Subscribe
Analysis

Long-form pieces

Theses, breakdowns, supply-chain deep-dives. New analysis lands roughly weekly during the build phase.

Utility-scale solar array in early morning light, panel rows angled toward the horizon (file photo).
thesissolar

T-minus 24 hours to July 4: the 5-percent cost-incurred pool is being finalized through post-vacatur equipment purchase agreements, and the documentation quality gradient is steeper than tax-equity is pricing

The physical-work pathway on the July 4, 2026 begin-construction deadline has been the visible headline through Q2 2026. The quieter half of the safe-harbor pool sits on 5-percent cost-incurred, and the last-day filings on that pathway are running through equipment purchase agreements drafted after the June 6 vacatur of Notice 2025-42. Post-vacatur EPAs carry deposit, delivery, and 3.5-month-payment structures that the vacated notice would have narrowed. On the wire count, roughly 55 to 70 GWdc of the announced 216 to 240 GWdc solar safe-harbor pool is coming through the cost-incurred pathway, and the last-24-hour documentation stack is where the audit-quality gradient will concentrate. Tax-equity is not fully pricing the intra-pathway bifurcation yet.
Onshore wind farm at dusk with utility-scale turbines against a clear sky (file photo).
thesiswind

The wind side of the July 4 OBBBA safe-harbor runs on a different clock than solar, and the repowering pathway is where the audit-lookback exposure concentrates

The July 4, 2026 begin-construction deadline is being read primarily as a solar-ITC event because the safe-harbored solar pool at 216 to 240 GWdc is the visible number. The wind side of the OBBBA safe-harbor is quieter and structurally different, and the wind pool has been drafted less carefully. The 80/20 repowering rule and the treatment of partial-generator replacements is where the wind audit-lookback exposure sits. On a run of the numbers, roughly 45 to 60 GW of onshore wind repowering has been rushed into the July 4 window with documentation packages that are lighter than the equivalent physical-work stacks on new-build wind, and the tax-equity pricing on wind repowering is starting to bifurcate.
Aerial view of the Prairie Island Nuclear Generating Plant in Red Wing, Minnesota (file photo).
weekly-digestai-demand

weekly digest, june 22-28, 2026

Two weight-bearing nuclear data points arrived in the same 24 hours. LevelTen Energy's Q2 2026 PPA index put hyperscaler-counterparty corporate procurement at 9.4 GW signed for the quarter, with nuclear at 3.9 GW (42 percent), the first quarter on record above 40 percent. The NRC's new mandatory-hearings policy statement (effective June 8, surfaced this week) moves uncontested licensing hearings from the end of the staff technical review to about 30 days after docketing, implementing Section 207 of the 2024 ADVANCE Act and pulling the most-cited timeline-compression lever in the advanced reactor pipeline. China's Ministry of Commerce added MP Materials and USA Rare Earth to its export-control list on June 22, a curb that bites no current orders but hardens the federal capital case for domestic heavy rare-earth separation. Tesla, Sunrun, and Renew Home announced a 16 GW VPP framework anchored at 300 MW in northern Virginia that intends to bid into PJM's Reliability Backstop. The week's cross-vertical thread: the AI-demand supply chain is being sorted by the federal procurement and permitting stack, not by price.
Utility-scale solar array under partly cloudy sky (file photo).
thesissolar

The IRS audit lookback on July 4 safe-harbored solar runs 6 years on aggressive readings, and the project-finance market is already pricing the documentation risk into tax-equity yield

The credit eligibility on the 216 to 240 GWdc of utility-scale solar safe-harbored into the July 4, 2026 OBBBA begin-construction deadline is, as a legal matter, locked at the project level. The exposure that remains, and that is now showing up in tax-equity term sheets and tax-insurance pricing, is the IRS audit lookback on whether each individual project actually established begin-of-construction under physical-work or 5 percent cost standards. Section 6501 baseline is 3 years from the placed-in-service return, but the 25 percent omission rule and substantial-understatement triggers extend the practical exposure window to 6 years, and the IRA-era 1603 grant precedent shows what a granular audit cycle looks like across a stockpile this size. The structural read for H2 2026 is that audit risk has rotated from a tail concern to a primary diligence axis for tax-equity investors, and the pricing reflects it.
High voltage transmission towers and conductors against an overcast sky (file photo).
thesissolar

The 216 to 240 GW solar safe-harbor stockpile coming out of the July 4 deadline is mortgaged against an interconnection queue that cannot clear it on the 2030 placed-in-service runway

Wood Mackenzie's estimate of 216 to 240 GWdc of US utility-scale solar safe-harbored between mid-2024 and July 4, 2026 is, by the industry's own framing, enough to cover projected installations through end-of-decade. The unwritten line in that framing is that the four-year continuity window only converts a safe-harbored project into a placed-in-service asset if the interconnection queue clears the corresponding capacity on a similar runway. Run the math by RTO using FERC Order 2023 cluster-study throughput, the FERC PJM expedited interconnection track, and the ERCOT large-load queue against the safe-harbored pool, and the post-July 4 binding constraint rotates from credit eligibility to interconnection deliverability. The 2030 question is not whether the stockpile exists. It is which fraction of it actually energizes.
Rows of utility-scale lithium-ion battery storage containers at a US grid site (file photo).
thesisstorage

The domestic-content BESS cell supply curve coming into H2 2026 is being rebuilt through ownership, not capacity, and the eligible-cell math is tighter than the announced gigawatt-hours suggest

Three deals in the last twelve days (AESC-Prevalon, Panasonic De Soto, T1 Energy-KORE) plus the Pentagon's June 8 Section 1260H expansion mark a shift in how the US battery-supply thesis should be read. The story is no longer 'how many GWh of cell capacity is sited in the US,' it is 'how many of those GWh are eligible for IRA-credit projects after the One Big Beautiful Bill Act's material-assistance test.' Run the math on credit-eligible incremental cell supply available to US grid-storage developers in H2 2026 and the curve is materially tighter than the announced 100-plus GWh of nameplate domestic capacity implies. Eligible supply is the binding constraint on storage deal flow into 2027, not nameplate.
High voltage transmission towers and conductors against an overcast sky (file photo)
thesispjm

PJM's 2028/2029 base residual auction closes July 7 with the price collar in place, and the question is not the clearing price but the cleared composition

The third PJM capacity auction under a $175 floor and $325 cap (with annual adjustments) closes July 7, 2026. The prior two auctions cleared at the cap, and the consensus expectation is that this one does as well. The data worth reading is not the headline number but the megawatt count of new entry, the share of storage cleared with current ELCC accreditation, the demand-resource volume, and which LDAs separate at higher locational caps. Those four lines determine whether the policy lever pulled at FERC on April 28 is shaping supply or only redistributing rent.
Sunlight Storage II battery energy storage system at the Desert Sunlight Solar Farm, Riverside County, California (file photo).
weekly-digestai-demand

weekly digest, june 8-14, 2026

The first of last week's three policy clocks landed: FERC approved PJM's Expedited Interconnection Track on June 9, opening a state-gated fast lane for up to ten 250 MW projects per year that can come online inside three years. The bigger demand-side clock (FERC RM26-4-000 on large-load interconnection) is now scheduled for the June 18 open meeting. Across the supply stack the same theme repeated in different chemistries and balance sheets: CATL signed the first commercial-scale sodium-ion offtake (60 GWh to HyperStrong), Panasonic disclosed plans to convert part of its 32 GWh De Soto cell factory from EV to data-center BESS, and the Pentagon added JA Solar, Trina, EVE, and CALB to the Section 1260H list with the direct DoD contracting ban starting end of June. Stoel Rives' OBBBA read confirmed that storage kept both safe-harbor paths and the 2027 placed-in-service backstop while solar above 1.5 MW lost both. ACP's Q1 2026 quarterly put solar and storage at 93 percent of new utility-scale grid additions. Thesis intact across all five active verticals; the supply chain is being reshaped in policy and capital structure faster than in price.
High-voltage transmission lines and substation infrastructure at dusk
thesisferc

FERC Order 2023 cluster studies, two years in: the queue is shorter, the withdrawals are larger, and the projects that survive are not the ones the pre-reform queue prioritized.

Order 2023 transitioned interconnection studies from first-come-first-served serial review to first-ready-first-served cluster review, with deposits, commercial readiness criteria, and withdrawal penalties. PJM, MISO, CAISO, ISO-NE, NYISO, and SPP have now run at least one full cluster cycle under the reformed rules. The headline result is a sharper queue, not a faster queue. The mix of projects clearing studies has shifted toward storage, hybrid storage-plus-solar, and load-paired generation. The pure-merchant solar queue has thinned. Whether the reformed process actually accelerates clean firm capacity onto the grid depends on the next cluster, not this one.
Drilling rig at sunset against arid mountain landscape, similar to southwest Utah geothermal sites
thesisgeothermal

Enhanced geothermal is the 2030 to 2032 lever, not the 2027 to 2029 lever. Fervo Cape Station is the test case.

Fervo Energy's Cape Station in southwest Utah is the first commercial-scale enhanced geothermal project to put steel in the ground in the US. The first 90 MW phase targets 2026 commissioning, with 400 MW total by 2028. The drilling cost reductions Fervo has demonstrated are real and large. They are not large enough to make EGS a 2027 to 2029 reliability story. They make it a 2030 to 2032 story, with the size of the role determined by the next 18 months of well results and the speed of the second-mover ramp behind Fervo.
Industrial gas turbine assembly hall with large rotor sections under overhead crane
thesisgas-turbines

The binding constraint on the 2027 to 2029 reliability window is no longer interconnection studies. It is the gas turbine OEMs.

Heavy-duty gas turbine orders at GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries have pushed delivery slots past 2029 for new bookings. The procurement queue, not the interconnection queue, is now the slowest piece of the reliability stack. Utilities chasing the data center load are paying reservation fees on equipment they will not see until 2030. Capacity auctions are pricing the gap.
High-voltage transmission tower against a blue sky, lines stretching to the horizon
thesisgrid

Grid-enhancing technologies were the throwaway line in FERC Order 1920. Two years in, they are the load-bearing assumption.

FERC Order 1920, finalized in May 2024, told regional transmission planners to consider grid-enhancing technologies (dynamic line ratings, advanced power flow controls, topology optimization, dynamic transformer ratings) before defaulting to new steel. The compliance filings RTOs submitted in early 2025 treat GETs as a checkbox. The IRPs and capacity expansion plans utilities are now writing for the 2027 to 2029 reliability window treat GETs as a delivery mechanism: a real megawatt of incremental transmission capacity, available in 12 to 24 months, while the multi-billion dollar reconductoring and new-build projects work their permitting timelines. That gap between regulatory framing and operational dependence is where most of the friction sits right now.
High-voltage transmission towers and lines at sunset in East Texas (file photo).
weekly-digestai-demand

weekly digest, june 1-7, 2026

Three federal and state policy clocks are running in the next 36 days: the July 4 IRA safe-harbor cliff for utility-scale solar, FERC's large-load interconnection ruling on Docket RM26-4-000, and MISO's first Order 1920 regional plan filing on June 12. A Carnegie Endowment paper put a number on the gap they are racing: every announced hyperscaler nuclear deal totals roughly 13 GW (about 102 TWh per year), less than 20 percent of projected US data-center demand through 2035. Private capital is already positioning around the answer: T1 Energy bought KORE Power for $32 million to add a 1,100-project BESS integrator to its US solar stack, a second solar-plus-storage tuck-in inside two weeks after Nextpower-Prevalon. USA Rare Earth committed $1.2 billion to a Cherokee County, South Carolina sintered NdFeB magnet and heavy-rare-earth metals plant, the first US site sized for heavy rare earth refining at scale. South Australia awarded 5.3 GWh of 15-year, 8-hour committed long-duration storage in its inaugural FERM tender, all lithium-ion. Painesville broke ground on a $80 million CPRG-funded coal-to-solar-plus-storage replacement, proof that IRA-obligated dollars are still moving into steel. Cross-vertical thread: policy is racing demand, and the marginal capex is already positioned for whichever way the clocks land.
Rows of server racks inside a large hyperscale data center with cooling infrastructure overhead
thesisgrid

Demand-side reliability is the third lever, and the data-center interruptibility deals are doing most of the work

Lever 3 in the 2027 to 2029 gas turbine gap series is demand response, but the version utilities are actually building bears little resemblance to the residential air-conditioner cycling programs that defined DR in the 2010s. The growth is concentrated in large-customer interruptibility contracts, with hyperscaler data centers now signing curtailment commitments that count for capacity-market accreditation and IRP planning. PJM's 2026 capacity auction cleared roughly 10.3 GW of demand resources, up from 7.8 GW the prior cycle. MISO and ERCOT show similar shifts. The shape of these contracts, what they actually obligate, and how reliably they perform under stress are the questions that will decide whether the lever holds in the years where capacity margins are tightest.
Large coal-fired power station with cooling towers and stacks against a hazy sky
thesisgrid

Coal life-extension is the second lever, and most state PUCs are quietly going along

Yesterday's piece identified four levers utilities are pulling to cover the gas turbine gap between 2027 and 2029. Lever 2, coal life-extension, has the lowest profile of the four and is also the one with the most regulatory inertia behind it. Across PJM, MISO, SPP, and the Southeast, a working list of at least 18 coal units originally scheduled for retirement between 2025 and 2029 has been deferred, re-permitted, or moved to seasonal or reliability-must-run service in the last 18 months. The state public service commissions handling these IRP amendments have largely approved them, often citing the same load-growth and reliability language utilities are submitting. The lever is unglamorous, carries real fuel, environmental, and rate-base risk, and is being pulled harder than the public reliability narrative reflects.
A high-voltage transmission corridor at dusk with lines fanning out from a substation toward distant generation
thesisgrid

Bridging the gas turbine gap: the four levers utilities are pulling for 2027 to 2029

With H-class delivery slots booked into 2030 and beyond, the load growth filings IRPs assumed are getting back-solved into something else. Storage acceleration, coal life-extension, demand-side reliability programs, and behind-the-meter customer deals are the four levers carrying the gap between when capacity is needed and when new gas can show up. None of them is a clean substitute. All four are already in the filings, and they are reshaping the 2027 to 2029 reliability picture in ways that capacity auctions have not yet caught up to.
An industrial power generation facility with cooling towers and transmission infrastructure
thesisgrid

Gas turbine lead times have stretched past 60 months at the H-class tier, and the orderbook is now a binding 2029 constraint

Heavy-duty gas turbines from the three major OEMs (GE Vernova, Siemens Energy, Mitsubishi Power) are now quoting 48 to 60 month delivery windows for H-class units, with some J-class slots booked into 2031. Utility IRPs across PJM, MISO, ERCOT, and SPP have leaned on new gas capacity to backstop reliability through the data center load build, and most filings published in the last six months are not adjusting their COD assumptions for what the OEMs are actually quoting. The next chapter of the grid-buildout supply-chain story is gas, not transformers.
A row of utility-scale battery storage containers at a substation under clear sky
thesisreliability

NERC's 2026 summer assessment puts the elevated-risk zones in regions that did not build storage

The 2026 Summer Reliability Assessment names three elevated-risk subregions: NPCC New England, MRO SaskPower, and WECC Northwest. None of them has built meaningful battery storage at the scale ERCOT and CAISO now run. The contrast is now sharp enough to read as a thesis: regions with 4-hour storage in the gigawatts have aged out of the summer-risk list, and regions without it have not.
Utility-scale photovoltaic power station in the United States (file photo).
weekly-digestai-demand

weekly digest, may 25-31, 2026

Last week was the institutional-scale story: a $67 billion utility merger, a 220 GW PJM intake, a 500 MW long-duration storage entry, a $450 million ERCOT BESS close. This week was the supply-side execution that those moves implied. Kings Mountain cleared federal permitting and the USGS sized the Appalachian lithium resource at 2.3 million tonnes, putting the Carolinas into the credible-near-term-supply column. Chinese NdPr oxide ran above the Defense Department's MP Materials price floor, meaning the Western rare-earth thesis is now being validated by spot pricing rather than federal backstop. DOE selected five private advanced-reactor developers to negotiate access to 20 metric tons of surplus weapons-grade plutonium, routing fissile inventory around the HALEU bottleneck. The NRC opened public comment on Part 57, a microreactor-specific licensing track projecting 6-to-12-month application-to-deployment timelines. Enbridge and Meta signed a $1.2 billion Wyoming solar plus 8-hour BESS deal sized for AI training load. SEIA logged the largest Q1 storage quarter on record (9.7 GWh, +32% YoY) and flagged hyperscaler procurement moving past lithium-iron-phosphate into iron-air and solid-oxide chemistries. Nextpower acquired Prevalon Energy for up to $365 million, pulling a US solar tracker incumbent into the BESS-integrator stack to capture hyperscaler offtake. Cross-vertical thread: AI demand is no longer an input to investment cases; it is the operating assumption now rewriting federal fuel allocation, M&A logic, and permitting reform across all five verticals in the same five business days.
Nuclear power station cooling towers at dusk (file photo)
thesisnuclear

NRC's proposed Part 57 reshapes microreactor unit economics, not just timelines, and that is the more important story

The Nuclear Regulatory Commission's Part 57 proposed rule, published May 1, 2026 and open for comment through June 15, does three things at once: fleet approval of identical designs, manufacturing licenses for factory-built units, and explicit permission for autonomous remote operation. Each one separately would matter. Together they change which microreactor projects pencil, which customers can actually take delivery before 2030, and where the next binding constraint lands. The 6-to-12-month application-to-deployment number the agency is publicizing is the headline. The deeper story is that the rule moves the chokepoint off licensing and onto HALEU fuel supply, project financing, and customer offtake.
High voltage transmission lines against a blue sky (file photo)
thesisferc

FERC's PJM co-location order is the most consequential grid-policy decision of 2026, and the tariff language now being drafted will set the terms for every nuclear-data-center deal in the eastern interconnection

On December 18, 2025, FERC directed PJM to draft a tariff framework for large loads co-located with generation, with PJM's terms-of-service brief due February 16, 2026 and a separate reliability informational report due January 19, 2026. The order proposes two new transmission products, Firm Contract Demand and Non-Firm Contract Demand, that a hyperscaler can take on behalf of a co-located data center in lieu of full Network Integration Transmission Service. The shape of those products, and the cost-allocation rules behind them, decides whether the next wave of nuclear-paired AI campuses gets built behind the meter, in front of the meter, or somewhere in between.
High-voltage transmission lines feeding into a utility substation (file photo)
thesisirp

Data center load is rewriting the utility IRP, and the resource plan is starting to push back against announced coal retirements

In the 2025 and early 2026 integrated resource planning cycles, the largest US investor-owned utilities have roughly doubled their forward load forecasts, with data center demand carrying the bulk of the revision. The new forecasts are flowing into the resource stack as deferred coal retirements, expanded gas peaker buildouts, and earlier retirement-replacement nuclear, not as a faster ramp on renewables. The constraint pattern is interconnection, transformer supply, and the gas turbine orderbook, in that order.
An electrical substation with power lines and transformer equipment under a clear sky
thesisgrid

Large power transformer lead times have stretched past 30 months, and the orderbook is now the binding constraint on grid buildout

Grid-scale step-up transformers in the 138 kV and 230 kV class are now quoting 24 to 30 month lead times at the major OEMs, with some 345 kV and 500 kV units pushing past 36 months. The bottleneck is not modules, cells, or capital. It is grain-oriented electrical steel, skilled assembly labor, and an OEM footprint that was sized for a different decade of demand. The constraint applies equally to solar, storage, nuclear, and data center substation builds.
High-voltage transmission lines crossing rural landscape (file photo).
weekly-digestpjm

weekly digest, may 18-24, 2026

The week broke wide across verticals. NextEra agreed to acquire Dominion in an all-stock $67 billion deal that, if cleared, makes the combined entity the second-largest US nuclear operator and the dominant generation counterparty inside PJM. Two days later PJM disclosed that Cycle 1 of its reformed interconnection process drew 811 projects at 220 GW, with nuclear posting a 17.9 GW queue entry that does not have a recent precedent. Hydrostor entered a 500 MW, 8,000 MWh A-CAES project into Ontario's long lead-time RFP, the third non-lithium long-duration capital event inside eight days. FERC affirmed PJM's at-risk readiness deposit framework, holding the cost discipline that keeps the cleaned-up queue clean. Spearmint closed $450 million on a 600 MWh ERCOT BESS with ITC transferability inside the stack and a three-month finance-to-COD window. Moment Energy notched a world-first UL safety certification for a second-life BMS, formalizing one of the few legitimate non-Chinese cell supply paths under the new FEOC material-assistance regime. And an independent lab found hexavalent chromium and arsenic in discharge from Tesla's Robstown lithium refinery, the most prominent US-domiciled lithium hydroxide facility, with the state's permit and prior investigation both missing the relevant analytes. Cross-vertical thread: AI-driven hyperscale load is now visible inside the same week across nuclear consolidation, PJM queue intake, and ERCOT BESS capital stacks, with the policy stack (FEOC, IRA transferability, FERC queue discipline) doing exactly the routing work it was designed to do.
Rows of utility-scale lithium-ion battery storage containers at a US grid site (file photo)
storagebatteries

US battery storage hit a record 9.7 GWh in Q1 2026, with 71% of utility-scale capacity going up in red states

SEIA and Benchmark Mineral Intelligence reported Q1 2026 US energy storage installations at 9.7 GWh, the strongest first quarter on record and a 32% year-over-year gain. Utility-scale accounted for 7.8 GWh of that. Texas, Arizona, and California led; together with the next handful of conservative-leaning states they hold 71% of utility-scale installs, even as federal policy uncertainty grows.
Nuclear power plant cooling towers (file photo)
thesisnuclear

Nuclear's comeback is real. The timeline isn't what the headlines say.

AI power demand, Palisades proving brownfield economics, and the NRC advancing multiple advanced-reactor permits have made nuclear the most-discussed energy story of 2025. The tailwinds are real. But new-build timelines run 8–15 years, SMRs won't move the needle before 2032, and the near-term investment thesis is brownfield restart and power uprates, not greenfield.
Battery storage containers at a US utility-scale site (file photo)
weekly-digestrebalance

Weekly digest, May 11–17, 2026

Three more operator prints completed the four-leg rebalance cross-confirmation (LAR, SGML, plus the bilateral read on Ganfeng); Lithium Americas put Thacker Pass capex on the guide line with an open $80–120M tariff bill disclosed; two named-risk events showed up in the same week (Alsym + Juniper 500 MWh sodium-ion procurement, Eos + Cerberus Frontier Power USA zinc-bromide IPP). The thesis frame holds. Risk #4 (alt-storage capturing grid share) gets re-weighted from named-but-unfunded to named-and-partially-funded.
Offshore wind farm at sunset (file photo)
thesispolicy

What the IRA cuts actually did and didn't do

The 'One Big Beautiful Bill' clawed back some clean energy credits and created real uncertainty for projects in development. But the most commercially embedded credits (Section 45X for domestic manufacturing and investment credits for operating facilities) survived largely intact. The transition math doesn't reverse. It slows, unevenly.
Open-pit mine (file photo)
thesiscritical-minerals

The US is ally-shoring, not reshoring, and that's probably the right call

The DPA executive orders, IRA sourcing requirements, and the Pentagon's MP Materials equity stake are building allied-nation supply chains, not domestic ones. True US domestic mineral production requires permitting timelines of 7–15 years. The real chokepoint (refining, not mining) is where the domestic strategy has the most leverage.
Utility-scale solar farm (file photo)
thesissolar

The tariff shock is accelerating the US solar supply chain rebuild

CVD rates above 3,400% on Southeast Asian imports are painful for projects buying modules today, but they're also the forcing function the IRA alone couldn't deliver: a US domestic manufacturing buildout with real urgency behind it. First Solar's structural position is understood; what's less discussed is how the module-cost shock reshapes developer strategy from 2026 onward.
High-voltage transmission lines (file photo)
thesisgrid

Transmission is the binding constraint and we're not building fast enough

There are 2,600+ GW of clean energy projects sitting in the US interconnection queue. The median wait is now 5 years. Every GW of solar, wind, and storage that gets built but can't connect to the grid is money and policy credibility lost. Reconductoring and MISO's LRTP are two of the few concrete moves. Neither is moving fast enough.
Lithium-carbonate refining facility (file photo)
weekly-digestalbemarle

Weekly digest, May 4-9, 2026

Albemarle's Q1 prints +148% adjusted EBITDA on a +51% pricing snap; Century Lithium drops a draft mine plan into BLM review at Angel Island; SQM calls 25% market growth for the year. The rebalance arrived. Underneath, the capital cycle keeps biting marginal names. Nothing yet that bends the thesis.
Lithium-carbonate refining facility (file photo)
weekly-digestganfeng

Weekly digest, May 4-10, 2026

Ganfeng's Q1 print confirms the rebalance on the China side, the chinese spot tape clears the structural threshold last weekend's digest set, Rio Tinto formally splits Aluminum & Lithium into its own division, and AMG's Q1 sets up Bitterfeld as the first western refining tell. The signal got bilateral; the alpha-in-being-early window is closing faster than we expected.
EV charging station (file photo)
thesisdemand

The EV slowdown is a misread of the lithium decade

Q1 2026 prices nearly doubled. Morgan Stanley and UBS now both forecast 2026 deficits. Storage-application demand grew ~71% in 2025. The 'oversupply' narrative that ran through 2024–2025 is breaking down, and EV-cycle thinking missed why.
Industrial refinery facility (file photo)
refiningpolicy

Refining is where the real chokepoint sits

Mining gets the headlines, but the supply-chain bottleneck for the next decade is refining. ~70% of refined lithium routes through China today, the spodumene-to-carbonate spread is the cleanest signal, and the marginal IRA-funded refining dollar is the most important number in the space.