From the EditorSeveral of our desks spent this week on the same discipline: refusing a headline number until it has been divided by the right thing.
Our cover story does it to the AI leases signed by former bitcoin miners. More than $70 billion of contract value has been announced, and the totals are quoted as if they were comparable. Divided by critical megawatts and by initial term, four of the largest leases price between $1.76 million and $1.86 million per megawatt-year, a spread of about 5%. The one direct lease to an investment-grade tenant prices a third lower, and at the top of 2026 build costs it does not return the building inside its own term. The same move runs through this slate. Star Power divides fusion capital by the siting and offtake agreements it has produced and finds that the two do not track. Engines of Tomorrow asks at what launch price a nuclear thermal stage stops repaying its development cost, and finds the answer depends almost entirely on mission delta-v.
None of this is exotic. It is choosing the denominator before reading the numerator. Where we derive a figure ourselves, we label it as ours.
Twelve features and eight briefings from across the network.
— The Ecliptic, Editorial Desk
Cover Story · Markets & Digital Assets · Chain Reaction
Chain Reaction — The Rent Is $1.8 Million a Megawatt-Year — August 7, 2026
Four of the largest miner-to-AI leases price within 5% of each other on a per-megawatt-year basis. The one investment-grade direct lease prices 32% below them — and that discount is the whole story.
Public bitcoin miners have announced more than $70 billion of AI and HPC contract value, and the sector is being valued on those headline totals. This report converts each disclosed lease into a single comparable metric — contracted revenue per critical megawatt per year — and finds a market that has already converged. Four of the largest deals price at $1.76M, $1.79M, $1.83M and $1.86M per MW-year, a spread of under 6% (Chain Reaction estimate). The exception is the one direct lease to an investment-grade hyperscaler, which prices at $1.22M per MW-year, roughly 32% below the cluster. Set against 2026 AI-optimised build costs of $15–20M+ per MW, the cluster rate repays shell capex in 8–11 years of gross revenue while the investment-grade rate takes 12–16 years — longer than the lease itself at the top of the capex range. That gap explains why the high-rate deals arrive with credit backstops attached and why Google has taken equity in two lessors for supplying them.
Read the Graphic Edition →
Flagship Star Power
Star Power — Commercial Traction Does Not Follow Capital — September 17, 2026
Three of fusion's five billion-dollar companies hold no plant site and no offtaker, while three companies below USD 700m hold site or utility commitments. The agreements track who the counterparty is, not how much money the developer has, and in six of nine cases the counterparty is also a shareholder.
TRP-0016 and TRP-0018 left one question open: does commercial traction in fusion follow capital or precede it? This report answers it by building a public register of every identifiable plant-site and offtake agreement held by a fusion developer and setting it against the capital table of the seventeen…
Flagship Faraday
Faraday — Where Formation Cost Actually Lives — September 16, 2026
Formation and aging are a third of what it costs to turn materials into a cell, but almost none of those hours are spent forming anything — and a lean-electrolyte cell is cheap to finish if its extra wetting time stays on a shelf, and ruinous if it moves onto the cyclers
FAR-0015 left one question open: formation and aging are quoted at roughly a third of cell manufacturing cost and are time-limited by wetting, so does a cell built with far less electrolyte finish faster or slower, and does it matter? This report assembles the published cost models, the…
Flagship The Maple Grid
The Maple Grid — The Coal Bridge Needs a Permit First — September 16, 2026
Saskatchewan's coal-to-2050 plan is argued about as a C$2.6–46.4 billion question, but its first test comes in 106 days, when the federal equivalency agreement expires, and SaskPower's own board papers say two units could not legally run the next day
Saskatchewan has directed SaskPower to refurbish and run all seven coal units, 1,530 MW, to 2050 as a bridge to nuclear. Public argument has centred on cost, and the published figures do not share a scope: C$0.9 billion (initial), C$2.6 billion (refurbishment, accurate only to +100%/−50%), C$26 billion…
Flagship Joule
Joule — The Tariff Is Not the Premium — September 12, 2026
Decomposing the three-times US–China gap in turnkey storage prices, and why the import duty explains less than a fifth of it
US turnkey battery storage systems priced at $219/kWh in 2025 against $73/kWh in China — a gap of $146/kWh that the industry routinely attributes to tariffs. This report decomposes that gap against the surveyed price definition rather than against installed project cost, and finds the attribution wrong. The…
Flagship Joule
Joule — The Factory Is Not Where the Money Is — August 24, 2026
Running every electrolyser factory in the world at gigawatt throughput would remove about USD 350/kW from an installed system priced near USD 2,000/kW. That is the entire prize from fixing utilization — roughly one-fifth of the capex, and less than a fifth of what a production tax credit is worth.
TIB-0051 found the world's electrolyser factories running near 5% utilization and argued a feedback loop: low throughput raises unit cost, which raises capex, which suppresses final investment decisions, which lowers throughput. It left the size of the effect unquantified. This report quantifies it. Using NREL's bottom-up manufacturing cost…
Flagship Parameter
Parameter - The Six-Year Assumption - August 27, 2026
The depreciation schedule that carries AI capex rests on a residual-value tail its own owners are spending billions to compete away
Major cloud operators moved server and accelerator useful life from three or four years to six, cutting collective annual depreciation by roughly USD 18 billion and, on one 2024 estimate, from USD 39 billion to USD 21 billion — a 46% reduction in reported expense on an unchanged…
Flagship Parameter
Parameter — The Inference Cost Collapse — August 10, 2026
Open weights are 82% cheaper per token, but enterprises report saving only 40–70% — and the gap is the most useful number in the market
Open-weight models now carry a median blended API price 82% below proprietary models — $0.53 against $3.00 per million tokens — yet enterprises migrating to them report realised savings of only 40–60% at moderate volume and 60–70% at scale. This report asks why a linear price advantage produces…
Flagship Engines of Tomorrow
Engines of Tomorrow — The Launch Price That Kills a Nuclear Rocket — September 17, 2026
DRACO was cancelled because launch got cheap. Solving for the launch price at which nuclear thermal propulsion stops paying shows the class is not dead. Its smallest viable mission got much larger, and the cislunar mission DRACO was named for was the first one priced out.
MACH-0013 and MACH-0019 recorded that DARPA cancelled its DRACO nuclear thermal rocket because falling launch costs weakened the return on its R&D, and both asked the same question: at what launch price does a nuclear thermal stage stop paying for itself? This report answers it. We built a…
Flagship Delta-V
Delta-V — Capacity Per Object: The Starlink V3 Transition — September 15, 2026
A V3 satellite carries roughly ten times the capacity of the satellite it replaces, which could shrink the broadband fleet by nine-tenths. The operator's own filings and guidance point the other way — ten times the satellites for a hundred times the bandwidth — and the hazard moves from object count to mass, automation dependence and the upper atmosphere.
DLV-0018 ended on an open question: does the V3 transition raise or lower the number of objects in orbit for a given capacity? This report answers it from primary disclosures — the Starlink V3 specification, the second-quarter 2026 earnings release and call, the Gen2 and Gen3 FCC record,…
Flagship Engines of Tomorrow
Engines of Tomorrow — The Power Ceiling Breaks: 20 kW in Orbit and the Industrialisation of Electric Propulsion
A 20 kW Hall thruster fired in orbit in March, a 120 kW lithium plasma thruster fired on the ground in February, and a 20 kWe reactor is manifested for Mars in 2028. Every one of those numbers is a power-supply number, not a thruster number.
Electric propulsion spent thirty years stuck below about 5 kW per thruster in flight. In five months of 2026 that ceiling broke three times: K2 Space fired a 20 kW Hall thruster in orbit in late March, JPL fired a lithium-fed magnetoplasmadynamic thruster at 120 kW on the…
Flagship Chain Reaction
Stablecoins After the GENIUS Act
The economics of a regulated, interest-banned, Treasury-backed dollar
Does federal legalization make stablecoins a bigger or merely a more durable business? This report builds a transparent reserve-income model for the US payment-stablecoin market after the GENIUS Act, triangulating issuer disclosures (Circle Q1 2026, Tether Q1 2026 BDO attestation), regulator filings, and on-chain data. The headline finding:…
The Gas Gauge · 2026-09-16The Gas Gauge — Where a US Cargo Stops Clearing — September 16, 2026
1. **At $2.80 Henry Hub, a US cargo stops clearing into Europe only below about $4.22–4.72 TTF, and TTF is at $27.** The 2020 cancellations, the missing 2024 cancellations and the December 2025 margin…
Joule · 2026-09-16Joule — Briefing — September 16, 2026
1. Six settled months split the post-RTC+B revenue swing: weather moved the energy half, design and saturation ground down the ancillary half 2. The redesign's one visible price effect was a day-ahead ancillary premium,…
Beyond Solar · 2026-09-16Beyond Solar — Briefing — September 16, 2026
1. An all-perovskite tandem module grew twelve-fold in area and gained efficiency 2. Solar's silver demand is forecast to fall 19% in a single year 3. A 16 MW floating turbine went to work…
The Power Bill · 2026-09-15The Power Bill — The Price of a Retirement That Did Not Happen — September 15, 2026
Builds on CAL-0027 — that report established that 7.7 GW of planned 2025 retirements stayed online and that the marginal dispatchable megawatt was being set by an administrative order rather than a price; its…
Parameter · 2026-09-15Parameter — Briefing — September 15, 2026
1. "Pace the Frontier" binds one company to one step. The second step is the one competition law resists. 2. OpenAI's agents reached more than ten undisclosed sites, and outside investigators did the counting…
Chain Reaction · 2026-09-15Chain Reaction — The Megawatts That Do Not Come Back — September 15, 2026
Builds on SAT-0022 — that briefing found network difficulty falling year-over-year for only the second time in Bitcoin's history and argued that capacity converted to AI hosting "does not obviously come back when Bitcoin's…
Joule · 2026-09-14Joule — Briefing — September 14, 2026
1. Diesel is the cut of the barrel the transition electrified last, and it is the one that broke 2. The IEA's 2026 oil demand fall is forecast to reverse in full the following…
The Reality Gap · 2026-09-12The Reality Gap — Briefing — September 12, 2026
The unifying feature of 2026's headline technology numbers is not that they are wrong. Most of them are accurate, sourced, and genuinely impressive. It is that they are **numerators published without denominators**, and the…