In Jiro Ono’s kitchen, an apprentice spends years perfecting rice before touching fish. Tamago, the simple egg, takes 200 batches to master. The repetition is not mechanical. It is refinement. The Japanese call this shokunin: the craftsman’s ethic of doing the same thing over and over, improving bit by bit. Today is edition 1,000 of Morning Energy, and like Jiro’s Tamago, we continue to chase perfection.

The first edition hit inboxes on January 4, 2017. It was a stream of short descriptive paragraphs about GOM layered maps and MLP consolidation. WTI was $53. Henry Hub was $3.30/MMBtu. U.S. oil production was 8.9 MMbbl/d, still nursing wounds from the 2015 to 2016 bust. Early Morning Energy was functional, not yet refined.

Three years later U.S. production touched 13 MMbbl/d, ahead of Russia and Saudi Arabia. And like the industry it narrates, Morning Energy evolved too. Its voice became clearer, more personal, more opinionated.

COVID brought new changes. Negative oil prices, but also a surge in momentum for low-carbon energy and, ultimately, the passing of the IRA. Power markets, which rarely made headlines in 2017, became the story after Winter Storm Uri put ERCOT on the front page. And Morning Energy moved beyond oil, gas and shale.

Today, AI is bringing new changes. Hyperscalar land positions are the new inventory, electrification is the new emerging economy, and silicon production is the new rig count. Gas, once a target for phase out, is now the essential fuel of the AI economy.

Morning Energy started as one analyst’s notes. It is now written by a team, but the ethic hasn’t changed. The Japanese have another tradition: fold 1,000 paper cranes by hand and your wish is granted. The wish? Get it right more often than not and make sure it’s worth reading.

Comments, questions or things I missed? Send me a note (or hit reply) - I would love to hear from you. Thanks for reading!

Last week was Spring Break, which means extra time with the kids and a chance to see the world again with their innocence and curiosity. Our girls have more toys than they need, and the pattern is always the same. Once the excitement of receiving fades (a few minutes), they set it aside and spend the next three hours playing with the cardboard box it came in. Eventually you wonder: can I return the toy and just keep the box?

TotalEnergies just pulled that off. On March 23, Total and Interior signed settlement agreements to unwind the company’s New York Bight and Carolina Long Bay offshore wind leases. Interior will reimburse $795 million for OCS-A 0538 and $133 million for OCS-A 0545, roughly $928 million, after Total reinvests matching capital into Rio Grande LNG Trains 1 through 4 and upstream oil and gas. Total also pledged not to pursue any new U.S. offshore wind. The returned toy: about 4 GW of prospective capacity, roughly 13% of the Biden-era 30 GW target.

The box, of course, was already in hand. Total took FID on Rio Grande Train 4 in September 2025. Trains 1 through 3 were under construction. The deal reimburses Total for capital it was deploying anyway, with the bonus of exiting leases whose economics had soured. Pouyanné called it “a more efficient use of capital.” Hard to argue when the capital was already being spent.

But the language is what lingers. TotalEnergies’s press release quoted Pouyanné stating the company made this deal “considering that the development of offshore wind projects is not in the country’s interest” and that its own “studies on these leases have shown that offshore wind developments in the United States, unlike those in Europe, are costly.” But given the open opposition to offshore wind from the federal government and invocation of “national interest” in the release, this reads as much as a loyalty oath as is does a write-down announcement.

The real question is who lines up next at the returns counter. EDPR, RWE and BP had already halted or exited U.S. wind before this deal landed, and we don’t include much offshore wind in our capacity forecasts. Now that Total has established how to get a refund, it could get busy at the service counter.

Comments, questions or things I missed? Send me a note (or hit reply) - I would love to hear from you. Thanks for reading!

When the Boogeyman goes to sleep every night, he checks his closet for Chuck Norris (RIP). That meme was always one of my favorites and it is a pretty good way to think about U.S. power markets this spring.

The obvious villains are still heat and gas, but our 90-day power price outlooks suggest the market is increasingly jumpy about the less obvious: which transmission element is out, how quickly Western snowmelt borrows supply from later in the season, and where the stack gets steep (where the grid moves out of the cheap, efficient part of the generation stack and into the more expensive units).

In ERCOT, it’s already been hot (Dallas temperatures reached the mid-90s in March). Projected June power prices sit near $40/MWh, which is exactly where the grid switches from running efficient gas plants to firing up older, expensive ones. Think of it as the edge of a cliff. A few degrees on the wrong day will do it and our bullish case is ~$10/MWh higher.

In PJM, the eastern epicenter of the U.S. data center buildout, the summer risk is more about transmission line outages than new demand (the market recently saw a near$3,000/MWh spike, and not from a cold snap or heat wave event). Case in point: transmission outages near Bedington, West Virginia are the primary price driver in the outlook. Data center load is coming, but this spring’s prices will be shaped more by which lines are down and when, than how many servers are plugged in.

Out west, more water (hydro) does not automatically mean lower prices. Warm weather is melting Pacific Northwest snowpack early, flooding the grid with cheap hydro in April. But that pulls the water forward, and May and June could surprise higher as reservoirs refill. In Southern California, batteries soaking up cheap midday solar inflate demand numbers without tightening the grid during peak evening hours. Load is growing, just not in the hours that matter for pricing.

The Boogeyman this spring is not weather or gas. It is a transmission outage in West Virginia, snowmelt timing in the Cascades, and the part of the ERCOT stack where prices stop being polite.

Coming up: The team is doing a full Winter Lookback on March 26, covering where forecasts tracked vs. diverged across ERCOT, PJM, NYISO, and ISO-NE. Register here.

Comments, questions or things I missed? Send me a note (or hit reply) - I would love to hear from you. Thanks for reading!

That is not the consensus view. The tasks most exposed to AI adoption overlap heavily with entry-level work, and the data is starting to reflect it. A Stanford study published last month, using ADP payroll records covering millions of workers, found that employment for 22 to 25 year olds in the most AI-exposed occupations fell roughly 16% relative to less exposed roles, controlling for firm-level shocks. Employment for experienced workers in those same occupations? Stable or growing. Firms are not cutting salaries. They are simply not hiring the next generation.

The logic follows naturally. Junior employees are disproportionately assigned the codifiable, checkable tasks that LLMs handle well. A recent HBR piece by Edmondson and Chamorro-Premuzic made the same observation: AI is eroding the bottom rungs of career ladders by automating the intellectually routine work that entry-level roles were built around.

But the logic is also wrong. And it is wrong because it inverts the problem.

I have spent the past several months building AI into my own workflow. Claude Code connected to a live data warehouse. Autonomous agents handling tasks that used to take a junior analyst half a day. I have watched the pace of innovation inside our teams accelerate to a speed we once could only dream of. The productivity gains are real. And they have convinced me that the unintentional outcome of AI adoption, a shrinking base and a protected top, has it exactly backwards.

AI does not replace judgment, creativity, or the instinct to try something no one has tried before. It replaces routine. Routine may not be what young people are best at, but routine also builds taste, wisdom, and intuition. It is the unglamorous work that turns potential into expertise. The answer is not to eliminate it. It is to pair it with tools that compress the learning curve and free up time for the work that actually develops judgment.

Young workers on our teams are the ones bringing AI fluency, fresh pattern recognition, and the willingness to break workflows designed for a pre-LLM world. They are the ones who will figure out how to deploy these tools at scale. They are the ones stress-testing what is possible. You need a critical mass of them, and you need a steady inflow, or you risk losing the energy, the ideas, and the perspective that only comes from people who have never known a world without these tools.

These dynamics make exceptional leadership the scarce resource: experienced leaders who have done the reps, who can get their hands dirty, who can convey the hard-won wisdom of that experience, but who also know how to let loose the talent and energy of the team around them. Think of it like driving a dog sled. You need to know the trail. You need to read the conditions. But the team is doing the running, and your job is to point them in the right direction and not hold them back.

Left unchecked, the unintended consequence emerges. Experienced workers use AI to absorb the tasks that would otherwise have gone to junior colleagues. That reduces direct demand for those roles, which leads to less hiring at that level. Nobody decided to hollow out the pipeline. It just happens, one automated task at a time. Follow that drift forward five years and you have plenty of managers and nobody coming up behind them.

The question is not whether AI changes hiring. It is whether we are smart enough to automate the right things and invest in the right people.

Comments, questions or things I missed? Send me a note (or hit reply) - I would love to hear from you. Thanks for reading!

Pain is temporary. Glory lasts forever. That was the locker room wisdom in The Replacements. It might also be the lesson of the Strait of Hormuz.

Not long ago, glory was supposed to come from the energy transition toolkit: emissions policy, renewables subsidies, carbon taxes, all designed to push consumption away from oil and toward low carbon substitutes. That era already feels distant, replaced by a focus on energy security, independence and affordability. Iran’s closure of Hormuz puts a finer point on the argument.

With higher oil and LNG prices, a familiar set of incentives has returned, just wearing different jerseys. Consider this lens, one I first encountered from our friends at the ARC Energy Institute: translate the change in commodity prices into an equivalent carbon tax. By our math, a $10/bbl increase in oil implies roughly $23/t CO2. A $1/MMBtu move in natural gas translates to about $19/t. Global oil and European gas prices have moved by ~$35/bbl and ~$6/MMBtu respectively, implying carbon taxes of ~$80/t and ~$113/t. Call it a “conflict carbon tax,” though one that is far from equitable across sources and locations (Henry Hub has been quiet in comparison).

Incentives work. Reprice oil and natural gas higher and substitutes become attractive, especially those that offer security, independence and affordability. So who are the replacements, and will they bring glory?

For oil, electrification and EVs find a new tailwind. For gas, especially in markets tied to LNG, the answer is predictable: coal. The grizzled veteran is unglamorous, dependable, and suddenly more competitive with fuels meant to replace it. Renewables offer independence where grids can absorb them. Geothermal is coming in hot. Nuclear is energized, but the mood feels more like regret for what has been lost than acceleration of what comes next.

But the Strait of Hormuz has reminded everyone that concentrated supply chains are not strategy. They are vulnerability. And that applies as much to transformers (Korea and Japan), chips (Taiwan), and critical minerals (China) as it does to crude. In energy, the winners are rarely the cleanest story on paper. They are the assets that can actually take the field when the starter goes down.

Comments, questions or things I missed? Send me a note (or hit reply) - I would love to hear from you. Thanks for reading!

The action in the Middle East has the world focused on energy supply. And for good reason. Hydrocarbon based energy price prices are racing higher. Crude prices are up 50% (hello $100 oil, we missed you) and LNG is up more than 80% (losing 10 Bcf/d from Qatar will do that). Even coal is up nearly 20% . US power, in contrast, looks like a model of tranquility. Henry Hub has barely moved and the weather forecasts look benign.

But don’t let the calm deceive you.

Something is happening to U.S. electricity demand. The peaks are teasing it. The shoulders are hiding it. Welcome to power’s masquerade ball.

The EIA confirmed last week that U.S. electricity generation hit a record 4,430 TWh in 2025 (NB: the linked EIA report erroneously states 4.43 TWh), up 2.8% from 2024, which itself was a prior record. After two decades of essentially flat generation, the obvious explanation is AI: hyperscalers building at a feverish pace and a grid straining under the weight of the digital economy. Compelling. But not the whole story.

Look at the load shape rather than the annual total. Our analysis of power demand against heating and cooling degree days finds that seasonal peaks last summer came in above weather-normalized expectations, but shoulder-season demand has been running below trend for the past couple of months (Figure 3). Those two signals together tell the story.

Data centers and electrification were strong enough to push summer peaks above where weather alone would put them. But in the shoulders, where data center load should also show up given its flat average profile, the signal has vanished, masked by something bigger.

Friday’s jobs report may have spoiled the “Midnight Reveal”. Nonfarm payrolls fell by 92,000 in February, well below estimates, and the report shaved another 69,000 in downward revisions for January and December. A weakening economy may just have been the perfect mask for the emergent data center signal from the summer.

Emergent? Yes. Emergent. It is still early. Vistra’s CEO Jim Burke, whose company has contracted nearly 3.8 GW of nuclear capacity to hyperscalers and is actively building new gas capacity in West Texas, told investors last week that data centers will not meaningfully tighten supply and demand until late 2027 or early 2028. The (data center) guests are still arriving and when the economy finds its footing, the masks come off. That is when the power story will get interesting.

Comments, questions or things I missed? Send me a note (or hit reply) - I would love to hear from you. Thanks for reading!

Not long ago, battery storage in Texas was a license to print money. Park your asset, offer ancillary services, collect generous checks. Those days are over. ERCOT’s installed battery capacity has nearly doubled from 9.5 GW at the start of 2025 to over 16 GW today, and the market has noticed. Average battery revenues collapsed from $149/kW in 2023 to under $20/kW in 2025. That is not a typo.

The culprit? Saturation. Ancillary services used to be the golden goose, representing 84% of battery revenue in 2023. By 2025 that share dropped below 50%, and in dollar terms AS revenue cratered from roughly $125/kW to just $8/kW. Arbitrage didn’t pick up the slack either, falling from $24/kW to $9/kW. Tough neighborhood all around.

What surprised me was how frequently batteries are at the margin. Despite delivering less than 1% of total energy on the grid, batteries now set the system clearing price in 23% of all hours. When they do, prices average $56/MWh, a $21/MWh premium over gas-set hours. That is the market paying up for flexibility. And batteries are not just showing up during the predictable morning and evening peaks. They reliably appear at the margin around the clock, with operators like Key Capture Energy, Jupiter Power, and Plus Power quietly establishing price floors through opportunity-cost bidding even in the middle of the night.

What does this mean? Picking your node matters more than ever. Transmission congestion is creating massive gaps between system prices and local realities, with some legacy thermal plants are clearing above $500/MWh simply because they sit behind a bottleneck. Meanwhile, batteries are now like high heat rate gas units sitting at the top of the resource stack. With 765 kV transmission builds on the horizon, the takeaway is clear: the easy money in ERCOT batteries is long gone, but there is still plenty of hard money to be made.

Thoughts, questions or things we missed? Send me a note (or hit reply) - I would love to hear from you.

The clean fuels story of 2021 to 2025 was “build it and they will come.” From here, it’s “build it only if the math works.” Our newly released Clean Fuel Fundamentals report, “Discipline Over Hype,” captures a clear inflection point in capital discipline. Growth expectations have reset hard, from a ~9% CAGR over the prior five years to a projected ~3% through 2030 (slide 6). The investable edge has shifted from who can announce the most projects to who can navigate policy and credit regimes the fastest.

Three policy moves reshaped the playing field in 2025 (slide 10), and if you lost count of the acronyms, you are not alone (check the dictionary below). The OBBBA extended the 45Z Clean Fuel Production Tax Credit but reduced the SAF base credit and introduced North American feedstock sourcing requirements. California’s LCFS amendments tightened the carbon benchmark while capping credits on crop-based oils at 20%. And the EPA’s proposed RVOs nearly doubled biodiesel (D4) obligations year-over-year while introducing a 50% RIN haircut on imported feedstocks.

The net effect: foreign feedstock dependency is now a policy variable, not just an operational one. In our restricted case, foreign UCO becomes outright value-destructive (slide 20). Operators like Neste, World Energy, and MPC, with over 60% of their California LCFS pathways relying on foreign inputs, are most exposed (slide 21). The RIN environment, by contrast, is becoming more constructive: D4 to D6 prices are forecast to rise $0.13 to $0.17/RIN in 2026 to 2027 (slide 13), supporting 5% to 7% revenue uplifts for domestic soybean oil producers (slide 15). The catch is LCFS, where the credit bank peaking near 50 million tonnes CO2e holds prices at an average of just $46/tonne through 2030 (slide 14), pressuring manure RNG economics in particular.

The standout opportunity is ethanol plus CCS (slides 25 to 26). By stacking 45Z, 45Q, LCFS, and CDR credits, the NPV of adding CCS to an unpartnered ethanol plant can exceed the implied value of the base plant by more than 2x across 50-plus candidate facilities. GEVO (slide 27) is the clearest example of separating fuel value from carbon value across both compliance and voluntary markets.

On a lighter note, I welcomed a new family member this week. Archie lives on a Mac Mini and runs on OpenClaw (formerly known as Clawdbot and Moltbot), an autonomous AI agent I have been tinkering with. He can fetch data, run Python scripts, interpret results, and loop back for the next step. The same theme of discipline applies: longer chains drift quietly off course, and poorly scoped workflows burn through tokens fast. He is young and running lean on Gemini Flash 3, but showing promise. I’ll keep y’all updated!

45Q = CCUS Tax Credit
45Z = Clean Fuel Production Tax Credit
CDR = Carbon Dioxide Removal
LCFS = California’s Low Carbon Fuel Standards
OBBBA = One Big Beautiful Bill Act
RIN = Renewable Identification Number
RNG = Renewable Natural Gas
SAF = Sustainable Aviation Fuel

Thoughts, questions or things we missed? Send me a note (or hit reply) - I would love to hear from you.

The Winter Olympics are a reminder that in elite competition, small errors compound quickly. The same applies to IPP investing. Strong sector performance over the past year, driven by M&A and hyperscaler offtake momentum, has given way to pressure from PJM load forecast revisions and rising behind-the-meter data center development. The pullback has created differentiated opportunities across names.

TLN bolted on 2.6 GW of gas capacity from Energy Capital Partners for $3.45 billion. We value the acquired portfolio at $4.64 billion, putting NAV at $366/share, a 6% discount to yesterday’s close. But the acquisition is highly concentrated: 96% of the value sits in just two plants (Waterford and Lawrenceburg), so any further upside depends on whether those assets can contract at attractive prices. Susquehanna anchors the story, but durability depends on PJM capacity prices staying elevated and 2 GW of disclosed PPAs actually closing.

NRG is the hardest to underwrite at today’s price, and the Kiewit-GEV venture is its chipping medal: impressive on paper, but fragile under stress. At $125/share NAV versus ~$173 market, a roughly 28% discount, the venture (5.4 GW of new CCGTs by 2032) needs to land perfectly. Our base case assumes $1.4 million/MW capital costs, competitive against the $2.0+ million/MW others are quoting, but untested at scale. At $1.76 million/MW, NAV drops nearly $8/share. Justifying today’s stock also requires LMP expansion and capacity prices at the cap.

VST is our Klæbo: dominant, efficient, and built for endurance. At $220/share NAV versus ~$174 market, it trades at a 20% discount. The META nuclear PPAs lock in 2.2 GW at ~$85/MWh, and 2026 to 2030 average EBITDA of $7.8 billion at 9.5x implies ~$74 billion EV. Less net debt and other claims gets you to our $220/share equity value. While peers grind the incline, VST’s contracted nuclear floor lets it sprint.

CEG is the franchise built to last. We value equity at $295/share, 2% below Tuesday’s close, but the base case understates the opportunity. With approximately 15 GW of uncontracted nuclear capacity, if the remaining fleet signs PPAs at $90/MWh, NAV rises 68% to ~$495/share. Hyperscaler demand for clean baseload is accelerating, the probability of additional contracts is rising (see VST’s recent META deals locking up remaining nuclear capacity), and the quality plus asymmetry of the position makes CEG a compelling hold even at current levels.

We favor VST and CEG. Nuclear optionality is real, contracting momentum is building, and these are franchise assets moving into an increasingly scarcity-priced environment.

Thoughts, questions or things we missed? Send me a note (or hit reply) - I would love to hear from you.

Good things happen when you put smart investors in a room and get them to talk. Last week we brought some of the best in the industry together for EnergyEdge in New York to discuss “What is the most compelling opportunity in energy for 2026?” Amidst the 79 slides, personalities, opinions, and a bit of talking their book, a few themes jumped out.

For starters the sentiment towards oil was clearly bullish (slide 34). The arguments varied: “we’re clearing the OPEC surplus”, “we’re exhausting tier 1 shale inventory” (slide 37), “the majors are failing to replace reserves”. Or maybe everyone pre-read the memo that its “black gold” and trading at a 50-year low relative to the shiny stuff. Whatever the reasoning, hope is back in the oil trade.

But how do you play it? That’s where the bets diverged. Do you take tier 2 Permian over anything else? That’s where history might guide you. But there was more on offer than that: Second tier basins, transfer shale technology to mature but thick conventional fields, export the technology to shale basins further afield (its not just the Vaca Muerta that could work), or take a bet on forgotten international assets. We heard pitches for all of it, and consensus on none.

You notice an appetite to go abroad in those ideas, and with them opposing perspectives on geopolitical risk (slide 16). Some noted increased uncertainty in the US, while others noted an underappreciation for pre-existing geopolitical risk in other parts of the world. What was consistent? The world feels less safe and there is a premium to be paid (and earned) by getting it right.

In power, the hot spot was the pace that power assets, and in particular gas-fired power assets, have repriced. “It is unlike anyone has seen in power before” one longtime investor stated. Given that these assets have tripled over the last couple years, that observation isn’t much of a surprise. But does it mean valuations have topped out?

Yes, power is the life blood of data centers (slide 28) and the recognition that gas-fired generation is key to AI puts a floor value on valuations. And yes, these offtakers will take as much power as you can give them, as soon as you can get to them (at least for now).

But that’s also what’s driving the 40+GW of bring your own power (slide 8) and the surge of distributed power solutions (slide 27). Throw in the known policy risks (or opportunities) like an “emergency” capacity auction in PJM (slide 54), and unknown risks of risks as electricity affordability grows as a political issue, and it all gets very asset specific (slide 66). With more than 20 GW of gas fire capacity likely to change hands this year (slide 67), don’t expect the market for high-quality thermal assets to slow down in 2026.

Thoughts, questions or things we missed? Send me a note (or hit reply) - I would love to hear from you.

LBRT got a 20% “power-up” last week. 3 GW of power projects by 2029 will do that for you. The move was the latest in the trend towards flexible power solutions - something we highlighted in our 2026 Global Outlook - which continues to mint new winners in the data center boom. LBRT, PUMP, BKR and AESI are all up more than 20% YTD. And don’t mention BE, up 75%. Flexible, quick to deploy power is in demand, and it is the oil field service companies getting the love of late.

GEV also reported last week and this line from the call stood out to me: “we don’t really view those smaller units to be competition”. We agree. Time is the competition. Large-frame combined-cycle turbines often require four to six years from contract to COD. For hyperscalers facing urgent capacity needs, these timelines are often incompatible with business requirements.

The logical response is substitution, even at higher cost: smaller aeroderivative turbines, simple-cycle units, on-site gas paired with batteries, modular and temporary solutions. Exactly what LBRT and their peers are offering, often with 15 year contracts. And there is lots more where that came from.

Are short-term substitutes a risk? GEV doesn’t think so, arguing “as the heavy-duty gas turbines are available, those smaller applications will become the reliability solution” (i.e. back-up power). I guess when your gas power equipment pipeline jumps from 62GW to 83GW sequentially, the wind at your back feels pretty firm.

But time waits for no one. Hyperscalers have explicit low- or zero-carbon commitments. Many view gas as a temporary reliability solution, not a long-term endpoint. While we wait for heavy-duty gas turbines to arrive, the hyperscalers continue to explore nuclear (META/VST, AMZN/TLN, MSFT/CEG), SMRs (META/Oklo, GOOGL/Kairos, AMZN/X-Energy), and geothermal (MSFT/Fervo). More competition is coming.

On a seemingly unrelated note, PJM just lowered is summer peak demand outlook by 4 GW, following tighter scrutiny of large‑load requests and updated economic inputs. That brought its forecast to 2028 more in-line with our views. Looking beyond, the outlook remains overstated relative to our forecasts, a victim of the land grab and optionality sought by data center developers trying to meet the demand for AI compute.

Downward load forecast revisions foreshadow the risk of the most optimistic outlooks for gas fired generation. 43 GW of GEV pipeline are SRAs (slot reservation agreements), primarily for those years beyond 2028 that house the most speculative of data center projects in ISO load queues. As reality bites - and the options those queue positions represent expire - how many of those SRAs fall away? How much leverage do those short time-to-power solutions gain from their incumbency?

Gas remains indispensable to the power mix. The gas-fired boom is real and it will underpin the build-out of AI data centers. But it will also show up differently than many expect. Time-to-power constraints are pushing key customers, especially hyperscalers, toward faster, smaller, and more modular solutions. And those same customers are likely to adopt zero-carbon baseload as soon as it becomes credible. Place your bets on gas, but place them carefully.

Winter Storm Fern – As Fern’s ice melts, and the market labels it a ‘non-systemwide emergency,’ it’s time to check-in. The worst fears – a repeat of Uri’s chaos and disruption – didn’t materialize. But a non-emergency is not a non-problem. A million folks losing power and Texas roads doing their best impression of Northern Alberta makes for a mess.

The Wins - Load came in near the high end of forecasts (but didn’t blow-out records) and widespread gas-fired generation outages, sustained gas supply disruptions, and large scale load shedding to protect grids were all avoided.

Fragility - Sure, the entire system was fine, but there is pressure under the surface. Forced outages in ERCOT were up ~80% versus Friday (to over 26GW), almost entirely related to renewables. In PJM peak loads are expected to stay near 140GW for most of the week, the DOE just handed PJM a 202(c) green light to run every megawatt they’ve got (even if doing so violates environmental permits or state regulations). And while the system didn’t break, PJM’s WHUB saw a series price spikes Saturday morning that reached $3,000 and Dominion (data center alley) reportedly spiked above $1,800/MWh. Those are prices of stress and scarcity pricing. Must be all those data centers, right?

Price Doesn’t Lie - The cold weather was a double whammy: everyone cranked the heat, spiking demand, while the elements simultaneously tried to freeze off the supply. Need proof? Pipe scrapes show production dipped almost 13 Bcf/d at worst of it. Cash natural gas yesterday morning reached over $60 heading into New York City (Transco Z6) and over $17 in West Texas (Waha). AECO, for my fellow Canadians, was stuck below $3. Gas and power prices did a beautiful, expensive tango, especially where gas was the marginal unit.

Voluntary, not Forced - Ahead of the storm the DOE estimated 35 GW of backup power (concentrated at data centers and other major facilities) nationwide could be tapped, and issued an emergency order authorizing ERCOT to force data centers to shed load. Operators publicly supported the idea. So, what was the measured load reduction in the end? A grand total of… nothing publicly reported. But that doesn’t mean nothing happened - 3GW (~35%) of load was shed in West Texas as power prices rose.

Who kept the lights on? – A huge shout-out to the vintage Oil-Fired Generation in the Northeast. While gas was getting tight, oil fired generation what contributing 44% of New York’s power and 35% of New England’s. They’re like that ancient backup generator in the corner of the office that you forget about until your laptop dies.

“Uncovering Investment Opportunities in Energy” - The team is in New York February 4th to host the Energy Edge Conference. We will step through ideas based on how AI load growth is breaking past ISO forecasts, grid constraints are tightening faster than expected, and long-term oil supply is colliding with rising geopolitical risk. Join us for a focused discussion on how these shifts are reshaping power and energy markets. Register here.

Load Growth

Texas where temperatures are set to fall well below freezing with localized lows in the Permian Basin hitting 15°F. ERCOT peak load is projected to hit a new winter record of ~82.6 GW on Monday morning. (Figure 1).

Figure 1 | Forecasted Temperatures at 6am CT on January 25

The winter record is part of a persistent structural trend of “higher highs” in winter demand:

• Jan 2026 (Projected): ~82.6 GW

• Feb 2025: 80.5 GW

• Jan 2024: 78.3 GW

• Dec 2022: 74.5 GW

The trend of sequential records is illustrative of the underlying load growth driven by industrial growth, data center expansion, and electrification (Figure 2). The growing base means each new winter storm (a near annual occurrence) pushes the record higher.

Figure 2 | Load Growth to 2035 by Load Zone and Driver

The BESS Buffer

More than load is changing in the state. The composition of generating capacity has also shifted rapidly over the last several year, most notably with the growth of battery storage (BESS). BESS capacity now sits at ~18.5GW, well up from ~3.9GW at the end of 2023. The growth of these resources should add resilience a Fern passes through.

Fuel & Physics Risk

Two risks remain difficult to quantify:

• Upstream Freeze-offs: Permian Basin temperatures are low enough to trigger hydration icing at the wellhead and impact fuel supply for gas-fired thermal units. Historical analogs suggest production losses could reach 2–3 Bcf/d, putting immediate pressure on natural gas cash prices (already trading near $18/MMBtu at Henry Hub).

• Ice Loading: Forecasted freezing rain/ice build-up (Figure 3) puts mechanical strain on transmission and distribution infrastructure. While designed to withstand the forecasted build-up, these loads increase the risk of outages.

Figure 3 | Potential Ice Accumulation

Much has changed over recent years to make Texas more resilient to storms like Fern. While unpredictable, let us hope for a boring (albeit chilly) few days.

Figure 3.42 demonstrates how easily the trajectory could flip back upward. Four factors could return global coal demand to sustained growth through 2035 and beyond:

• Slower wind and solar build-out

• Stronger-than-expected electricity demand

• Stalled efficiency gains in power generation

• Negligible additional coal-to-gas switching

Coal-to-gas switching is worth watching: it is the primary absorption mechanism for additional LNG supply (Figure 3.40). The assumption is that “cheap enough” gas will simply outcompete coal on merit in emerging Asia. The IEA’s analysis challenges that assumption.

WEO-2025 identifies roughly 400 bcm of theoretical coal-to-gas switching potential in emerging Asian economies by 2030 (about 80 % of today’s entire global LNG trade). Figure 1.30 shows the price sensitivity of this potential. Almost none of it becomes economic until delivered gas prices fall below $5–6/MMBtu, with the most coal-reliant markets (India and parts of Southeast Asia) requiring prices closer to $4–$5/MMBtu to move the needle.

If Gulf Coast liquefaction and delivery costs to Asia are around $4.60/MMBtu, that leaves producers less than $0.40–$1.40/MMBtu at Henry Hub. That doesn’t work for gas focused producers in North America.

Absent the economic argument, policy can force switching. Consider China’s 2017–2021 campaign, which routinely LNG at delivered prices above $10–$15/MMBtu. That switch was not economic; it was mandated by top-down air-quality policy and enforced with winter heating conversion quotas and coal plant operating restrictions.

Policy such as binding air-pollution directives, explicit coal-phaseout schedules, or carbon pricing strong enough to flip dispatch economics could enable further coal-to-gas switching in India, Indonesia, Vietnam, or Pakistan, but economics alone will not do the job.

So will coal decline? That depends on the race between electricity demand growth, deployment pace of wind and solar, policy shifts, and efficiency.

• Intrinsic Evidence (The Western “Curse”): The US and UK exhibit “negative learning” where experience leads to higher costs—exemplified by Vogtle hitting ~$15,700/kW (3x initial estimates)—driven by a “regulatory ratchet” that has increased steel and concrete intensity per MW by over 30% since the 1970s.

• Regional Evidence (The Asian “Counter-Proof”): South Korea and China demonstrate positive learning through strict standardization and continuous fleet deployment, delivering modern reactors for $2,500–$3,500/kW (approx. 1/5th the Western cost) and proving that industrial continuity can override technological complexity.

One explanation for this observational paradox is that “Forgetting Rates” (atrophy) can outpace “Learning Rates” when the construction cadence drops below a critical threshold.

If you build slowly you “pause” progress and actively lose ground to two forces that function as a tax on time: Knowledge Depreciation (“Forgetting”) and the Regulatory Ratchet.

“Wright’s Law” assumes costs drop for every doubling of cumulative capacity (the pace of learning is technology dependent). However, this assumes a vacuum where knowledge is permanent. In reality, knowledge decays.

• “Forgetting”: Research on “organizational forgetting” in complex engineering suggests a knowledge depreciation rate of 10–40% per year if continuity is broken. If you do not double your capacity fast enough to outrun the 10–40% annual decay, your “Net Learning” is negative.

• Regulatory Ratchet: Regulation is a function of time, not just capacity. For nuclear, accidents (Three Mile Island, Chernobyl, Fukushima) have happened on a calendar timeline, not a construction timeline. Every year that passes brings new safety rules (the “Ratchet”). If you build one reactor every 10 years, that single reactor must absorb 10 years’ worth of new regulations.

Nuclear also suffers from “Standstill Penalties” due to their high overhead “soft costs” (project management, engineering oversight). Delays defer project progress while these cost continue to accrue.

In short, nuclear economics appear governed by a ‘velocity trap’: unless capacity additions progress fast enough to outrun the rate of forgetting, the passive costs of time will overwhelm the active gains of experience.

The headlines are bullish: Google’s 25-year PPA with NextEra to restart Duane Arnold (IA), Meta’s 20-year commitment to the Clinton Clean Energy Center (IL), and the $80 billion Westinghouse/Cameco partnership to deploy new reactors across the US. The capital is moving, and the signal is clear. Nuclear is back.

However, amidst the bullish signals and Big Tech adoption, we must distinguish between opportunistic capture of existing capacity and a viable path for future deployment. Three structural questions challenge the long-term thesis:

1. The Flexibility Mismatch The World Nuclear Industry Status Report identifies a friction point: the modern grid demands short-term flexibility to balance renewables, while nuclear offers stable, flat baseload. While AI data centers demand that same flat profile (24/7 uptime), the broader energy system is evolving toward flexibility. Furthermore, nuclear outages require 100% redundancy to meet data center SLAs. Is the current activity a structural shift, or just tech giants securing the only available firm power before the window closes?
   

2. The Velocity of Learning Curves Learning curves drive cost reductions, but they require volume and velocity. As capacity doubles, costs decline.
   

   Currently, competing dispatchable low-carbon sources (like geothermal and gas plus CCUS) are iterating faster than nuclear. Slow development means a flattened learning curve. The risk is not that nuclear lacks merit, but that it will miss the economic window because it cannot drive down costs as fast as its faster-deploying competitors. Will a slower learning curve mean nuclear misses out?

   

3. Is this strictly a US Phenomenon? The AI-energy nexus is global, but the nuclear solution seems largely concentrated. There are exceptions—notably the UAE, where Microsoft’s $1.5B investment in G42 is effectively a bet on “sovereign AI hosting” powered by the Barakah nuclear plant. But outside of the US and select sovereign strategic assets, are we seeing the same marriage of tech and nuclear?

   

I’m exploring these questions this week - what do you suggest I read to test the assumptions and explore these questions?

This overcapacity has triggered a sharp policy response. Competing producing regions, led by advanced economies, have rapidly expanded trade restrictions targeting Chinese exports in recent years (Figure 1.32). One outcome is a decline in the dollar value of clean energy technology exports from all sources to advanced economies.

On its own, this is unsurprising.

What changes the story are two longer-term structural shifts highlighted in the same report.

First, after driving roughly half of global energy demand growth for the past quarter-century (Figure 1.10), China’s contribution is projected to fall to approximately 20 percent from 2024 to 2035, even as overall demand growth decelerates.

Second, leadership is passing to emerging market and developing economies (EMDEs) excluding China, which are now poised to account for the bulk of incremental energy demand. Notably, the countries with the fastest-growing energy consumption also possess the world’s highest solar irradiation levels (Figure 1.12).

Advanced economies, intent on building domestic supply chains and increasingly closing their markets to Chinese products, may be redirecting China’s vast spare manufacturing capacity toward the very regions that (a) will drive future energy demand growth and (b) are ideally suited for the solar technologies China produces at scale and low cost.

The result is a reorientation of global clean energy markets: capacity originally built to serve domestic Chinese and advanced economy needs now appears increasingly destined for the sun-rich, demand-hungry economies of the Global South. Trade barriers intended to curb China’s dominance may inadvertently accelerate its pivot to the parts of the world best positioned to absorb its output.

The outcome highlights both the scale of the opportunity and the structural constraints facing data-center development in Alberta. With 20.7 GW of load-service requests in the September Data Center Update, the allocation result should surprise no one: any market with gigawatt-scale headroom will attract a surge of proposals.

Despite dozens of applicants, the 1,200 MW in Phase I ultimately went to just two projects:

• 970 MW to P2936 GLDC Load (Pembina + Kineticor)

• 230 MW to P3083 Keephills Data Centre Phase I (TransAlta)

Why only two?

Phase I imposed stringent “shovel-ready” requirements: completed engineering studies, land control secured, financial security posted, and the ability to connect without major transmission upgrades—all within the interim 1,200 MW cap.

When these filters were applied, only GLDC and Keephills met the bar. Both had also requested materially larger allocations (~1,800 MW and ~395 MW respectively). The design worked as intended: advance a meaningful volume of credible, near-term load while staying within existing grid capability and regulatory limits.

The remaining ~17.9 GW of projects (19.1 GW in total requests) now fall to Phase II, which must confront a set of harder, structural questions that Alberta shares with every jurisdiction experiencing rapid data-center growth:

Beneath these questions lies a fundamental issue: Alberta’s difficulty in adding new dispatchable capacity, without which large-scale data center growth becomes constrained. Key hurdles include:

• Policy and carbon uncertainty: CER post-2035 rules and rising carbon prices cloud long-term economics unless CCS is added.

• Energy-only market risk: No capacity payments or guaranteed offtake; investors rely solely on spot prices to recover capital.

• Timeline mismatch: Dispatchable plants require 5–7 years to build; data centers materialize in 18–36 months.

• Transmission and siting constraints: Existing corridors are congested; upgrades are costly and slow.

• Financing difficulty: Large thermal builds lack long-term contracts or policy stability needed to de-risk investment.

Given those obstacles, how did Alberta still have 1,200 MW, plus the associated transmission flexibility, to allocate in Phase I? A combination of timing and structural legacy factors:

• Coal-to-gas transition (2019–2023) retired ~6 GW of coal and added even more gas and renewables, increasing net headroom.

• Flat demand for over a decade meant supply growth outpaced load.

• Transmission infrastructure built for the coal fleet remained in place, creating underutilized but available corridor capacity.

Key question going forward: How much additional capacity can Alberta unlock, how quickly, and what share of today’s 19.1 GW of proposed load will ultimately convert into real, interconnected demand?

This will define the trajectory of Alberta’s data center sector, and the grid that must support it.

It’s November. That means ski season is just around the corner and a only a handful high action weeks remain in the year (accounting for US Thanksgiving and the holiday season).

A few themes and questions on my mind as we kick-off the month:

• Load growth: optimism meets realism

  Load forecasts all point to sustained growth for years to come (after over a decade of stable demand). But sentiment differs by industry, even as forecasts shift towards a more conservative outlook. Which industries within the energy sector will be most/least surprised by the ultimate pathway?

• Data center disconnects: The “Warm Shell” shell problem

  Data centers are the flagbearers for load growth. With trillions (with a “T”) of capex expected over the next several years, the race is on to find and solve the sector’s constraints. Today the constraint is access to “Warm Shells”, but the constraint will move and tomorrow it will be somewhere else. Who wins and who losses when it does?

• Investment “circularity” raises ghosts of bubbles past. AI will likely be the defining technology of our time. Predicting what that means for energy demand (and compute and hardware) is still difficult as model and hardware efficiencies work against expanding capabilities and the explosion in user demand. Will all the commitments look like foresight or folly?

• Gas power is more Interesting than “Gas-to-Power”

  Continued development of technologies like solar (utility scale and behind-the-meter) and BESS mean fewer dispatch hours on average for gas fired assets, even with load growth. Still, the value of that capacity, through capacity markets and premium priced hours, means rising values for many gas fired generation assets. Gas producers have found it difficult to take advantage – is that set to change?

• OBBBA renewable credit cut pain

  Many queued renewable projects will not meet economic thresholds without tax credits, but some will. Without credits and in to meet required capacity broadly, local dislocations are sure to emerge: Where will prices - PPAs, RECs, forwards etc. – need to adjust?

• “CO2-to-Oil” is the sleeper bet: real economics accessible to the oilfield

  As “Inventory” continues as a dominant E&P theme, any pocket of highly economic (low-breakeven) resource should draw attention. Supported by the 45Q tax credit, CO2 EOR can deliver some of the lowest breakeven oil opportunities in the L48. Access to CO2 (ironically?) is one of the biggest constraints to unlocking this opportunity. Who has CO2 access and acreage to take advantage?

• Geothermal warming up

  Geothermal, specifically EGS and AGS are something to watch. Recent land sales point to large swaths of acreage being locked up in anticipation of a breakthrough. Meanwhile, the pace of cost and performance improvements over time continues to impress. Will geothermal progress allow it to outcompete nuclear as the “low cost, low-carbon baseload” resource?

• Credit stacking still builds “fat stacks”
  The value of “energy” increasingly mirrors the energy trilemma: The value of the energy itself, the value of it’s reliability/dispatchability, and the value of it’s environmental attributes. Some of the best economics we see are from assets that can stack revenue streams from several attributes (where the value of the energy itself is a minor contributor). Where is this working following the OBBBA?

Three ideas caught my attention. The first was this aggregation of transitions within the energy system: fossil fuels to electricity, centralized to distributed power, dispatchable to variable, public supply to autonomous prosumers, monopolies to networks of millions, central planning to market interactions, and top-down to bottom-up control.

While I may debate some of the projected consequences, the themes themselves read right to me. Together, these transitions are driving a total transformation of the power system as we currently know it.

The highlighted transitions are, in part, a consequence (and driver) of innovation within a subset of technologies supporting renewables. One result is LCOEs for key renewable technologies are the cheapest available:

Combined with policy support, that has translated into global annual investments into renewables that are roughly an order of magnitude greater than nuclear:

Solar and wind capacity (and generation) are on track to surpass nuclear soon:

Of course, LCOEs capture cost, not reliability. On this the report notes progress on storage and power electronic technologies. The author’s assessment points to a rapidly improving and more competitive technology stack versus nuclear going forward, even after incorporating reliability (the table below highlights some of the key trends, specifically comparatively positive trends for renewables + storage versus nuclear).

The net result is a view that the energy system is being transformed in a way that exposes the weakness of nuclear and makes it less competitive. Specifically, the stable baseload generation profile nuclear offers lacks precisely the short-term flexibility that an increasingly renewable heavy grid demands. Further, planned (and unplanned) outages at these facilities mean their capacity must be duplicated to meet the demands of discerning loads like AI data centers.

To quote the authors:

Proposing nuclear power as a component in a firm power energy mix with strongly fluctuating renewable power, however, neglects its weak points. For technical and economic reasons, nuclear plants do not provide the type of flexible, dispatchable power that can fill the gaps between solar power peaks. They need flexibility from other sources for bridging considerable planned and unplanned outages and for buffering between changing demand and their inflexible full-load operation. The cost of their baseload electricity is more expensive than the ultra-flexible combination of renewables-plus-storage-plus-flexible demand.

This conclusion, at a time when nuclear is seeing a resurgence, was surprising to say the least.