Odd Lots (Bloomberg): "The Utilities Analyst Who Says the Data Center Demand Story Doesn't Add Up"

Date: February 2, 2026
Hosts: Tracy Alloway, Joe Weisenthal
Guest: Andy DeVries (Head of Investment Grade Credit & Head of Utilities and Power, CreditSights)

Episode Overview

This episode explores the frenzy around utilities and the power demands of the ongoing AI/data center buildout. While the consensus is that AI will drive unprecedented growth in power and infrastructure—which has turned once-overlooked utility analysts into hot commodities—guest Andy DeVries offers a contrarian, numbers-driven perspective. He argues that current and planned supply is much higher than sober demand estimates, hinting at looming overcapacity, misaligned incentives, and major risks for investors, utilities, and ratepayers alike.

Key Discussion Points & Insights

The New Spotlight on Utilities (01:36–04:17)

• Utilities were long seen as boring, bond-like investments, valued for steady dividends and growth roughly in line with inflation.
• The explosive growth of data centers, driven by AI, has catapulted utilities into a high-demand, high-growth narrative.
• Quote:
  “If you were a utilities analyst…sudden you are very in demand. Right, because all you hear about nowadays is the AI build out and energy constraints on that.” – Tracy Alloway (02:40)

Historical Context & The Analyst’s Perspective (04:17–06:12)

• Utilities weren’t always sleepy—they’ve weathered Enron, TXU, Calpine, wildfires, and other crises.
• Utility analysis pre-data centers was “a lot of rate cases, policy, tracking natural gas prices, renewables displacing coal.”
• Policy—both state and federal—is crucial.

Current Mood & Conference Mania (06:42–07:23)

• Recent utility conferences (e.g., EEI in November) packed to the brim; “CNBC broadcasting from the floor…probably a sign of a top.” – Andy DeVries (07:07)
• Utilities’ projected growth rates have moved from 4–6% to 5–8%—largely attributed to data center demand.

The Bull Case vs. The Contrarian View (07:23–08:52)

• Stocks like XLU have risen, but remain interest-rate sensitive.
• Industry narrative: Data centers will decouple utilities from interest rate sensitivity due to secular growth from AI/digital transformation.

Quote:
“As soon as the Fed starts jacking up rates, ChatGPT comes on the scene and all of a sudden there’s data centers…now you’re taking the leg up on growth.” – Andy DeVries (08:37)

The Core Contrarian Argument: Data Center Math Doesn’t Add Up (08:52–13:16)

Demand Forecasts (09:26–10:17)

• Current data center consumption: ~45 GW
• By 2030: Consensus forecasts (from banks, consultants, BNEF, etc.) average around 90–95 GW (so, +50 GW).
• By 2035: Forecasts approach 160 GW, but estimates are fuzzy.

Supply Side Overlooked—But Critical (10:15–13:16)

• Utilities are working to connect ~140 GW of capacity to the grid in the near term, of which roughly 110 GW is firm, committed supply (apples-to-apples, adjusted for PUE—Power Usage Effectiveness).
• “On the demand side between now and 2030 you need 50 GW; utilities are working on 110 GW.” – Andy DeVries (11:47)
• There’s a risk of overbuilding supply relative to credible demand, especially as commitments multiply.

Tracking & Methodology

• DeVries’ team mixes data from Gmail alerts for demand (old-school “AI agent”) and a diligent analyst (Diego) for utility-sourced pipeline numbers (13:00).

Deep Dive: Interpreting Power Markets & Forward Curves

Texas & the Forward Power Curve (16:20–20:23)

• Texas is seen as the epicenter, with wild demand forecasts (+30 GW peak by 2030 questioned).
• Utilities claim real demand is coming—signaled by $2.5 billion in cash collateral posted by data center customers (19:20).
• But forward power prices are flat—if grid demand were truly going to rocket, prices should be up several dollars, not $1 (20:12).

Quote:
“If you’re going to add 20% to your grid demand…you’d expect [the price] to go up…several dollars.” – Andy DeVries (20:12)

Natural Gas & LNG Exports (17:34–18:57)

• Gas futures curves are downward sloping despite surging US LNG exports and data center growth—contradicting the viral “supply crunch” narrative.

What If the Demand Doesn’t Materialize? Who Bears the Risk? (23:04–25:48)

• If there’s an overbuild and demand lags, someone pays—the ratepayer or the utility/shareholders.
• Best practice: deals that shield residential users (e.g., NiSource/Amazon in Indiana), but many states lack strong protections.

Quote:
“The political risk of having mom and pop bail out Zuckerberg, Bezos, you can’t have that happen.” – Andy DeVries (25:31)

Will Bureaucracy and Construction Lags Save Us? (25:48–28:09)

• Delays more likely on power generation (6–7 years), not data centers (2–4 years)
• Grid can handle more than headline numbers suggest, especially if data center load is flexible (e.g., shifting chemical/refining activity).

The Price of Capacity

• Gas plant construction costs have soared from $1,200/kW to $3,000/kW, but that’s tiny compared to data center build costs (~$40,000/kW).
• “For Big Tech…to spend another three to lock in their gas price or their fuel source, it’s nothing. It’s de minimis.” – Andy DeVries (27:10)

Regional Transmission & the US-China Angle (30:09–32:11)

• Trouble spots: Midwest ISO (“retired the most coal, worst shape”), New England (highest power prices, tightest grid).
• US transmission build-out is slow, but not decisive versus China—power won’t be the bottleneck for the US AI race.
• “From your perspective, power is not going to be the decider here.” – Joe Weisenthal (31:44)

The Data Center Debt Gold Rush: Private Credit Moves In (34:48–39:46)

• Private credit funds are pouring into data center finance—PIMCO made $2 billion in a single day lending on a Meta (Facebook) data center.
• Concern: as returns and competition rise, standards and covenants are likely to deteriorate.

Quote:
“We all know how this ends. Covenants start falling, rates start falling…If you’re Big Tech, who cares if you overspend?” – Andy DeVries (35:39)

• Meta/Big Tech often use off-balance-sheet vehicles for data center debt, possibly to maintain flexibility or to walk away from liabilities.

Shady Circles: Incestuous Financing & Rating Agency Views (39:18–40:25)

• Tech/cloud companies investing in each other and securing reciprocal business (e.g., NVIDIA and OpenAI with "vendor financing").
• Ratings agencies do consider lease obligations as debt for ratings once operational.

Nuclear: Still a Pipe Dream? (40:25–42:21)

• Mega-projects (e.g., Vogtle) now seen as unbankable; only hope is Small Modular Reactors (SMRs) with Big Tech buying and equity investment.
• Trend unlikely to reverse without massive political/economic incentives.

Hosts' Closing Reflections (43:01–47:47)

• Parallels to previous tech booms: huge transformative impact can coexist with bubbles and overcapacity.
• Efficiency gains in AI hardware may blunt explosive demand—the “Jevons Paradox” isn’t destiny.
• The disconnect between market narratives (shortages, bottlenecks) and observed pricing curves (flat/declining forward power and gas).

Quote:
“As he pointed out, okay, ChatGPT came out—2 gigawatts…There aren’t a ton of ChatGPTs out there. That’s one of the most computationally intensive things.” – Joe Weisenthal (46:58)

Notable Quotes & Memorable Moments

• On utilities’ newfound star status:
  “Sudden you are very in demand. Because all you hear about…is the AI build out and energy constraints.” – Tracy Alloway (02:40)

• Breakdown of overcapacity risk:
  “Utilities are working on already connecting almost as much as you need by 2035. There is a lot of supply of data centers coming and it’s very unclear if there’s going to be demand for this.” – Andy DeVries (11:45)

• On Texas grid risk:
  “Texas is a walled-off market…demand estimates are they’re going to add 30 GW by 2030…no way, the forward power curves don’t reflect that at all.” – Andy DeVries (13:41)

• On ratepayer protections:
  “The political risk of having mom and pop bail out Zuckerberg, Bezos…can’t have that happen.” – Andy DeVries (25:31)

• Efficiencies blunting demand mania:
  “What we’ve seen so far is that these things are getting more and more efficient…faster than anyone expected.” – Tracy Alloway (45:11)

Key Timestamps for Major Segments

• 00:00–01:36 — Ad spots / intro skip
• 01:36–04:17 — The utilities analyst's “moment”
• 04:17–07:23 — Historical crisis/opportunity in utilities; policy focus
• 07:23–08:52 — Stock/bond market response, growth narratives
• 08:52–13:16 — Contrarian math: demand vs. supply for data centers
• 16:20–20:51 — Forward power/gas curves; implication for true demand
• 23:04–25:48 — Who pays: customers or shareholders; protection mechanisms
• 25:48–30:09 — Delays in supply, construction costs, actual system slack
• 30:09–32:11 — Grid/transmission, US-China, is power a constraint?
• 34:48–39:46 — Private credit booms in data centers; off-balance-sheet structures
• 39:18–40:25 — Incestuous industry financing; vendor lock-ins
• 40:25–42:21 — Nuclear verdict: only SMRs with Big Tech as anchor?
• 43:01–47:47 — Reflection: bubbles and true demand, efficiency drivers

Conclusion

This Odd Lots episode delivers a much-needed reality check on the AI-data center-utilities narrative. Andy DeVries’ methodical approach shows utilities are racing ahead with supply that may vastly overshoot even aggressive demand forecasts. Meanwhile, market pricing (power, natural gas) remains skeptical of an imminent crunch—contrasting with pervasive hype from industry and media alike.

The conversation closes with a nuanced take: revolutionary technologies can transform society while still wildly overpromising, leading to cycles of overbuild and correction, both for equity—and, increasingly, for debt investors.

For more from Odd Lots, subscribe to their newsletter and join the discussion on Discord (info at bloomberg.com/oddlots).