control in like minutes or seconds”
Listen at 1:07:03
Future AI systems will need built-in controls for self-policing.
“you're going to have to have those controls built into the systems to self-police them.”
Listen at 1:08:26
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.