Latest / The Macro Memo with Fexingo: Daily Conversations on Inflation, GDP, and Federal Reserve Policy / The Real Cost of AI Energy Consumption
Transcript
- Lucas: Let’s talk about something that isn’t on the Fed’s immediate radar but is quietly reshaping the cost structure of the entire digital economy: the sheer physical weight of artificial intelligence. Luna: You mean the servers? The ones humming away in warehouses we never see? Lucas: Exactly. Not just the hardware, but the electricity it devours. If you look at the broader macro picture as of September first, twenty twenty-six, we are seeing a fundamental shift in how power is consumed. It used to be manufacturing or housing. Now, it is data centers. Luna: And why does that matter to someone trying to figure out their mortgage payments or business overhead? Lucas: Because energy costs flow through everything. When the grid gets strained by massive AI workloads, utility rates go up. And when utility rates go up, every company’s operating expenses rise, which eventually trickles down to the CPI. Luna: So you’re saying the boom in AI is actually an inflationary force right now? Lucas: It is a structural one. Let’s look at the numbers. The ten-year breakeven inflation rate is sitting at two point three five percent as of yesterday. That’s higher than it was last month. Investors are pricing in sticky long-term inflation. Luna: Which is weird, because usually tech brings deflationary pressure with efficiency gains. Lucas: That was true for software alone. But generative AI requires massive computational power. And computation requires cooling, power, and infrastructure. It is incredibly energy-intensive. We saw private payrolls rise by only thirty-eight thousand in August according to the ADP report. That is weak job growth. Luna: But if those jobs are high-skill engineering roles building these data centers, they don’t show up in broad consumer spending immediately. Lucas: Right. And meanwhile, the core PCE price index, the Fed’s preferred gauge, is at one hundred thirty point seven. It’s still ticking up. Why? Because the underlying costs of running the digital backbone of our economy are rising. Luna: Is this happening everywhere, or just in places with heavy tech adoption like Virginia or Northern California? Lucas: It’s becoming global. Look at the headlines from the Bank of England. Their chief recently warned that new AI models threaten global financial stability. Part of that threat is operational risk, but also resource scarcity. Luna: Resource scarcity for what exactly? Chips? Or actual electrons? Lucas: Electrons. The demand for electricity from hyperscale data centers is projected to double in the next few years. This isn’t just about buying more GPUs. It’s about building substations, laying new transmission lines, and securing natural gas or nuclear contracts. Luna: That sounds like a capital expenditure nightmare for utilities. Lucas: It is. And who pays for that capital expenditure? Ratepayers. So your electric bill goes up. Then, companies factor that higher energy cost into their pricing models. That creates a feedback loop where inflation persists not because wages are skyrocketing, but because the base cost of doing business has shifted. Luna: I hadn’t thought about the utility bill connection. Most people blame wage growth for sticky inflation. Lucas: Wage growth is a factor, sure. Average hourly earnings are at thirty-seven dollars. But energy is a universal input. If the cost of moving data doubles because of power constraints, then cloud services, streaming, e-commerce, and even fintech apps become more expensive. Luna: So the 'free' internet we grew up with is getting a price tag. Lucas: Precisely. And here is the kicker: the Fed watches the labor market closely. The unemployment rate is four point one percent. Job openings are at seven million two hundred seventy-one thousand. The labor market is still tight. Luna: But if energy costs rise, doesn't that slow down economic activity? Lucas: It can. We saw real GDP growth slow to one point five percent annualized in the second quarter. That is a deceleration from two point one percent. Part of that slowdown could be attributed to businesses hesitating to invest due to uncertainty around energy availability and costs. Luna: So we have slowing growth but persistent inflation. That sounds like stagflation lite. Lucas: Stagflation lite is a good way to put it. The Fed is caught in a bind. They can’t cut rates aggressively because inflation is sticking around, partly due to these structural energy costs. But they can’t keep rates high too long if growth is stalling. Luna: What’s the solution? Build more power plants? Lucas: They are trying. But permitting takes years. Nuclear is promising but slow. Renewables are intermittent. Data centers need constant, baseload power. This mismatch between the speed of AI deployment and the speed of energy infrastructure build-out is the core tension. Luna: It feels like we’re underestimating the physical footprint of virtual technology. Lucas: We absolutely are. We think of AI as code. But code runs on silicon, which runs on copper, which runs on electricity. You cannot separate the two anymore. Luna: Do you think this will lead to a split in the market between energy-rich and energy-poor regions? Lucas: Already happening. Countries with abundant cheap power, like parts of Scandinavia or Canada, are attracting massive data center investments. Places with constrained grids are being left behind. It’s a geopolitical shift driven by kilowatt-hours. Luna: Wow. So if I’m investing in infrastructure, I should look at energy providers, not just tech stocks? Lucas: If you want exposure to the AI boom without the valuation risk of individual chipmakers, yes. Utilities and grid modernization companies are the toll roads of this new economy. Luna: That’s a much more tangible play than betting on which chatbot wins. Lucas: Exactly. Betting on the shovel sellers during a gold rush is usually safer. Here, the shovels are transformers and turbines. Luna: It really changes how you view the news headlines about AI breakthroughs. Lucas: Every time there’s a new model release, ask yourself: how much power did it take to train? That number is going to dictate margins for the next decade. Luna: Okay, before we dive back into the specifics of grid capacity, I want to pause for a moment. Lucas: Sure thing. Luna: If walking through the economy with us has made something click, especially on these complex structural shifts, please consider supporting the show directly. Lucas: A handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that's literally what funds making this many of these shows possible. Luna: It keeps us independent, no ads, just straight analysis. We appreciate every single contribution. Lucas: Back to the grid. The issue isn't just total power, it's location. Luna: Right, because you can't just plug a data center anywhere. Lucas: No. The grid has local limits. Even if there is enough national power, a specific substation might be maxed out. So tech companies are having to negotiate directly with power producers, bypassing traditional utility channels in some cases. Luna: Direct power purchase agreements. That must drive up costs for everyone else on the grid. Lucas: It does. When a massive buyer secures all the available green energy, smaller businesses get priced out or stuck with dirtier, cheaper power that might face future carbon taxes. Luna: So it creates a two-tier system for energy access based on company size. Lucas: Pretty much. Small and medium enterprises are already struggling with overhead. Adding higher energy costs on top of wage pressures makes the margin squeeze real. Luna: This makes the Fed’s job even harder. They are fighting inflation that isn't just demand-driven, but supply-constraint driven. Lucas: Correct. And you can’t print more electrons. That’s a hard limit. So the economy has to adapt to a higher energy cost baseline. Luna: Does that mean we’ll see slower overall economic growth permanently? Lucas: Not necessarily permanent, but lower potential growth. If energy is more expensive, productivity gains from AI might be offset by higher input costs. It’s a tug of war between innovation and infrastructure. Luna: I wonder if investors are finally starting to price this in properly. Lucas: Some are. Look at the small-cap Russell 2000. It’s down two point one percent over the last five days. These companies are more sensitive to interest rates and energy costs than the mega-cap tech giants. Luna: So the divergence is widening between the AI winners and the rest of the market. Lucas: Yes. And that’s a risk. If the energy costs rise faster than AI revenue, those big tech valuations could come under pressure. Luna: That would be a major correction trigger. Lucas: It could be. We are watching this space very closely. The intersection of energy policy and tech regulation is going to define the next bull or bear market cycle. Luna: Interesting perspective. It makes you look at your own smartphone differently. Lucas: It does. Every swipe has a carbon footprint and a cost basis. Luna: Thanks for shedding light on that, Lucas. Lucas: Anytime. Let’s see where the power flows next week.