The Power-Compute Nexus:Compute Is the New Tollbooth
AI’s real bottleneck was never the chips. It was electricity, land, and a grid that cannot connect fast enough — the demand shock underneath the entire infrastructure decade.
Fenrir Research · Jul 2026 · Yggdrasil Ledger / latticelog.in
They built the forge before they had the fire to feed it — a great engine of iron standing cold, waiting on a river of heat that no one had yet learned how to carry. The furnace was never the hard part. The hard part was the fuel, and the road it had to travel.
Original epigraph, in the register of Tolkien’s forge- and fire-verses
Section 01
The Demand Shock
For two decades, electricity demand in the United States was essentially flat. Efficiency gains cancelled out growth, utilities planned around a stable load, and power was the last thing anyone in technology thought about. Artificial intelligence ended that in about eighteen months.
The clearest way to see the scale is through what the hyperscalers are spending. Combined capital expenditure at the four largest — Amazon, Alphabet, Meta and Microsoft — ran near $226 billion in 2024, roughly $410 billion in 2025, and is guided at about $725 billion for 2026, a jump of roughly 77% in a single year. Fold in Oracle and the five-firm total pushes toward $700–900 billion, and analysts already pencil in more than a trillion dollars for 2027. Goldman Sachs puts cumulative hyperscaler capex across compute, data centres and power at something on the order of $7.6 trillion between 2026 and 2031. The Stargate venture alone — OpenAI, SoftBank and Oracle — carries a $500 billion headline. This is, by a wide margin, the largest private construction programme in history.
Hyperscaler Capex Has Tripled in Two Years
Combined capex, four largest US hyperscalers (Amazon, Alphabet, Meta, Microsoft), US$bn. Sources: company guidance; FT / Tom’s Hardware tally; CreditSights. 2026 figure is guidance, up ~77% YoY.
The number that matters for this section, though, is not the dollars — it is where they go. A rising share of every capex dollar now lands not on chips but on the physical shell around them: the buildings, the cooling, the substations, and the power contracts. Memory alone is set to absorb roughly 30% of hyperscaler data-centre spending this year, four times its 2023 share. The AI story, told honestly, is an infrastructure story wearing a software costume.
Section 02
Why Power, Not Chips, Is the Bottleneck
The tell came from Microsoft, which disclosed an Azure order backlog it could not fill — not because it lacked chips, but because the GPUs it already owned sat idle in inventory, waiting for power. When the most valuable company in the world is capacity-constrained by electricity rather than silicon, the binding constraint has moved.
The physics explains why. A traditional server rack draws 5 to 15 kilowatts; an AI training rack draws 30 to more than 100. A single cluster of 100,000 GPUs pulls 70 to 80 megawatts — the continuous draw of a small city — from one connection point. Multiply that across the build-out and the load curve goes vertical:
US Data-Centre Power
31 → 66 GW
2025 to 2027, more than doubling (Goldman Sachs)
Share of US Peak Demand
4.1 → 8.5%
Summer peak, 2025 to 2027 — a national market tightening
Global DC Electricity
~945 TWh
IEA base case by 2030, roughly double 2024 (~415 TWh)
Data-Centre Build Time
~12 mo
vs. 4–7 yrs to connect, 10+ yrs for new transmission
That last comparison is the whole problem in one line. The compute can be built in a year; the power to run it cannot. And the queue to connect has become surreal: at the end of 2024, roughly 2,300 gigawatts of generation and storage sat waiting in US interconnection queues — more than the entire installed US generating fleet of about 1,280 GW. In Texas, the ERCOT large-load queue hit 410 GW by spring 2026, 87% of it data centres, of which under 2% had actually energised.
The Queue Now Exceeds the Grid Itself
US installed generating capacity vs. capacity waiting in interconnection queues (end-2024), and the ERCOT large-load queue (spring 2026). Sources: LBNL “Queued Up: 2025 Edition”; ERCOT. Queued capacity is not all real — much is speculative — but the ratio signals the scale of the access problem.
Analyst Read — The Bottleneck Is the Investment
Most of that queued capacity will never get built; a large share is speculative, duplicative, or filed to hold a place in line. But the queue’s sheer size is the signal: the scarce, valuable thing is no longer the ability to build a data centre — it is the ability to power one. Whoever controls interconnected capacity, firm generation, or a site with secured power holds the real asset. Everyone else holds a waiting ticket. That is the single most important reframe in this entire section.
Section 03
The Workarounds — and Where They Lead
Because the grid cannot connect fast enough, capital is routing around it. Three responses dominate, and each is a full topic in its own right:
Skip the grid entirely. Site the data centre directly at a power source — behind the meter, on-site generation, or straight off a plant — and avoid the interconnection queue altogether. Developers going this route (xAI, Crusoe, Oracle) are energising in one to two years against four-plus on the grid.
Revive firm power. Renewables can’t run a 24/7 training cluster alone, so the build-out is dragging gas turbines (now backordered for years) and nuclear — including the hyperscaler-nuclear deals — back to the centre of the conversation.
Squeeze the existing grid. Reconductoring, grid-enhancing technologies and undergrounding aim to move more power through lines that already exist, faster than new transmission could ever be permitted.
Run this through the primer’s models and the investable shape comes into focus quickly.
How does it get paid? The prize is contracted, investment-grade offtake — a long power-purchase agreement or a take-or-pay lease with a hyperscaler whose credit is impeccable. That is the low-risk, financeable end, and it is why infrastructure and private-credit capital is piling in. The danger zone is merchant exposure: speculative “neocloud” capacity built on the assumption that demand will show up to fill it. Same sector, opposite risk.
What stage is it at? Almost all of this is greenfield — new data centres, new generation, new lines. On the primer’s risk spectrum that lands in value-add and opportunistic territory, not core. Whatever else the AI build-out is, it is not a bond proxy, and pricing it like one is the first mistake.
Where does policy become the cash flow? Everywhere. Power-market rules set what generation earns; interconnection reform decides who connects and when; and the firm-power response leans heavily on tax credits for nuclear, storage and gas. Change any of those and the economics move.
The Fenrir Question
Is the interconnection queue a temporary bottleneck, or a structural moat?
For a developer without power, it’s a bottleneck — a problem to be waited out or engineered around. For an incumbent that already holds interconnected capacity, firm generation, or a permitted site, it is a moat — a multi-year barrier that competitors cannot cross at any price. The durable value in this whole build-out clusters around the second group. Own the scarcity, not the queue.
Section 05
The Investment Map — and the Tail Risk
If the scarce input is power and the connective capacity to deliver it, the beneficiaries are the picks-and-shovels of the electrical system, not the model-builders:
Power Generators / IPPs
Firm capacity re-rated
Independent producers with dispatchable, 24/7 generation near demand centres are suddenly holding the scarcest asset in the system.
Grid Equipment
Multi-year backlogs
Transformers, switchgear and high-voltage kit are supply-constrained worldwide — a hard bottleneck with pricing power.
Gas Turbines
Backordered to 2029+
The fastest firm power available at scale; order books are full, which is bullish for makers and a constraint for everyone else.
Data-Centre Developers / REITs
Power = the moat
Value accrues to those holding secured power and interconnection, not to raw shell capacity.
Nuclear / SMR
Optionality, not yet delivery
The hyperscaler-nuclear deals price in firm, carbon-free baseload — real demand, uncertain timelines.
Infra & Private Credit
Funding the build
The financing stack for the 2025–28 cycle looks like project finance — an ~$800bn private-credit opportunity.
But this is where discipline earns its keep, because the consensus is now very crowded — and 2026 brought the first real cracks in it.
The Bull Case
Two demand shocks stack on the same base: AI compute and broad electrification both need power
Contracted, investment-grade hyperscaler offtake underwrites the strongest projects
The build-time mismatch (12 months of compute vs. years of power) is a durable, ownable scarcity
Inference demand is still early; if agentic AI scales, today’s capex looks small
The Tail Risk
The capex-to-revenue gap is widening; markets are starting to flinch (Meta fell ~9% on raising guidance)
DeepSeek showed training costs can collapse overnight — efficiency is the bear’s friend
Concentration: the whole edifice rests on a handful of hyperscaler counterparties
If demand lags the build, merchant and speculative capacity strands first
Ratepayers are already absorbing the cost — PJM capacity prices jumped ~$9bn, a political backlash risk
Analyst Read — Underwrite the Offtake, Not the Narrative
The single discipline that separates durable exposure from stranded exposure is the contract. Assets with committed, investment-grade offtake — regulated power, contracted hyperscale capacity, take-or-pay firm generation — survive even a sharp demand disappointment. Merchant compute and speculative capacity built on a forecast do not. The AI build-out is real; the question is never whether to have exposure, but which cash flows are contracted. That is the line between owning a tollbooth and owning a bet.
Bottom Line
AI turned a decade of flat electricity demand vertical, and in doing so revealed that the true scarce inputs in computing are physical: power, land, water and the connective capacity of the grid. The chips were never the constraint. The tollbooth is the connection to the electrical system, and it is the thing worth owning.
Read every deal in this build-out through the same lens: where is the power, is the offtake contracted, and is the scarcity a moat or a queue? The bull case — two demand shocks stacking on one physical base — is strong. The tail risk — a debt-funded, concentrated bet on demand that may arrive late — is real. Both are true at once, and the contract is what tells them apart.
The smiths who prospered were not the ones with the finest hammers. They were the ones who had secured the fire — who owned the heat itself, while their rivals stood in line at a forge they did not control.
Original epigraph, in the register of Tolkien’s forge-verses
Bifrost Systems: An Infrastructure Primer — Fenrir Research
Fenrir Research · Bifrost Systems · The Infrastructure Primer
Bifrost Systems:An Infrastructure Primer
Before the deep-dives — the mental models. What infrastructure is, how these assets actually earn and get valued, and why the whole game has changed.
Fenrir Research · Jul 2026 · Yggdrasil Ledger / latticelog.in
A bridge belongs to neither shore. It is built by people who may never cross it, to carry weight they will never see, between a world that exists and a world that does not yet — and everything of consequence, sooner or later, has to pass across.
Original epigraph, in the register of Tolkien’s bridge- and road-verses
Section 01
Infrastructure Moved From the Background to the Foreground
For most of a generation, infrastructure was something you noticed only when it failed — a blackout, a burst main, a closed bridge. It was the dependable substrate underneath the interesting parts of the economy, and being boring was the whole point. That era is over.
Three forces arrived at almost the same moment, and each one turns on the same physical base. The energy transition is not a policy aspiration but a multi-trillion-dollar construction programme — generation, grids, storage, all of it. The AI build-out made electricity and land the scarcest inputs in technology, turning power — not chips — into the true bottleneck. And a wave of geopolitical rewiring — war, protectionism, resource nationalism, reconstruction — put physical control of energy, minerals and trade routes back at the centre of statecraft. Underneath all three, the ageing infrastructure of the developed world needs refinancing at the same time.
The result is that infrastructure has become the binding constraint on almost everything the world is trying to do at once. This primer is the foundation for Bifrost Systems — Fenrir Research’s standing coverage of that shift. It won’t list every topic (a short roadmap sits at the end); it will give you the mental models you need to read any of them. In the old stories, Bifrost was the bridge between worlds, the structure that made everything else reachable. That is what infrastructure is to a modern economy.
The Scale of the Convergence — Annual Capital Demand
Sources: IEA World Energy Investment 2025 (clean energy, generation, grids, fossil); Dell’Oro / Morgan Stanley (data-centre construction, annualised). Figures indicative, US$/year. Note how little flows to grids relative to generation — the bottleneck in one bar.
Section 02
What Infrastructure Actually Is
Infrastructure is defined less by what the assets are than by how their cash flows behave. A toll road, a power grid, a fibre network and a hospital look nothing alike, but strip away the concrete and a common economic signature remains — and it is that signature, not the physical form, that makes something investable as infrastructure:
Essential services. Water, power, roads, data — demand is inelastic and largely indifferent to the economic cycle.
High barriers, often a monopoly. These assets are capital-intensive, usually regulated or concession-based, and hard to replicate. Competition is limited by design.
Long-lived. Useful lives of 25 to 99-plus years, matched by concessions or contracts of similar length.
Predictable cash flows. Revenue is regulated, contracted or availability-based — it doesn’t depend on selling a product at a volatile price.
Inflation-linked. Tariffs and regulated returns are frequently indexed, passing inflation through to revenue.
The One Mental Model to Keep
Infrastructure charges a toll on a flow.
Electrons, water, vehicles, data, freight — an infrastructure asset sits on a flow that has to happen anyway and collects a toll on it, under a long, protected contract. That is the whole thing. Real estate rents space. Commodities own the flow itself and its price. Infrastructure owns the toll booth. Once you see an asset that way, its risk and its value both become far easier to read.
Dimension
Infrastructure
Real Estate
Commodities
Gets paid by
A toll on an essential flow, under contract or regulation
Rent on space; occupancy-driven
The spot price of the thing itself
Duration
Very long (decades)
Medium (lease cycles)
Short, price-cyclical
Inflation
Explicitly linked
Partial (rent reviews)
Volatile, unlinked
Correlation to GDP
Low (at the core end)
Moderate-to-high
High
Because the cash flows are long, contracted and inflation-linked, core infrastructure behaves like a long inflation-linked bond with a little equity upside — which is exactly why pensions and insurers, whose liabilities have the same shape, are its natural owners. But hold that thought loosely: as the next section shows, only part of the universe behaves that way, and the label “infrastructure” on its own tells you almost nothing about the risk you’re taking.
Section 03
How to Read Any Infrastructure Asset
Two questions decide almost everything about an infrastructure asset’s risk. First, how does it get paid? Second, what stage is it at? Answer those two and you can place any asset on the risk spectrum — and price it roughly — before you know anything else about it.
Question one: how does it get paid?
There are only three answers, and they sit in ascending order of risk:
Revenue model
How the money arrives
What you’re exposed to
Example
Regulated (RAB)
A regulator sets an allowed return on a defined asset base
Regulatory risk, not market risk
Electricity & water networks
Contracted / availability
Fixed payments for making the asset available
Counterparty credit risk
PFI hospital, contracted wind farm
Merchant / demand
Revenue rises and falls with how much the asset is used
Volume, price & GDP risk
Toll road, airport, merchant power plant
Question two: what stage is it at?
A brownfield asset is already built and operating, with a cash-flow history you can underwrite. A greenfield asset still has to be constructed — carrying permitting, construction and ramp-up risk before it earns a cent. Greenfield sits a full rung higher than brownfield on everything.
Combine the two questions and the entire universe sorts onto one spectrum — from bond-like at the left to equity-like at the right:
Core
Brownfield, regulated or contracted, mature. Yield-led, low leverage.
~6–9%
Core-plus
Some growth or contract-renewal risk, light expansion capex.
~9–12%
Value-add
Repositioning, platform build-out, more leverage.
~12–16%
Opportunistic
Greenfield, merchant, or emerging-market. Equity-like risk.
~16%+
← lower risk · incomehigher risk · capital growth →
Why This Matters More Than Ever
A decade ago, “infrastructure” mostly meant the left of that spectrum — core, regulated, brownfield yield. But the energy transition and the AI build-out are inherently greenfield and value-add: you are constructing new capacity, not buying a settled toll road. The centre of gravity has drifted right. So when someone describes a strategy simply as “infrastructure,” the useful question is never whether it’s infrastructure — it’s where on this spectrum it sits, because a regulated grid and a merchant hydrogen plant share a label and almost nothing else.
Section 04
How It Earns, and How It’s Valued
Infrastructure returns come mostly from yield — the cash the asset distributes each year — plus modest capital growth, the reverse of the buyout world where the gain is almost all at exit. That yield-led shape, and the sheer length of the cash flows, drives how these assets are valued, and it produces two quirks worth understanding.
The first is duration sensitivity. Infrastructure is valued by discounting decades of future cash flow back to today. When the cash flows stretch that far out, small changes in the discount rate move the valuation a great deal — which is why rising interest rates hit infrastructure valuations harder than their stable cash flows alone would suggest, and why the asset class is more rate-sensitive than its “defensive” reputation implies.
The second is the regulated-asset-base anchor. For a regulated network, value is roughly the regulated asset base multiplied by the allowed return — a valuation floor and ceiling handed to you by the regulator. No other real asset has anything like it; it is the closest thing in equities to a set of published rules for what the asset is worth.
The Thing That Makes Infrastructure Unique
In infrastructure, policy doesn’t just influence the return. Very often, policy is the return.
A production tax credit, a capacity payment, a regulated allowed return, a carbon price, an availability contract with a government — in this asset class, a policy decision frequently doesn’t nudge the cash flow, it creates it. That is why so much of Bifrost Systems comes back to policy: a change in a subsidy or a price signal can turn a project from uninvestable to compelling overnight, and back again. Reading infrastructure well means reading the policy that manufactures its cash flows — not as background, but as the asset.
One caveat to carry throughout. Because private infrastructure is valued by appraisal rather than by a live market price, its reported values move far less than listed equivalents. This smoothing flatters volatility and correlation statistics — the diversification benefit is real, but partly an artefact of how infrequently the assets are re-priced. Treat the “low volatility” of private infrastructure as partly genuine and partly a measurement effect.
Section 05
The Tension That Organises Everything: Build vs. Strain
Here is the single most important shift, and the reason this section is structured the way it is. For most of modern infrastructure history, the binding constraint was capital — the hard part was raising the money. That is no longer true. Institutions are under-allocated to infrastructure against rising targets, private managers sit on record dry powder, and private credit is queuing to lend. There is a wall of money.
What is scarce now is everything physical: grid-connection capacity, permits, transformers, skilled labour, water, copper and critical minerals — and power itself. The bottleneck has moved from the balance sheet to the physical world. The question is no longer “can we finance it?” but “can we actually build it, and what will it run into?”
The Frame for the Whole Section
Build is the enormous, accelerating construction of new capacity — and the ingenuity being deployed to do it faster: colocation, second-life assets, hybrid renewables, small modular reactors, grid-enhancing tech.
Strain is what the build runs into — resource limits, physical climate risk, permitting walls and geopolitical friction.
The returns, and the mispricings, live in the gap between the two.
Section 06
The Lens: What Fenrir Research Actually Asks
There is no shortage of engineering writing on these subjects, and no shortage of advocacy. Bifrost Systems asks a narrower, less-crowded question: what should a long-horizon investor make of this? Three sub-questions do most of the work, and they’ll recur in every piece:
Where is a physical risk mispriced? Markets are good at pricing financial risk and clumsy at pricing physical risk — heat derating a grid, a basin running short of water, a mineral supply chain with one choke point. The gap between the two is where the opportunity usually sits.
Is this a structural moat or a temporary bottleneck? A four-year interconnection queue is either a durable barrier protecting incumbents or a problem about to be solved. Which one it is decides whether you want to own the scarcity or bet on its relief.
Where does a policy signal become a cash flow? Following Section 04 — when a subsidy, price or mandate turns into a contracted return, and how durable that return is to the next election.
The register is markets, not morality. The aim is to see the system clearly enough to position around it.
Section 07
Where This Goes: The Roadmap
With the models in place, the rest of the section applies them. The coverage runs across seven threads — three about building, three about the pressures the build runs into, and one about the markets where those conclusions change shape. This is the map; each piece is a full application of the frameworks above.
A note on the links. Bifrost Systems is published one piece at a time. Titles shown as live links are published; every other title is forthcoming and its link will stay inactive until that piece appears. Titles marked soon are drafted or in progress. Where a piece has a counterpart that reaches the opposite conclusion outside the OECD, both carry a ↔ mirrored by note.
01 — BuildThe Machines Getting Built
What’s being constructed, and the ingenuity used to build it faster.
Spotlight: India data centres as the demand anchor that pulls an unbuilt grid, a domestic equipment industry and captive clean power into being — the inversion of compute-as-strain.
For the demand story behind the urgency, start with the Power-Compute Nexus. For the constraints that will decide who wins, read the Interconnection Queue and Resource Adequacy: Power. For the finance, start with Infrastructure in Modern Portfolios and the four-part asset-class series. For the single most contrarian argument against the whole consensus, read The Demand Counterweight — then read its mirror, The Demand Multiplier, which argues the reverse for most of the world’s population. Reading a mirrored pair together is the fastest way to see why the same framework produces opposite answers in different markets.
Primer — Bottom Line
Infrastructure spent decades as the least interesting thing in the economy precisely because it worked. It is interesting now because it has become the constraint — on the energy transition, on AI, on national security, on growth itself. Everything the world wants to do next has to cross the same overloaded bridge, and the bridge is not yet built to carry the weight.
Hold the four models and you can read any of it: infrastructure is a toll on a flow; its risk is set by how it gets paid and what stage it’s at; its value is long-duration and often policy-made; and the binding constraint has moved from capital to physical capacity. Everything in Bifrost Systems is an application of those four. The rest is detail — and the detail is where the work is.
They did not ask whether the far shore was worth reaching. They asked only whether the span would hold — and then they laid the first stone, and trusted the ones who came after to lay the next.
Original epigraph, in the register of Tolkien’s bridge-verses
How the American grid works — a primer on generation physics, market architecture, and the supply-chain choke points reshaping US power.
BOTTOM LINE UP FRONT
The US power sector is the most complex regulated industrial system in the world. Four jurisdictional tiers, seven organised wholesale markets, three asynchronous interconnections, fifty state commissions, roughly 3,300 utilities, and a 2,300 GW interconnection queue all sit on top of a physical machine that must balance supply and demand in milliseconds.
For two decades this complexity was hidden by flat demand. With the data centre boom, the rules of the machine now matter for equity returns. Part I lays the foundations: the timeline, the physics, the regulatory architecture, the fuel mix, the rate-base model, and the three supply-chain choke points — solar AD/CVD tariffs, the grain-oriented electrical steel monopoly, and the interconnection queue — that are now binding constraints on every project in development.
Part II addresses the demand shock, the new generation stack, and grid resilience. Part III translates the framework into portfolio positioning — the utility re-rating, the S5UTIL trajectory, and named positioning tracks.
PART I · CONTENTS
01
A timeline of US electric power, 1882–2026
02
The physics of generation — how electrons get made
03
The 2025 fuel mix and what produces a kilowatt-hour
04
The regulatory stack — FERC, NERC, ISOs, state PUCs
05
Generation, transmission, distribution & the three interconnections
06
The rate-base business model
07
Three choke points — solar AD/CVD, transformers, the queue
G
Glossary — abbreviations and technical terms
01 · A timeline of US electric power, 1882–2026
Every structural feature of US electricity in 2026 is a fossil of a specific historical decision. The shape of the regulatory pyramid, the existence of seven different wholesale markets, the legal status of nuclear plants, the divide between vertically integrated utilities and merchant generators — none of these are technically optimal. They are political settlements layered on top of older political settlements. We organise the chronology around five inflection points.
Formation (1882–1935). Thomas Edison’s Pearl Street Station in Manhattan, commissioned September 4, 1882, was the first commercial electric utility. The current War — Edison’s DC versus Westinghouse and Tesla’s AC — ended with AC’s victory by the 1890s, fundamentally because AC could be transformed to high voltages for long-distance transmission. State public utility commissions began in 1907 (Wisconsin and New York simultaneously), establishing the principle that electricity was a “natural monopoly” requiring price regulation. The industry consolidated rapidly into pyramidal holding companies whose financial speculation contributed to the 1929 crash.
The regulated era (1935–1978). Roosevelt’s Public Utility Holding Company Act (PUHCA, 1935) broke the holding companies. The Federal Power Act (also 1935) created the Federal Power Commission (renamed FERC in 1977) to regulate interstate wholesale sales and transmission. The 1965 Northeast blackout — 30 million people, nine hours — drove the formation of the North American Electric Reliability Council (NERC’s predecessor). Through this entire period, the dominant business model was the vertically integrated, state-regulated, cost-of-service utility. Nuclear plants were built on this model from 1965 to 1985; after Three Mile Island (1979), new construction effectively halted for thirty years.
Deregulation (1978–2008). The Public Utility Regulatory Policies Act (PURPA, 1978), passed in response to the 1973 oil shock, required utilities to purchase power from “qualifying facilities” — small renewable and cogeneration plants. This opened the door for independent power producers. The Energy Policy Act of 1992 expanded wholesale competition. FERC Order 888 (1996) mandated open, non-discriminatory access to transmission systems. FERC Order 2000 (1999) encouraged the formation of Regional Transmission Organisations. The Energy Policy Act of 2005, following the 2003 Northeast blackout (50 million people, four days), made NERC’s reliability standards mandatory and enforceable. By 2008, seven ISOs/RTOs were operational, serving roughly two-thirds of US load.
Decarbonisation and the shock (2008–present). The shale gas revolution after 2008 collapsed Henry Hub prices and made combined-cycle gas the marginal generator across the eastern interconnection, displacing coal. Wind and solar costs fell 70% over the 2010s. The Inflation Reduction Act of August 2022 was the largest climate spending package in US history. The One Big Beautiful Bill Act of July 4, 2025 rewrote much of it. Layered on top: the AI data centre demand shock that broke the flat-demand consensus and rewrote utility valuations.
FENRIR VIEW
Three patterns recur across the timeline. First, every major regulatory restructuring follows a blackout. Second, every consumer subsidy regime gets reversed within 10–15 years (PURPA → 1990s deregulation; IRA 2022 → OBBBA 2025). Third, the deepest structural reforms (PUHCA, Order 888, EPAct 2005) preserved the rate-of-return business model intact. The architecture has been remarkably durable; only the prices and the politics change.
02 · The physics of generation — how electrons get made
Six technologies produce the vast majority of US electricity. Each operates on different physical principles, has different cost structures, different ramp characteristics, and different regulatory treatment. Understanding the physics is not optional — it determines which technologies can scale, which can be dispatched on demand, and which compete on cost versus reliability.
A CCGT is two thermodynamic cycles bolted together. First, a Brayton cycle: natural gas is compressed, mixed with air, combusted at ~1,500°C, and expanded through a gas turbine that spins a generator. The hot exhaust — still around 600°C — would normally be wasted. Instead, it passes through a heat recovery steam generator (HRSG) that boils water into high-pressure steam. The steam drives a second turbine (a Rankine cycle), producing additional electricity. The combined system reaches 60–63% thermal efficiency in modern H-class units — nearly double a coal plant’s ~35%.
Operating characteristics: Capex ~$800–1,100/kW. Construction time 24–48 months historically, now extending to 60+ months due to turbine supply chain. Ramp rate ~50 MW/minute from spinning reserve. Heat rate around 6,400 Btu/kWh. At Henry Hub $3.50/MMBtu, a CCGT’s marginal fuel cost is about $22/MWh; at $7/MMBtu, $45/MWh. The gas price is the single biggest driver of US wholesale power.
Why this matters: CCGTs are the marginal generator in almost every US RTO, almost every hour. They set the wholesale clearing price. Their abundance over the 2010s collapsed coal generation; their relative scarcity now is one reason PJM capacity prices have exploded. The OEM oligopoly (GE Vernova, Siemens Energy, Mitsubishi Power) is the binding supply constraint on new build.
A US nuclear plant is fundamentally a giant kettle. Enriched uranium-235 (3–5% U-235) sustains a fission chain reaction in a pressurised reactor vessel. The heat boils water (or, in pressurised water reactors, heats a primary loop that boils water in a secondary loop). The steam drives a conventional steam turbine. The entire fission process produces no CO₂ at the point of generation.
The US fleet is 94 reactors totalling roughly 95 GW of capacity, mostly pressurised water reactors (PWRs) of the Westinghouse three- and four-loop designs or General Electric BWRs. Capacity factors are exceptional: the fleet averaged over 92% capacity factor in 2024 — well above any other generation technology. Marginal operating costs are about $30/MWh, dominated by fuel and labour; the rest of the cost is sunk capital.
Why this matters: Nuclear is dispatchable, zero-carbon, and continuous — exactly what hyperscalers want for 24/7 data centre load. Existing US reactors are the most valuable underpriced assets in the US power complex. Microsoft–Constellation revived Three Mile Island for this reason. NextEra–Google is restarting Duane Arnold. Meta has signed 20-year PPAs with Constellation and Vistra totalling 3.2 GW of nuclear capacity. We address Small Modular Reactors (SMRs) — a different physical architecture — in Part II.
Solar PV exploits the photoelectric effect Einstein described in 1905. A photon of sufficient energy strikes a semiconductor (typically silicon doped with phosphorus and boron to create a p-n junction), liberates an electron from the valence band into the conduction band, and the resulting potential drives current through an external circuit. No moving parts; no thermodynamic cycle. The DC output is inverted to AC for grid synchronisation.
Modern utility-scale silicon PV achieves about 22–24% module efficiency. Capex has fallen below $900/kW for utility-scale fixed-tilt systems. There is no fuel cost — marginal cost is essentially zero. Capacity factor is the binding constraint: a fixed-tilt installation in Arizona produces around 30% capacity factor; in the Pacific Northwest, closer to 18%. Output is also highly correlated across geographic regions, producing the so-called “duck curve” of negative-priced midday hours in CAISO.
Why this matters: Solar’s near-zero marginal cost makes it always-dispatched-first when the sun shines, but its output is uncorrelated with peak demand (evening) and zero overnight. This is what drove storage adoption. First Solar’s cadmium telluride thin-film modules are the only major US-manufactured PV technology and the dominant beneficiary of post-OBBBA domestic content rules.
A wind turbine converts the kinetic energy of moving air into rotational mechanical energy, then into electricity via a generator. Power available in wind scales with the cube of wind speed (P ∝ v³) — doubling wind speed produces eight times the power. This explains why turbine economics are dominated by site selection. The Betz limit (1919) caps maximum extraction at 59.3% of incoming wind energy; modern utility turbines reach 45–50%.
US onshore wind is dominated by the Great Plains corridor (Texas, Oklahoma, Kansas, Iowa) where capacity factors reach 40–45% — among the highest globally. Offshore wind, almost exclusively along the Northeast Atlantic coast, can achieve 45–55% capacity factors but at 2–3× the capex of onshore. As we discuss in Part II, US offshore wind is currently in regulatory crisis.
Why this matters: Wind is the largest non-hydro renewable source in the US (11% of 2025 generation). Its diurnal pattern is opposite to solar — windier at night in most regions — making wind-plus-solar a complementary mix. The post-OBBBA tax credit construction-start deadline of July 4, 2026 is forcing a developer scramble. Wind component domestic content credits (§45X) terminate after 2027, threatening US wind manufacturing.
Conventional hydro extracts gravitational potential energy from water falling through a turbine. Output is determined by head (vertical drop) and flow rate. Modern francis and kaplan turbines achieve 90%+ mechanical efficiency. The US has about 80 GW of conventional hydro, concentrated in the Pacific Northwest (Bonneville Power Administration) and the Southeast (Tennessee Valley Authority).
Pumped storage hydropower (PSH) is the original utility-scale battery: pump water uphill when power is cheap, release through a turbine when prices are high. The US has about 22 GW of PSH, almost all built in the 1960s–80s to load-follow nuclear plants. Round-trip efficiency is 75–80%. New PSH projects face decade-long permitting timelines, but interest is reviving — Rye Development is partnering with Kentucky utilities on a 266 MW / 2,128 MWh PSH project.
Why this matters: Hydro is one of two zero-carbon dispatchable resources (the other being nuclear). Its long-duration storage capability is essentially impossible to replicate with batteries at current technology. As we discussed in the ENSO Primer, US hydro output is highly correlated with ENSO state — El Niño typically reduces Pacific Northwest hydro, La Niña enhances it.
Hydrogen is not itself a primary energy source — it is an energy carrier. The dominant US production route is steam methane reforming (SMR): natural gas is reacted with steam at ~800°C over a nickel catalyst to produce hydrogen and CO₂ (CH₄ + H₂O → CO + 3H₂, followed by a water-gas shift). About 95% of US hydrogen comes from SMR — colloquially “grey hydrogen”. Combining SMR with carbon capture produces “blue hydrogen.”
Electrolysis — splitting water with electricity (2H₂O → 2H₂ + O₂) — produces “green hydrogen” if the electricity is renewable. Polymer electrolyte membrane (PEM) and alkaline electrolysers dominate. The fundamental challenge is energy: about 50 kWh of electricity to produce 1 kg of hydrogen, which contains ~33 kWh of energy. Round-trip efficiency through a fuel cell is around 35–40%.
Why this matters for power: Hydrogen is being positioned as the long-duration storage solution and the decarbonisation pathway for hard-to-abate industrial sectors. The IRA’s §45V production tax credit (up to $3/kg for green hydrogen) was the largest single subsidy in the bill; the OBBBA has compressed it. Practical deployment timeline for hydrogen in US power generation: 2030+. Note: “SMR” here means steam methane reforming. The other “SMR” — Small Modular Reactor — is a different technology covered in Part II.
A coal plant is a pure Rankine cycle: pulverised coal is combusted in a boiler at ~1,200°C, the heat boils water into superheated steam, the steam drives a turbine generator. Most US plants are subcritical (~35% efficient); newer supercritical and ultra-supercritical units reach 40–45%. Average capacity factor for the US coal fleet has fallen from over 70% in the 2000s to about 42% today as plants run mid-merit rather than baseload.
US coal generation peaked at 2,016 TWh in 2007 (49% of generation). By 2025 it was about 750 TWh (17%) — a 63% decline driven by cheap gas, environmental regulation (MATS rule 2012, Clean Power Plan / replacement), and renewables economics. The 2025 uptick (+12% YoY) was a temporary gas-price-driven reversal that the EIA expects to unwind in 2026.
Why this matters: Many coal plants scheduled for retirement have been deferred to meet rising demand. PJM reports 17 power plants postponed retirement since the 2024 auction, retaining 1,100 MW of capacity — mostly coal. The political question of forced coal-plant retention is now active. Equity-wise, the few remaining coal-heavy IPPs (NRG, Vistra’s legacy coal fleet) are hedges, not theses.
03 · The 2025 fuel mix and what produces a kilowatt-hour
In 2025, the United States generated a record 4,430 TWh of electricity, up 2.8% year-on-year, ending a decade and a half of essentially flat aggregate demand. The fuel mix has shifted dramatically over fifteen years.
The 2025 mix breaks down to: natural gas 41%, nuclear 18%, coal 17%, wind 11%, utility-scale solar 7%, hydropower 6%, and a residual of biomass, geothermal, petroleum, and other sources (less than 1% combined). Renewables collectively contributed 24% of generation. Power-sector CO₂ emissions rose 4.4% on the year — the second consecutive annual increase after a long decline — driven by the brief coal resurgence on higher gas prices.
The dispatch stack and how a kWh gets priced
In every organised wholesale market, generators offer their output in bid stacks. The system operator dispatches in merit order — cheapest marginal cost first — until forecast demand is met. The bid of the last unit needed (the marginal unit) sets the locational marginal price (LMP) that every dispatched generator receives. This is fundamental to how US power markets work.
Three implications follow immediately. First, because gas is almost always the marginal generator, US wholesale electricity prices are essentially a leveraged play on Henry Hub. Second, zero-marginal-cost generators (wind, solar, nuclear, hydro) earn economic rents — they capture the gas-set clearing price despite paying nothing for fuel. Third, the steepening of the right tail of the curve (when the dispatch stack runs out of cheap generators and must call on oil peakers or scarcity-priced units) is what produced the PJM capacity price explosion we’ll detail in Part II.
04 · The regulatory stack — FERC, NERC, ISOs, state PUCs
“You don’t get to commit the sin and then ask all of us to feel sorry for you when there are consequences.”
— LEWIS STRAUSS · OPPENHEIMER
No serious investor can underwrite a US utility name without understanding which level of the regulatory stack governs which decision. The American power system is a four-tier pyramid layered onto a tripartite physical machine. The two do not map cleanly. This is the source of most analytical errors.
FERC regulates interstate wholesale electricity sales, transmission service, and the tariffs filed by ISOs and RTOs. It does not set retail rates. Established as the Federal Power Commission in 1935 and renamed in 1977, it answers to Congress and approves market designs proposed by ISOs and RTOs. FERC has five Commissioners; the current chair under the Trump administration has prioritised data centre integration, gas turbine permitting, and dynamic line ratings.
NERC — the North American Electric Reliability Corporation — is the Electric Reliability Organisation. Designated by FERC in 2006 following the Energy Policy Act of 2005, it develops and enforces reliability standards for the bulk power system across the US, Canada, and a sliver of Mexico. Its six regional entities (WECC for the West, SERC for the Southeast, MRO for the Midwest/Plains, NPCC for the Northeast, RF for the mid-Atlantic, Texas RE for ERCOT) audit compliance. NERC penalties for reliability violations can reach millions of dollars per day per violation.
ISOs and RTOs are non-profit grid operators that emerged from FERC Order 888 (1996) and Order 2000 (1999). They run day-ahead and real-time energy markets, dispatch generation by economic merit, manage congestion, and (in four of seven) run forward capacity markets. They serve about two-thirds of US electric load.
State Public Utility Commissions are the most important and least-discussed actors. They set retail rates, approve or deny utility integrated resource plans (IRPs), issue siting permits, and approve every dollar of utility capital expenditure that enters the rate base. In non-RTO regions (most of the Southeast, Southwest, and Northwest), they regulate vertically integrated monopolies directly.
The seven organised markets — at a glance
RTO / ISO
Footprint
Peak load
Capacity mkt?
PJM
13 states + DC; data centre alley (Northern Virginia)
~155 GW
Yes (BRA)
MISO
15 states across Midwest and South
~127 GW
Yes (seasonal)
ERCOT
Texas (~90%); outside FERC jurisdiction
~86 GW
No (energy-only)
CAISO
California + slivers of Nevada
~52 GW
No (RA-based)
SPP
14 states across central plains
~57 GW
No (energy + AS)
NYISO
New York State
~32 GW
Yes (ICAP)
ISO-NE
Six New England states
~25 GW
Yes (FCM)
Day-ahead, real-time, and capacity — three markets in one
Each RTO runs two energy markets and (in four cases) a third capacity market. The day-ahead market clears at noon the day before delivery: generators submit hourly bids for the next 24 hours, the RTO clears the market against a forecast load curve, and binding financial schedules are issued. The real-time market clears every 5–15 minutes against actual load, with deviations from day-ahead positions settled at the real-time price. Capacity markets (PJM’s Base Residual Auction, NYISO ICAP, ISO-NE Forward Capacity Market, MISO Planning Resource Auction) procure firm capacity 1–3 years forward to ensure system adequacy.
ERCOT is the major exception: it operates as energy-only, relying on scarcity pricing (capped at $5,000/MWh through 2025, with proposals to raise) to incentivise new capacity. This design works during normal years but failed catastrophically during Winter Storm Uri in February 2021, contributing to a multi-day blackout and at least 246 deaths.
FENRIR VIEW
The single most underappreciated structural fact in US power: roughly a third of the country — the Southeast (TVA, Duke, Southern Company), the Pacific Northwest (Bonneville Power Administration), most of the Mountain West — has no organised wholesale market at all. These are bilateral, vertically integrated utility territories where investor returns are set by regulated rate-of-return formulas, not by capacity auctions. The data centre boom is now forcing this part of the country into bilateral PPAs and resource adequacy debates it has never had to run before. This is where the longest-duration utility re-rating sits.
05 · Generation, transmission, distribution & the three interconnections
The physical chain runs in three segments. Electricity must be generated, transmitted, and consumed within milliseconds — it cannot be meaningfully stored at grid scale (battery storage is changing this but is still under 1% of grid energy). Each segment is regulated differently.
The three interconnections
North America’s bulk power system is divided into three asynchronous AC interconnections, each operating as a single synchronous machine at 60 Hz: the Eastern Interconnection (east of the Rockies, plus a bit of Texas), the Western Interconnection (west of the Rockies, extending to British Columbia and Baja California), and the Texas Interconnection (most of ERCOT’s footprint). Power flows freely within each interconnection but cannot flow directly between them — they would tear themselves apart if synchronised. Limited transfer between interconnections happens through DC ties, which can move several hundred MW each.
This three-grid architecture is the binding constraint on the long-discussed “macro-grid” — a national HVDC overlay that could move excess wind from the Plains to coastal demand centres. The DOE’s National Transmission Planning Study (2024) modelled scenarios requiring 2–3× transmission expansion by 2035. Whether any of this gets built is now an active question; the IIJA provided some transmission funding, and FERC Order 1920 (May 2024) mandated long-term transmission planning, but the political and siting hurdles remain extreme.
AC, DC, and the new HVDC opportunity
US transmission is overwhelmingly alternating current (AC) at 345 kV, 500 kV, or 765 kV. AC has dominated since the 1890s because it can be transformed easily between voltages. But at distances above roughly 500 miles, high-voltage direct current (HVDC) becomes economically competitive: it suffers no reactive losses, no skin effect, and can be controlled point-to-point. Modern HVDC links use voltage-source converters and operate at ±525 to ±800 kV.
The US has only about 8 GW of operational HVDC — Pacific DC Intertie (3.1 GW, Oregon to LA, since 1970), TransWest Express (planned), SunZia (under construction), and several merchant DC ties. By comparison, China has built over 400 GW of HVDC since 2010. The case for US HVDC expansion is strongest for offshore-wind-to-coastal-load and for moving Plains wind to PJM. The political case is much harder.
06 · The rate-base business model
For investor-owned utilities in regulated jurisdictions, returns are not market-driven. They are determined by a deterministic formula approved by the state Public Utility Commission. This is the most important equation in US utility equity investing.
The rate base is the depreciated book value of utility assets — generators, transmission lines, distribution poles, substations, software, vegetation management equipment, wildfire mitigation infrastructure — that the state PUC has determined are “used and useful”. The allowed ROE typically ranges between 9.0% and 10.5% across US jurisdictions, set through periodic rate cases. Most jurisdictions use a hypothetical capital structure (typically 50–55% equity) for ROE calculation.
Three implications for equity
Earnings grow with capex. A utility that doubles its rate base doubles its earnings power, holding ROE constant. This is why utility EPS growth tracks rate-base growth, not GDP or industrial demand. A utility growing rate base at 7% per year typically grows EPS at 6–8% per year.
Capex must be approved. Imprudent or duplicative spending can be disallowed by the PUC, sitting on the balance sheet earning nothing. PG&E’s bankruptcy from wildfire liability illustrated this in the negative; the systematic approval of large data centre transmission build in Virginia (Dominion) and Georgia (Southern) illustrates it in the positive. Regulatory risk is the dominant risk in this business model.
Returns are bond-like — until they are not. A regulated utility growing rate base 3% per year is a bond proxy. One growing rate base 8–10% per year on the back of data centre build-out, electrification, and climate resilience capex is something else entirely. This is the heart of the 2024–26 utility re-rating, which we address in detail in Part III.
Many large utilities are hybrid: NextEra Energy (NEE) is the parent of regulated Florida Power & Light and merchant NextEra Energy Resources. Public Service Enterprise Group (PEG) operates regulated PSEG New Jersey alongside merchant nuclear at PSEG Power. Vistra is mostly merchant but owns regulated retail at TXU. Understanding the regulated/merchant split in any utility name is the single most important valuation step.
07 · Three choke points reshaping the build-out
“It’s a long way from theory to a working bomb.”
— LESLIE GROVES · OPPENHEIMER
Capital is not the binding constraint on US power expansion. Three more prosaic problems are: solar module tariffs, the transformer supply chain, and the interconnection queue. Each deserves its own dedicated framing.
Choke point 1 — Solar AD/CVD tariffs and the US–China dimension
The US has been imposing trade remedies on solar imports since 2012, when the original Commerce Department investigation found Chinese manufacturers were dumping crystalline silicon modules in the US market. Subsequent investigations targeted the workaround through Southeast Asia. On April 21, 2025, the Department of Commerce announced its final affirmative determinations in AD/CVD investigations of crystalline solar cells from Cambodia, Malaysia, Thailand, and Vietnam. The country-wide rates are extreme:
Country
Antidumping rate
Countervailing rate
Notes
Cambodia
125.4%
3,403.96%
Country-wide adverse facts
Vietnam
271.3%
542.6%
Country-wide adverse facts
Thailand
111.5%
263.7%
Country-wide adverse facts
Malaysia
81.2%
168.8%
Lower than peers; some company-specific rates exist
These rates stack on top of pre-existing Section 201 tariffs (~14%) and the new “reciprocal” tariffs that took effect April 2025 (10–49% depending on country). Anti-circumvention duties from earlier Auxin investigations apply where applicable. The petitioners — the American Alliance for Solar Manufacturing Trade Committee — include Korean-owned Qcells and Mission Solar, Swiss Meyer Burger, and US-based First Solar.
The effect on US solar economics is severe. Roughly 80% of US solar modules historically came from these four Southeast Asian countries. Domestic cell and wafer manufacturing capacity is ramping (Qcells in Georgia, First Solar in Ohio/Alabama/Louisiana) but cannot fully fill the gap until 2027–28. The OBBBA’s domestic content thresholds compound the pressure. Net result: utility-scale solar capex in the US has risen materially since 2024 and will remain elevated through the rest of the decade.
Choke point 2 — The transformer supply chain and grain-oriented electrical steel
A transformer’s core is made of grain-oriented electrical steel (GOES), a highly engineered silicon-iron alloy with anisotropic magnetic properties that minimise hysteresis losses. The United States has exactly one domestic GOES producer: Cleveland-Cliffs, operating plants in Butler, Pennsylvania and Zanesville, Ohio. Every domestic transformer manufacturer draws GOES from this single source.
Demand for power transformers has exploded. Wood Mackenzie data show generator step-up (GSU) transformer demand up 274% from 2019 to 2025; high-voltage power transformer demand up 116%; substation demand up 91%. Supply has not scaled proportionally:
As of Q2 2025, large power transformers averaged 128 weeks (2.5 years) for delivery; GSUs averaged 144 weeks (2.75 years). Some classes of distribution transformer have risen 95% in price since 2019. GOES prices have roughly doubled since 2020; copper prices are up over 50%. Roughly 80% of large US power transformers are imported, primarily from Mexico, South Korea, and China — exposing the supply chain to trade policy and global demand.
The structural response is in motion. ArcelorMittal announced a $1.2 billion non-grain-oriented electrical steel (NOES) plant in Alabama (first production 2027). GE Vernova completed its acquisition of Prolec GE in February 2026, consolidating North American transformer capacity. Hitachi Energy, Hyundai Electric, and HICO have all announced US capacity expansions. Wood Mackenzie forecasts the supply deficit for large power transformers narrowing from 30% in 2025 to roughly 5% by 2030 — but only if announced investments materialise on schedule.
Choke point 3 — The interconnection queue
A generator cannot connect to the grid without an interconnection agreement. The process — system impact study, facilities study, generator interconnection agreement — is operated by the relevant RTO or transmission owner. It has become the single largest bottleneck on US power expansion.
According to the Lawrence Berkeley National Laboratory’s “Queued Up: 2025 Edition” report, approximately 2,290 GW of generation and storage capacity was actively seeking interconnection at the end of 2024 — roughly 1.8× the entire installed US generation fleet. The composition: solar 956 GW, storage 890 GW, wind 271 GW, natural gas 136 GW (up 72% YoY). 408 GW already has a draft or executed interconnection agreement but has not yet reached commercial operations.
Two facts about the queue matter for equity investors. First, completion rates are low: historically, only about 14% of capacity that entered queues between 2000 and 2019 reached commercial operation by end-2024. The rest is withdrawn, fails studies, or is indefinitely delayed. Second, wait times have doubled: median duration from interconnection request to commercial operation date has risen from under two years (for projects built 2000–2007) to over four years (for projects built 2018–2024).
FERC Order 2023 (effective 2024) is attempting to fix this by moving from “first-come, first-served” to “first-ready, first-served” cluster studies with stricter financial commitments. Early indications are that the queue is rationalising — total active capacity decreased 12% YoY in 2024 as historic withdrawal waves cleared speculative projects. But the binding constraint remains: a US project applying for grid connection today cannot reasonably expect to be operational before 2030 in most jurisdictions.
FENRIR VIEW
The three choke points share a common analytical implication: incumbency is worth a premium. Generators with existing grid connections, transformer fleets, and operating PPAs have an option that new entrants cannot replicate inside this decade. This is why nuclear restarts (Three Mile Island, Duane Arnold, Palisades) command 20-year PPAs at premium prices — they offer immediate, dispatchable, zero-carbon capacity that no greenfield project can match. It is also why co-located behind-the-meter deals are commanding premium economics. Part III translates this into specific equity positioning.
Bottom line · what Part I established
“Now I am become death, the destroyer of worlds.”
— J. ROBERT OPPENHEIMER · BHAGAVAD GITA, AS QUOTED IN OPPENHEIMER
We have laid the foundations. The US electricity system is the product of 144 years of layered political settlements operating on top of immutable physics. Six generation technologies produce 99% of US electrons. Four regulatory tiers govern who gets paid what. Seven organised wholesale markets, plus a third of the country still running on bilateral utility tariffs, set the prices. The rate-base business model translates capex into earnings through a deterministic formula. Three supply-chain choke points — solar AD/CVD tariffs at up to 3,400%, transformer lead times of nearly three years, and a 2,300 GW interconnection queue with a 14% historical completion rate — bind the speed of any response to changing demand.
Part II addresses the demand shock: data centre and electrification forecasts, the new generation stack (gas turbines, nuclear restarts, SMRs, storage, CCUS, geothermal), the IRA-to-OBBBA policy reset, the offshore wind freeze, the shift from net-zero rhetoric to “energy abundance” policy framing, and grid resilience capex including covered conductors and undergrounding. We will reference our existing ENSO Primer and ENSO Markets piece for the climate overlay on wildfire and storm activity.
Part III translates everything into portfolio positioning: the S5UTIL index trajectory, the P/E re-rating, the shift from defensive bond-proxy framing to growth narrative, five named positioning tracks with anchor stocks, and the risks that could compress the trade.