Bayes Group
← All Insights
·16 min read·Updated : auction price corrected; section added on events since May 2026

After PJM Cleared at the Cap: How Multi-Strats Built the New Power Desk in 2025

By Richard H., Founder

When PJM's December 2025 capacity auction cleared at the FERC-approved price cap with data centers responsible for forty percent of the cost, the multi-strats had already spent the year acquiring physical power businesses, opening Houston and Dubai pods, and hiring ex-utility and ex-trading-house talent into structures that look closer to a senior systematic-PM hire than to the sell-side energy trader of five years ago. The power-desk hire of 2026 is a different job from the one most BD and HR teams are still resourced to fill — and the firms that get the profile right will define the next phase of multi-strat hiring.

A Capacity Auction That Read Differently

Multi-strat hedge funds spent 2025 quietly buying physical power and gas businesses. The auction print that made the trade unmistakable arrived in mid-December 2025: PJM Interconnection's 2027/2028 Base Residual Auction cleared at $333.44 per megawatt-day across the full footprint, at the FERC-approved cap, and more than eleven times the $28.92 the same market cleared for delivery year 2024/2025. The PJM Independent Market Monitor subsequently estimated that data-center load was responsible for roughly forty percent of the $16.4 billion in capacity costs the auction will recover from ratepayers across the footprint.

The new power desk is a physical power and gas trading pod built inside a multi-strat platform, assembled through acquisitions and senior hires rather than inherited from the sell-side. The auction was not surprising to anyone who had been watching power markets through 2025. It was a confirming print on a regime change the multi-strat platforms had already been positioning around.

By the time PJM cleared, Citadel had spent the year building a physical power and gas business in earnest: acquiring Paloma Natural Gas for roughly $1 billion in March 2025, agreeing to acquire German power trader FlexPower in October 2025, and recruiting Mitchell Walk, the former head of North American power, gas and emissions research at Energy Aspects, into the seat. Verition opened a Houston office in April 2025, hired ex-Millennium trader Alex Mouturat into Dubai in September 2025 to build a European gas, power, carbon and oil pod, and through November 2025 added a London team led by Beaufort Energy founder David Brookes alongside Houston refined-products hires James Hutchinson from Millennium and Gregory Galimberti from TotalEnergies. Balyasny had built out a Denmark physical power and gas unit through 2024 and 2025 with hires including Joe Constantinou from Bank of America and Federico Censi as a gas trader. Squarepoint, primarily a quant firm, had pushed into physical metals trading through August and September 2025, and in March 2026 launched STG Securities, an electronic market-making affiliate. Bobby Jain's commodities build under Jain Global, before its April 27, 2026 pivot to manage capital exclusively for Millennium, included ex-BlueCrest commodities specialist Bruce Kish.

These are not isolated hires. Read in sequence, they describe a single phenomenon: the multi-strat platforms spent 2025 buying the ability to trade physical power and gas, and they did it with a hiring template that does not look like the sell-side energy desk of 2018.

What did the April 2026 hyperscaler capex print confirm?

The April 2026 hyperscaler earnings print made the demand side of the trade legible. Microsoft, Alphabet, Amazon, Meta and Oracle collectively committed somewhere between $660 billion and $725 billion of 2026 capex, with roughly three quarters of that earmarked for AI-related infrastructure. Microsoft guided to about $190 billion, and Alphabet's capex guidance ran to as much as $190 billion, the top of a 2026 range it raised that week to $180 billion to $190 billion. The binding constraint named on call after call was not GPU supply or model architecture. It was electricity at the substation.

The EIA's January 2026 release forecast the strongest four-year growth in U.S. electricity demand since 2000 and revised commercial-sector demand growth to roughly 5% for 2026, driven principally by data centers. ERCOT forecast its peak load to grow 9.6% in 2026 and roughly 50% by 2029, with AI and high-density compute the dominant driver. The IEA's 2025 Energy and AI report made the same point at global scale. Gas did not follow power: the EIA's October 2026 outlook has Henry Hub averaging $3.48 per MMBtu in 2026 and $3.16 in 2027, with storage above the five-year average in most months, which is the cheap-fuel, dear-product configuration a spark-spread book wants. ERCOT North forward summer on-peak contracts traded well above $50/MWh through Q1 2026, and forward winter peak contracts continued to grade upward through 2029.

The arXiv literature has caught up to the same observation. The Carnegie Mellon team's 2025 paper on the electricity demand and grid impacts of AI data centers flags exactly the dynamic the multi-strats are trading: large, often unpredictable AI loads that amplify locational marginal price volatility in transmission-congested zones, and capacity-market clearing prices that have moved by an order of magnitude in three years. ACER's April 2026 review of European gas wholesale markets through winter 2025-26 reaches the parallel conclusion for TTF: extrinsic value has overtaken intrinsic in the storage curve, and the pricing of optionality has become the dominant feature.

Read together, the December 2025 PJM clear, the January 2026 EIA revision, and the April 2026 hyperscaler capex print are three disclosures of the same regime. The multi-strat power-desk build-out of 2025 was a positioning bet on that regime, taken before the print confirmed it.

What does the new power-desk hire look like?

The reference profile being hired into these seats is a composite that did not exist as a clean single hire five years ago.

The first leg is a physical trader with sourcing, scheduling and logistics fluency, usually built at a major (BP, Shell, TotalEnergies), an integrated utility (EDF Trading, Engie, RWE) or a merchant (Vitol, Trafigura, Mercuria, Gunvor). The Verition Houston refined-products hires are typical. Mouturat's prior tenure trading European power at EDF Trading and Engie before Millennium is the lane the Dubai pod was built around. Citadel's FlexPower acquisition imports a desk that already trades more than 1,700 megawatts across six European countries and places over 11 terawatt-hours annually on short-term power markets: physical infrastructure as much as personnel.

The second leg is a fundamental research function with depth in regulatory mechanics, gas-power coupling, capacity markets and weather. Mitchell Walk's move into Citadel from Energy Aspects is the canonical example: research seniority that can sit adjacent to the trading book and feed it with structured macro and microstructure intelligence. The hires being made into the European power and gas pods at Balyasny and Verition follow the same template.

The third leg, and the one that genuinely separates the 2025-26 build from the previous wave of multi-strat energy desks, is the systematic overlay. The 2025 arXiv work on maximizing battery-storage profits via high-frequency intraday trading, a German order-book backtest showing that an order-book-aware policy earns 58% more than re-optimizing hourly and 14% more than re-optimizing every minute, describes a class of strategy that requires a quant infrastructure the trading houses generally do not have. The continuous intraday market in Europe trades to within a few minutes of physical delivery, and the value at stake versus a perfect-foresight benchmark is now well-quantified. The deep-learning literature on day-ahead and intraday electricity price forecasting has moved from academic curiosity to production tooling at the firms that can absorb it. The multi-strats can; the trading houses, with rare exceptions, still cannot.

The senior power-desk hire of 2026 is a profile that reads against all three legs at once. It is closer to the senior systematic-PM hire we wrote about in May 2026 in the context of elite market-makers than to the senior energy trader of 2018. The reference compensation is also drifting toward PM-pod economics: base bands roughly in the range of senior PM hires, with pod payouts in the 20-30% of P&L range on the multi-strat side, against the more lump-sum book-share economics that have historically prevailed inside the trading houses.

Why is the talent pool shallower than the demand implies?

The talent pool that fills this composite is structurally narrow.

Senior physical power and gas traders at the merchant houses have historically traded inside compensation structures that the multi-strats can outbid in cash terms but do not necessarily out-build in optionality. Major multi-strat hiring slowed in 2024 but, on Hedgeweek's April 2025 count, the five largest platforms still added more than 550 portfolio managers and analysts, and Bloomberg's read on the cross-firm reneging dynamic between Citadel and Millennium confirms that the market for senior commodities talent now operates with the same offer-volatility dynamics as senior PM hiring more broadly. Pablo Duran Steinman reneging on Millennium twice for Citadel is the macro-trader version of the dynamic; the energy-pod equivalents are not yet as visible publicly, but the offer rounds inside the merchant houses through Q4 2025 and Q1 2026 say the same thing.

The peer-reviewed work on commodity factor risk premia adds a structural reason. The Nakagawa-Sakemoto paper estimating prices of risk for commodity factors argues that the equity market factor has become more strongly associated with cross-sectional commodity portfolios in the post-2010 financialization regime, a reason that allocators looking for diversifying alpha now need pods that can credibly produce idiosyncratic, supply-and-demand-driven energy P&L rather than re-priced beta. The Federal Reserve's 2025 working paper on whether financial stress affects commodity futures traders' positions reaches a related conclusion from the position-data side. The combination is: idiosyncratic energy alpha is rarer and more valuable than the headline coverage suggests, which raises the bar on the talent that can produce it.

The talent comes from a small number of seats. The major and integrated-utility sources are well-mapped. The merchants (Vitol, Trafigura, Mercuria, Gunvor, Glencore, Hartree) guard their senior physical traders carefully and have themselves been hiring across base metals and energy through 2025 as the metals and physical-oil books have stayed profitable. The fundamental-research lane (Energy Aspects, S&P Commodity Insights, Wood Mackenzie, BloombergNEF, Rystad) has a shallower senior bench than the trading-side lane and has already been picked at. The systematic-overlay lane is the rarest of the three because it requires a researcher who can speak both the order-book language of the European intraday market and the fundamental language of the gas-power coupling, and that profile is still being trained more often than it is being hired.

For a search firm sitting in front of a power-desk mandate, the practical consequence is that the candidate pool reads narrow on first inventory and widens only with cross-asset triangulation: systematic macro researchers with rates and FX backgrounds who have built coupling models for energy as a side project, equity-derivatives quants with mid-frequency factor experience who can extend to power forwards, ex-utility scheduling leads with quantitative aptitude who have not yet been credentialled into the trading-desk seat. Each of those lanes contains real candidates; none of them is the obvious first call.

What This Changes for Multi-Strat BD and HR Teams Through 2026

Three implications matter for a Head of BD or Head of Human Capital sitting on a multi-strat platform that is either building or being asked about an energy pod.

First, the diligence question allocators are starting to ask is not "do you trade commodities" but "what is the source of the energy P&L". A cross-asset book that prints commodity returns by way of macro overlay is a different product than a pod that has physical sourcing, scheduling and logistics, and the second is what the AI-capex regime change has made more valuable. Allocators who have absorbed the Hedgeweek and With Intelligence outlook framing for hedge fund growth into 2026 are increasingly able to distinguish between the two.

Second, the comp envelope for the senior power-desk hire of 2026 sits closer to a senior systematic PM than to a sell-side energy trader. The implication for hiring infrastructure is that the platforms competing here need PM-style search processes: long lists, real triangulation across the major / utility / merchant / quant lanes, sit-out-window planning per the eighteen-to-twenty-four-month garden-leave cycles we wrote about for senior PMs, and the ability to hold a candidate through an 18-to-24-month sit-out without losing them to a competing offer.

Third, the firms doing this well are vertically integrating. Citadel's FlexPower acquisition is not a hire; it is the acquisition of a full physical book, with the assumption that the senior leadership stays and a systematic overlay gets added to it. Squarepoint's STG Securities affiliate launch follows a comparable logic on the market-making side. Verition's Houston-and-Dubai twin-pod build is a smaller-scale version of the same template: the firm bought the geography and the physical fluency as well as the names. For a BD team, this matters because the conversation with allocators about commodity exposure is now framed against a peer that has already done the integration.

What has happened since May 2026?

PJM cleared at the cap a third time. The 2028/2029 Base Residual Auction, reported on July 14, 2026, cleared at $325.00 per megawatt-day across the whole footprint, 6,831 MW short of the reliability requirement, with 525 MW of new generation and uprates clearing against a load forecast that rose 1,374 MW largely on new large loads; PJM's own simulation without the collar puts the uncapped clearing price at $554.72, and $776.69 in ComEd. The cap held because FERC extended the price collar to the 2028/2029 and 2029/2030 auctions on April 28, 2026, after all thirteen PJM-state governors and two cabinet secretaries signed a statement of principles on January 16, 2026 asking for exactly that and for a backstop auction aimed at data-center load. The market monitor's July 2026 read is that data-center load accounted for $6.3 billion, or 38.2 percent, of the $16.4 billion in capacity charges from that auction, and $29.4 billion, 46.2 percent, across the last four; its position is that data-center load should leave the capacity market for a dedicated auction. The backstop itself slipped. PJM's board approved a fifteen-year reliability backstop procurement for the 6.8 GW gap on July 27, 2026, FERC accepted it on September 29 and suspended it for five months over cost allocation, transmission-owner exit terms and a $1.5 million per MW collateral requirement, and PJM refiled on October 29 for a March 1, 2027 launch with costs following each zone's firm large-load growth. For a power desk the trade is no longer the auction print. It is the regulatory path between a capped capacity price and an uncapped energy price.

The demand side kept printing, and the volatility arrived in winter rather than summer. During Winter Storm Fern, January 23 to 26, 2026, PJM's day-ahead LMP averaged $362 per megawatt-hour, gas exceeded $100 per MMBtu at north-east hubs on January 27 and uplift ran to $805 million over ten days. The EIA then counted summer peak-load records in PJM at 162.6 GW on July 2, 2026, in ERCOT at 91.1 GW on July 22 and in SPP at 58.0 GW on July 27, in the hottest summer on record nationally, and its October 2026 outlook has US wholesale power averaging $52 per megawatt-hour in 2026, with PJM up 41 percent on 2025. ERCOT's own summer review complicates the Texas leg: the 91,134 MW record arrived with real-time hub prices below 2024 and 2025 and peaker margins lower still, because solar and storage growth outpaced large-load additions and real-time co-optimization lowered ancillary costs. Texas then paused the queue. On August 3, 2026 Governor Abbott ordered an audit of every data center seeking an ERCOT connection and halted approvals until it is complete, against a large-load queue of more than 474 GW, about 90 percent of it data centers. The capex behind the queue has not slowed: Meta's third-quarter 2026 release narrowed its 2026 range to $135 to $145 billion. The Dallas Fed's working paper, revised in October 2026, gives the price effect a number: data centers raised wholesale prices 6.9 percent nationally by 2025 and 15.0 percent in northern Virginia, with the national effect reaching 12.3 percent by 2028 under its preferred scenario.

The rulebook moved too. FERC did not write the national large-load interconnection rule the Department of Energy had asked for; on June 18, 2026 it issued six show-cause orders, one to each RTO, giving them 60 days to justify their tariffs or file large-load and co-location provisions. PJM's July 2026 package added an interim resource-adequacy service under which new large loads that do not bring their own supply curtail or run on backup generation when the system nears emergency conditions. The preprints have followed the policy. Liu and Mays' August 2026 paper on capacity markets for large loads under supply-chain constraints finds that bring-your-own-capacity mandates mostly shift procurement between grid-built and self-built capacity under efficient pricing, but raise welfare under administrative price caps because they make data centers pay the full cost, which is PJM's configuration. Pan, Alexakis and Konstantinou's price-chasing data centers paper, revised October 21, 2026, shows flexible loads that chase cheap hours compressing the low-price hours that storage arbitrage relies on. For the systematic-overlay quant on a power desk, that is a change in the order book he is modeling.

The hiring did not pause for any of it, and the supply side opened in two places. Verition had hired about ten energy traders by the end of 2025, and in February 2026 a Citadel power portfolio manager moved to a newer multi-strategy platform's commodities team in Houston. Balyasny added a New York gas trader from DRW in September 2026, one of several gas moves Hedgeweek's report counted alongside a hire at ExodusPoint, with that report putting the pay arithmetic that pulls traders out of merchants and utilities at about 15 percent of profits at a merchant, 10 percent at a utility and more than 20 percent at a fund. Roscommon shut its US gas desk in January 2026 and by August at least eleven of its people had landed at Balyasny, Citi, Marathon Petroleum, Trafigura, Castleton and others; Austin's e360 Power closed its flagship US gas and power fund in December 2025 after tariff-driven losses and redemptions. Citadel's commodities arm, the model for the whole build, had a quieter 2025 and first half of 2026 by Hedgeweek's account, and its natural-gas trader of almost two decades stepped back to an advisory seat in September 2026. The buyers widened beyond the funds: Deutsche Bank, out of commodities since 2013, hired two former Goldman traders and two Morgan Stanley commodities salespeople in August 2026 and in October a sixteen-year BP veteran to lead Americas power, and OpenAI posted a power-trading lead role in August 2026 to run a hedging program across its data-center power portfolio, asking for ten years in power or gas. The recruiters' read is that commodity hiring that normally peaks in the first quarter ran through the second and into the third in 2026, with the Hormuz disruption from late February putting volatility back into every book.

Europe gave the gas and power pods their own winter set-up. EU storage, which the May 2026 version of this piece expected near 96 percent by 1 November, stood at 74.6 percent, 845 TWh, on October 31, 2026, after ACER warned in July that LNG imports would need to rise about 13 percent on 2025 to reach the 90 percent target from a mid-year level near 49 percent. The TTF front month topped €100 per megawatt-hour intraday on October 28, 2026 with storage at a record low for the date, and the funds had been stepping up gas-trader hiring through September 2026 ahead of another potentially volatile winter. The European intraday pods built in 2025 were built for exactly this tape.

What to watch into 2027

The next read on the regime is the 2029/2030 Base Residual Auction, which closes on December 15, 2026, the last auction under the extended collar. A fourth clear at the $325 cap with the reserve margin still below 20 percent would make the backstop procurement, now targeted for March 1, 2027, the real price discovery for new capacity in PJM. FERC's answer to PJM's October 29, 2026 refiling, and the RTOs' responses to the June 2026 show-cause orders, set how large loads connect, pay and curtail through 2027. In Texas the question is whether the ERCOT audit clears the 474 GW queue or shrinks it, and whether the solar-and-storage build that kept summer 2026 prices down keeps pace with the load that does connect. In Europe, a 74.6 percent storage start to winter against a TTF front month that has already touched €100 is the volatility surface the European power and carbon pods are trading, and the EU ETS reform under discussion through 2026, with the average CO₂ price projected toward €85/t for the year and triple digits by 2027, sets the carbon-and-power coupling underneath it. Every one of those is a regulatory or physical fact rather than a model output, and the three-legged hire is the only profile that reads all of them.

The directional read is straightforward. The multi-strat power-desk build of 2025 was an early bet on a regime that the December 2025 and July 2026 PJM prints and the April 2026 hyperscaler capex disclosure have confirmed. The hiring template the platforms used to assemble those desks (physical-trader leg, fundamental-research leg, systematic-overlay leg, all into a pod that pays on PM-style economics) is the template that will define the next eighteen months of senior commodities hiring across the multi-strat universe. The firms that can find and close against all three legs simultaneously will define what the new power desk looks like. The ones that try to hire any one leg in isolation will pay the same prices and end up with a different desk than they thought they were building.

Common Questions

Why did multi-strats build physical power desks in 2025?

They positioned ahead of a regime change that later printed publicly: PJM's December 2025 capacity auction cleared at $333.44 per megawatt-day at the FERC-approved cap, more than eleven times the delivery-year 2024/2025 clear, with data-center load responsible for roughly forty percent of the $16.4 billion cost, and the July 2026 auction cleared at the cap again. Hyperscalers then committed between $660 and $725 billion of 2026 capex, mostly AI-related.

What does the new multi-strat power-desk hire look like?

A three-legged composite: a physical trader with sourcing, scheduling, and logistics fluency from a major, utility, or merchant house; a fundamental research function deep in regulatory mechanics, gas-power coupling, and capacity markets; and a systematic overlay quant who can work the European intraday order book. Compensation drifts toward PM-pod economics, with pod payouts in the 20 to 30% of P&L range.

Where do firms find power-desk talent for these builds?

The pool reads narrow on first inventory and widens only with cross-asset triangulation: systematic macro researchers who built gas-power coupling models as a side project, equity-derivatives quants with mid-frequency factor experience extendable to power forwards, and ex-utility scheduling leads with quantitative aptitude. The obvious sources, the majors, integrated utilities, and merchant houses, guard their senior physical traders carefully.

Bayes Group

To respond to this piece, write to Richard: richard@bayes-group.com