Leadership Under Uncertainty

AI data centres (version 2) · All records

The numbers are stated judgments. The record began on 15 October 2026 with starting confidences that a model set as the average of five separate readings, from its general knowledge and up to ten developments for each assessment first reported in the three weeks before, after a second model had checked the developments the readings relied on, and a published rule has made every change since. They are not measured probabilities. The question and the rule are fixed before the record begins, and each change shows its evidence and its reasons. The rule and what is not claimed are on the method page.

This record corrects the first record of the same question, AI data centres, which began on 12 October 2026 and continues unchanged under the rules it began with. What changed, and why is on this record’s method page.

The update loop’s working notes

What sets the pace of AI data centre buildout in the United States. These notes move nothing: only a sealed change, made by the rule, moves a number.

Last checked for news: 15 October 2026, 11:10 UTC. The update loop checks once a day, and each day’s state is written at the end of the day.

What to expect. The rule moves a number only after at least 2 independent publishers report the same development within 7 days of its first report. One development then moves an assessment by at most about 2.2 percentage points when an outlet graded high reports it, and by under one percentage point when every outlet reporting it is graded unknown. Of the 75 outlets the update loop has read for this record so far, 63 are graded unknown. The update loop has not yet read a development new to the record since it began.

Since the record began

Developments first reported since the record began

These are the update loop's working notes: each development it read, how it read it, and what the published rule makes of that reading. They are not part of the sealed record and they move nothing. Only a sealed change, made by the rule, moves a number.

Since the record began, the update loop has not yet read a development first reported after it began.

The daily states record what the update loop read from the record's second day, and none has been sealed yet. Both count a development once for each assessment it bears on. The daily states also count reports, while these notes list developments, and a report retrieved after the record began but first reported before it is counted in the daily states and never weighed by the rule.

Each outcome is worked out by the published rule from what the loop recorded when it read the development: its outlets then, its reading, and the challenge step's verdict once that has run. The outlet grades are the loop's own.

No development first reported since the record began has been read yet.

Before the record began

Developments read before the record began

Before the record began, the update loop ran each day as it does now but could write nothing. The rule does not weigh a development first reported before the record began, and these readings did not set the starting point, which a separate reading set.

  • Read 15 October 2026, 04:12 UTC · The alternative
    • “Utilities Struggle to Confront Issue of ‘Phantom' AI Data Centers” · breitbart.com · 13 October 2026 · Open the report · archived copy
    First reported 13 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor C5 (Power still given as the reason for delays). Read as routine.
    • On the assessment: less likely.
    • On alternative CX1: more likely. This tells the alternative apart from the assessment.
    • On alternative CX2: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The report describes utilities tightening connection requirements due to speculative requests swelling grid queues, which is consistent with power/interconnection still being a binding constraint (C5 strengthened), making the demand-led 'if_wrong' scenario less likely. CX1 is more likely because the tightening of requirements in response to speculative/phantom requests suggests queues may shrink while wait times remain long.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    • “Utilities Struggle to Confront Issue of ‘Phantom' AI Data Centers” · breitbart.com · 13 October 2026 · Open the report · archived copy
    First reported 13 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M1 (Waits for large grid connections). Read as routine.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Utilities tightening connection requirements due to speculative data-center requests swelling grid queues directly supports longer effective waits for large grid connections (M1), reinforcing that grid power and interconnection remain the binding constraint. This is commentary/reporting on a trend rather than a specific regulatory order or project action, so it is routine.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · The condition it rests on
    • “SOAR AI Secures Up to $1 Billion Equity Facility to Fund Expansion of Its AI Infrastructure Platform” · businesswire.com · 12 October 2026 · Open the report · archived copy
    First reported 12 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor R3 (Financing for data centre campuses). Read as material.
    • On the assessment: more likely.
    • On alternative RX1: less likely. This tells the alternative apart from the assessment.
    • On alternative RX2: more likely.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): SOAR AI closing a $1 billion equity facility for an Ohio AI campus backed by customer orders is a concrete financing event that shows capital continues to flow into new data centre construction, keeping demand pressure on power. The facility is arranged outside the largest cloud companies' balance sheets, which directly favours RX2 (spending shifting to specialized providers) while contradicting RX1 (spending levelling off).

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    First reported 12 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M1 (Waits for large grid connections). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The report states that Oncor's interconnection queue reached 271 gigawatts in Q1, a concrete filed figure indicating massive backlogs of large load requests awaiting grid connection in Texas, directly evidencing long waits for grid interconnection as a binding constraint. This scale of queued demand favours power as the bottleneck over chips, capital, permitting, or demand-side limits.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    First reported 12 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as routine.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: more likely.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The report describes officials weighing moratoriums and tighter rules citing grid strain and electricity costs, which strengthens M2 (special terms and pauses for large loads). However, concerns about tax breaks and community opposition also favour MX3, since the political backlash may reflect land-use and fiscal grievances rather than purely grid-capacity constraints.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · The alternative
    First reported 11 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor C5 (Power still given as the reason for delays). Read as routine.
    • On the assessment: less likely.
    • On alternative CX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative CX2: less likely.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The report describes GPU supply constraints and rising rental prices as the binding factor, which would support the 'if_wrong' assessment, but it also notes that data-center power infrastructure lags—reinforcing C5 that power delays persist and the grid still binds, which weakens the 'if_wrong' scenario. The GPU scarcity element is commentary/market research rather than a material filing or contract, and the simultaneous mention of power infrastructure lag means the development does not clearly favour cheaper computation or demand softening (CX2).

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    • “Powering the AI revolution: Inside the grid, thermal, and interconnect challenges limiting data center growth” · techradar.com · 11 October 2026 · Open the report · archived copy
    First reported 11 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M1 (Waits for large grid connections). Read as routine.
    • On the assessment: more likely.
    • On alternative MX1: less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The report explicitly identifies grid connections and transformer availability as major constraints on AI data-centre growth, with connection waits cited as up to 14 years in Northern Virginia, directly supporting M1 (long waits for grid connections) and the thesis that power is the binding constraint. However, this is commentary/analysis rather than a regulatory order or project filing, so it is routine.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · The condition it rests on
    • “Alphabet's Google Cloud Is Surging at 82%, and the Heavy Spending Is Starting to Make Sense” · theglobeandmail.com · 11 October 2026 · Open the report · archived copy
    First reported 11 October 2026. Reported by 1 outlet when it was read. Outlet grade: high.
    Strengthens factor R2 (Capital spending outlook). Read as material.
    • On the assessment: more likely.
    • On alternative RX1: less likely. This tells the alternative apart from the assessment.
    • On alternative RX2: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Alphabet guided to $195–205 billion in 2026 capex, a substantial figure reflecting continued heavy investment in AI infrastructure, which strengthens the view that capital spending keeps rising and demand keeps pressing on power constraints. The note that supply is limiting ability to meet demand directly supports the thesis that power/interconnection is the binding constraint rather than capital. RX2 is neither favoured nor disfavoured since the report concerns Alphabet's own balance-sheet spending, not a shift to specialized providers.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    • “Electric avenues: AI data centers spur natural gas, solar; challenge power grids” · reflector.com · 11 October 2026 · Open the report · archived copy
    First reported 11 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M5 (Ways around the grid: on-site power, co-location and flexible load). Read as routine.
    • On the assessment: neither more nor less likely.
    • On alternative MX1: neither more nor less likely.
    • On alternative MX2: neither more nor less likely.
    • On alternative MX3: neither more nor less likely.
    • On alternative MXC: neither more nor less likely.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The report describes data-centre developers turning to on-site natural gas and solar to work around grid and equipment backlogs, which strengthens M5 (ways around the grid via on-site generation). However, the report also references 'equipment backlogs' alongside grid delays, and as a general commentary piece rather than a specific project approval or contract, it does not materially shift confidence in the main assessment or any alternative.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    • “AI Data Center Grid Interconnection - The Innovation Attorney | Michael Kimball” · theinnovationattorney.com · 11 October 2026 · Open the report · archived copy
    First reported 11 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M1 (Waits for large grid connections). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The report cites Berkeley Lab data showing 2,061 GW queued for interconnection with only 13% of 2000–2020 requests reaching commercial operation, and a PJM capacity auction clearing 6,831 MW short of the reliability requirement at the price cap—concrete figures indicating severe grid congestion and long waits for large loads, directly strengthening M1 and the thesis that power/interconnection is the binding constraint rather than chips, capital, permitting, or demand.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    • “Map Shows States Taking Action to Protect Ratepayers from Data Center Power Costs” · newsweek.com · 9 October 2026 · Open the report · archived copy
    First reported 9 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: more likely.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Multiple states are reported to be adopting cost-allocation rules, reviews, and permitting pauses specifically weighing data-center growth against grid costs, which constitutes special regulatory terms rationing connection capacity and strengthens M2. MX3 is also more likely because some of the cited actions involve permitting pauses and community impacts rather than purely grid capacity constraints.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    First reported 9 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): FERC directing regional grid operators to justify or revise tariffs for large-load interconnections is a regulatory action that directly bears on M2 (special terms for large loads), indicating regulators are actively managing how scarce interconnection capacity is allocated to data centres. This strengthens the view that grid power and interconnection are a binding constraint requiring regulatory intervention, not chip supply, capital, permitting, or demand-side limits.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 04:12 UTC · Main assessment
    • “Oracle Trucks in Natural Gas to Outrun Pipeline Delays at AI Data Centers” · webpronews.com · 9 October 2026 · Open the report · archived copy
    First reported 9 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor M5 (Ways around the grid: on-site power, co-location and flexible load). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Oracle is actively trucking natural gas to data centre sites to work around delayed pipeline and grid infrastructure — a concrete on-site power arrangement that exemplifies M5 (ways around the grid). The fact that this workaround is needed confirms grid/pipeline access is the binding constraint, not chips, capital, permitting, or demand.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “PJM Will File Its Own Data Center Ride-Through Rule in November, Before NERC Sets Thresholds” · mgrid.org · 10 September 2026 · Open the report · archived copy
    • “PJM Proposes Reliability Standards to Manage Large Load Disconnection Events | PJM Inside Lines” · insidelines.pjm.com · 10 September 2026 · Open the report · archived copy
    • “PJM proposes ride-through rules for large data center loads | TechTarget” · techtarget.com · 15 September 2026 · Open the report · archived copy
    And 2 more reports of the same development. First reported 10 September 2026. Reported by 5 outlets when it was read. Outlet grade: medium, the best of them.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): PJM's proposed ride-through rules, filed in response to a FERC order and triggered by a nearly 4,000 MW data center load disconnection event, represent new regulatory requirements specifically targeting large computational loads. These rules impose additional technical and compliance obligations on data centers connecting to the grid, reinforcing that grid reliability and interconnection management are active constraints on large-load buildout, consistent with M2's definition of special terms rationing scarce connection capacity.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports confirm that PJM has proposed new interconnection reliability requirements specifically targeting large computational loads like data centers. Imposing special technical terms on large new loads to manage grid reliability directly supports Factor M2 and strengthens the assessment that grid interconnection is a binding constraint. None of the alternatives are equally favoured by this grid-specific regulatory action, and it is a new development not previously counted.

    • Report 2, quoted: “PJM has proposed changes to interconnection reliability requirements for Large Loads, specifically, computational loads that include data centers and crypto-mining facilities.”
    • Report 3, quoted: “PJMs proposed ride-through rules would require large data centers to stay connected during grid disturbances rather than switching to backup.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · The alternative
    • “The machine that the AI boom cannot get enough of — and the wait is years-long” · washingtontimes.com · 30 September 2026 · Open the report · archived copy
    • “Commentary: The two-year waitlist to power modern life - Grand Forks Herald | Grand Forks, East Grand Forks news, weather & sports” · grandforksherald.com · 22 September 2026 · Open the report · archived copy
    • “The Steel Box Gatekeeping Trillions in Grid Spending – Stock Report” · stockreport.com · 26 September 2026 · Open the report · archived copy
    And 12 more reports of the same development. First reported 28 August 2026. Reported by 3 outlets when it was read. Outlet grade: medium, the best of them.
    Strengthens factor C5 (Power still given as the reason for delays). Read as material.
    • On the assessment: less likely.
    • On alternative CX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative CX2: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Multiple reports citing Wood Mackenzie, DOE-funded analysis, and CoBank data confirm transformer and substation equipment lead times of 3–4 years and gas turbine shortages, meaning power delivery constraints remain the stated reason for project delays. This strengthens C5 (power still given as reason for delays), which weakens the if_wrong scenario that demand rather than power sets the pace.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports directly state that shortages of gas turbines, multi-year lead times for transformers and substation equipment, and interconnection constraints are acting as critical scheduling constraints that threaten to slow data center buildouts. This matches Factor C5, as power and grid equipment bottlenecks continue to be cited as the reason for buildout delays, which weakens the assessment that demand alone sets the pace.

    • Report 1, quoted: “A shortage of gas turbines and long equipment lead times are complicating efforts to supply data centers with firm power, while tight grid capacity and interconnection constraints threaten to slow buildouts”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · The condition it rests on
    First reported 13 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor R3 (Financing for data centre campuses). Read as material.
    • On the assessment: more likely.
    • On alternative RX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative RX2: more likely.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Hyperscale Data is closing project debt financing for a Michigan AI data-centre campus serving a neocloud customer, which is a concrete financing event keeping construction moving. This also fits RX2 because it shows capacity being built through a specialized provider financed via a special-purpose debt structure rather than on a hyperscaler's own balance sheet.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “The machine that the AI boom cannot get enough of — and the wait is years-long” · washingtontimes.com · 30 September 2026 · Open the report · archived copy
    • “Commentary: The two-year waitlist to power modern life - Grand Forks Herald | Grand Forks, East Grand Forks news, weather & sports” · grandforksherald.com · 22 September 2026 · Open the report · archived copy
    • “The Steel Box Gatekeeping Trillions in Grid Spending – Stock Report” · stockreport.com · 26 September 2026 · Open the report · archived copy
    And 12 more reports of the same development. First reported 28 August 2026. Reported by 3 outlets when it was read. Outlet grade: medium, the best of them.
    Strengthens factor M4 (Lead times for transformers, switchgear and turbines). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Reports from Wood Mackenzie, DOE-funded lab analysis, and CoBank document that large power transformer lead times have stretched from 12–18 months to 30–36+ months, generator step-up units average ~144 weeks, and substation equipment broadly runs 3–4 years—directly strengthening M4 (lead times for transformers, switchgear and turbines) and confirming power-side equipment as a structural constraint on buildout pace.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports explicitly document multi-year lead times for large power transformers, substation equipment, and gas turbines, which directly strengthens factor M4 (Lead times for transformers, switchgear and turbines). While some cited analyses date to May and July 2026, the reports also reference Wood Mackenzie's newest survey and current ongoing shortages without an earlier date, supporting a small step in favour of the main assessment.

    • Report 2, quoted: “Standard power transformers now take 2.5 years to deliver, according to data analysis firm Wood Mackenzie’s latest survey ; the larger transformers used at substations — the fenced hubs that route power between plants and neighborhoods — are running three years in the firm’s newest survey and can take up to four.”
    • Report 4, quoted: “Large power transformer lead times have stretched from roughly 12 to 18 months pre-shortage to roughly 30 to 36 months today, based on a DOE-funded national laboratory supply chain gap analysis published in May 2026 and related industry survey data. Generation step-up units now run about 144 weeks on average in commonly cited industry surveys.”
    • Report 1, quoted: “A shortage of gas turbines and long equipment lead times are complicating efforts to supply data centers with firm power, while tight grid capacity and interconnection constraints threaten to slow buildouts and raise electricity costs.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “Energy Department Supports FERC’s Call for PJM to Protect Ratepayers and Build More Reliable Power | Department of Energy” · energy.gov · 8 October 2026 · Open the report · archived copy
    • “DOE presses PJM on ratepayer protections from large load costs | Utility Dive” · utilitydive.com · 8 October 2026 · Open the report · archived copy
    • “DOE Presses PJM On Ratepayer Protections From Large Load Costs” · zerohedge.com · 9 October 2026 · Open the report · archived copy
    First reported 8 October 2026. Reported by 3 outlets when it was read. Outlet grade: high, the best of them.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The DOE formally urged PJM to require large new electricity users (data centers) to bear the costs they cause, and FERC has called for ratepayer protections—this represents federal regulatory pressure to impose cost-allocation terms on large new loads, directly fitting the M2 definition of special terms rationing scarce connection capacity. This strengthens the case that grid interconnection rules are tightening around large loads, making power the binding constraint more likely.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports confirm that the Department of Energy urged PJM to require large new loads, including data centers, to pay for the generation and reliability costs they cause rather than shifting them to existing ratepayers. This federal push for stricter cost-allocation rules on large loads directly aligns with Factor M2 (special terms for large loads) and is a new development not previously counted. None of the alternatives are equally supported by grid regulators tightening cost terms on large power connections.

    • Report 1, quoted: “DOE urges PJM to require large new electricity users to pay for the generation”
    • Report 2, quoted: “The U.S. Department of Energy on Oct. 7, 2026, urged the PJM Interconnection to quickly revise its reliability backstop procurement plan to prevent cost shifts onto existing utility customers from data centers and other large loads.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · The alternative
    • “Data Center Physical Infrastructure Market Grows 27 Percent Y/Y To $14 Billion In 2Q 2026, According To Dell'Oro Group” · menafn.com · 14 October 2026 · Open the report · archived copy
    First reported 14 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Weakens factor C5 (Power still given as the reason for delays). Read as routine.
    • On the assessment: less likely.
    • On alternative CX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative CX2: less likely.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): DellOro reports that power, long-lead equipment, permitting, and grid interconnections are stretching project timelines, directly citing power and interconnection as near-term constraints. This strengthens C5 (power still given as reason for delays), which weakens the 'if_wrong' scenario that demand sets the pace. CX2 is also made less likely since capital spending is reported as growing strongly (27% Y/Y), not slowing.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · The condition it rests on
    • “Alphabet Signals More AI Spending After $80.6 Billion First-Half Outlay” · tokenpost.com · 13 October 2026 · Open the report · archived copy
    First reported 13 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor R2 (Capital spending outlook). Read as material.
    • On the assessment: more likely.
    • On alternative RX1: less likely. This tells the alternative apart from the assessment.
    • On alternative RX2: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Alphabet's signal of increased 2027 capital expenditure, following $80.6 billion in first-half 2026 spending, directly raises the forward spending outlook (R2) and confirms continued large-balance-sheet demand pressing on power infrastructure. This contradicts RX1's scenario of spending leveling off, while RX2 is unaffected since this concerns Alphabet's own balance sheet spending.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · The condition it rests on
    • “Fermi Signs Credit Support Agreement with AMD and Announces Partnership” · digitaljournal.com · 13 October 2026 · Open the report · archived copy
    First reported 13 October 2026. Reported by 1 outlet when it was read. Outlet grade: unknown.
    Strengthens factor R3 (Financing for data centre campuses). Read as material.
    • On the assessment: more likely.
    • On alternative RX1: less likely. This tells the alternative apart from the assessment.
    • On alternative RX2: more likely.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): A credit support agreement between Fermi and AMD for a Texas AI campus represents closed financing outside the largest cloud companies' balance sheets, directly evidencing R3 (campus financing that keeps construction going). It also supports RX2 because a specialized provider (Fermi) is securing capacity through a partnership/credit arrangement rather than through a hyperscaler's own capital spending, while the project's focus on dedicated power and compute underscores continued demand pressing on the power constraint.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “FERC approves Duke large load transmission rules despite cost-shift concerns” · utilitydive.com · 13 October 2026 · Open the report · archived copy
    First reported 13 October 2026. Reported by 1 outlet when it was read. Outlet grade: medium.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): FERC's approval of Duke Energy Carolinas' tariff provisions for large-load transmission upgrades and financial security requirements represents a regulatory order imposing special terms (financial security, cost-allocation rules) on large new loads, directly matching the M2 definition of utilities and regulators setting special terms to ration scarce connection capacity. This strengthens the case that grid interconnection is a binding constraint rather than chips, capital, permitting, or demand.

    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “PJM Interconnection Reforms Aim to Speed Up Grid Access | Utilities Curated” · utilitiescurated.com · 27 August 2026 · Open the report · archived copy
    • “PJM Revisits Surplus Interconnection Rules | Developments Today” · developmentstoday.com · 24 August 2026 · Open the report · archived copy
    First reported 24 August 2026. Reported by 2 outlets when it was read. Outlet grade: unknown, the best of them.
    Strengthens factor M1 (Waits for large grid connections). Read as routine.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The reports describe PJM's surplus interconnection pathway as largely ineffective in practice—only 2 of 8 applications approved since 2023—indicating that even reform efforts aimed at speeding grid access have delivered little, reinforcing that long interconnection waits remain a binding constraint. No evidence is presented that chips, capital, permitting, or demand are the limiting factor.

    Disputed by the challenge step, written by Gemini 4 Argon (Google): Factor M1 measures how long a large new load waits between requesting a grid connection and energization. The reports instead describe PJM's surplus interconnection rules as a pathway for 'new energy resources to integrate' at existing power plants, not connection waits for large new loads. Moreover, the current development is that PJM is revisiting and reforming these rules to 'Speed Up Grid Access,' which moves in the opposite direction from lengthening connection waits.

    Its ground: what the reports state is not a change in this causal factor.
    • Report 1, quoted: “PJM Interconnection Reforms Aim to Speed Up Grid Access”
    • Report 1, quoted: “provides a pathway for new energy resources to integrate without exceeding previously approved safety limits”
    • Report 2, quoted: “since 2023, PJM has received eight SIS applications and approved only two of them”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “PJM Fast-Tracks 2.1 GW Capacity to Meet Rising Demand — EnergyPulse” · opensentiment.org · 10 October 2026 · Open the report · archived copy
    • “PJM accepts two ENGIE batteries and a gas uprate on its fast track | The Duck Curve” · theduckcurve.com · 9 October 2026 · Open the report · archived copy
    • “PJM fast track approves first 2.1 GW: Engie storage, LS Power gas uprate — Grid Brief” · gridbrief.org · 10 October 2026 · Open the report · archived copy
    And 1 more report of the same development. First reported 8 October 2026. Reported by 4 outlets when it was read. Outlet grade: unknown, the best of them.
    Weakens factor M1 (Waits for large grid connections). Read as material.
    • On the assessment: less likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): PJM's Expedited Interconnection Track has approved its first 2.1 GW of projects with a target of signing interconnection agreements within 10 months of submission, which shortens the wait between request and energization for qualifying large resources. This weakens M1 (longer waits) and thus modestly reduces the tightness of the grid-as-binding-constraint thesis, though the track is capped at 10 requests per year and runs only through 2027, limiting its overall impact.

    Disputed by the challenge step, written by Gemini 4 Argon (Google): Factor M1 is defined as how long a large new load (such as a data centre) waits between its request for a grid connection and energization. The reports state that PJM's Expedited Interconnection Track is for 'new or uprated capacity resources' (power generation and storage, specifically two battery storage projects and a gas uprate) to sign a 'Generation Interconnection Agreement,' not for connecting new load.

    Its ground: what the reports state is not a change in this causal factor.
    • Report 2, quoted: “PJM accepts two ENGIE batteries and a gas uprate on its fast track”
    • Report 2, quoted: “The track runs beside PJMs regular interconnection queue and takes up to 10 requests per calendar year in 2026 and 2027 for large new or uprated capacity resources.”
    • Report 2, quoted: “PJM expects projects to sign a Generation Interconnection Agreement within 10 months”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    First reported 7 October 2026. Reported by 3 outlets when it was read. Outlet grade: medium, the best of them.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Duke Energy and North Carolina stakeholders reached a formal settlement requiring large-load customers (data centres) to fund dedicated facilities, provide security deposits, and meet minimum payment obligations for grid upgrades — precisely the kind of special financial terms and cost-allocation rules that define M2. This is a regulatory filing/settlement, not mere commentary, and it reinforces that grid connection capacity is being rationed through financial conditions rather than that chips, capital, or permitting are the binding constraint.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports state that Duke Energy and stakeholders reached a proposed North Carolina settlement requiring large-load customers to cover connection costs, provide security for grid upgrades, and make minimum payments. This directly matches factor M2, which covers utilities and state regulators setting special financial terms, collateral, and minimum-demand charges for large new loads. The development is new and is not explained equally well by any of the listed alternatives.

    • Report 1, quoted: “Duke Energy said a proposed North Carolina settlement would require large-load customers to fund dedicated facilities and provide security for shared grid upgrades, subject to state regulatory approval.”
    • Report 2, quoted: “Duke Energy and large customers proposed North Carolina rules requiring major electricity users to cover connection costs and make minimum payments, addressing who bears the cost of data-center-driven grid expansion.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “Constellation's 20-Year PPA Deal with Google to bring 890 MW of Nuclear Capacity through Uprates to PJM Grid | Energy Tech” · energytech.com · 6 October 2026 · Open the report · archived copy
    • “Constellation Signs 20-Year Deal with Google to Unlock 890 MW of New Nuclear Capacity - ESG Today” · esgtoday.com · 6 October 2026 · Open the report · archived copy
    First reported 6 October 2026. Reported by 2 outlets when it was read. Outlet grade: unknown, the best of them.
    Strengthens factor M5 (Ways around the grid: on-site power, co-location and flexible load). Read as material.
    • On the assessment: less likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Google has signed a 20-year PPA with Constellation for 890 MW of nuclear uprates delivered directly to the PJM grid, a large-scale arrangement that lets a major hyperscaler secure power outside a conventional grid upgrade queue. This strengthens M5 (ways around the grid) because it is an approved, contracted arrangement delivering substantial capacity sooner than new grid infrastructure would allow, which means grid power binds less for Google's buildout and thus weakens the thesis that grid interconnection is the binding constraint.

    Disputed by the challenge step, written by Gemini 4 Argon (Google): Factor M5 covers arrangements that bypass conventional grid upgrades or queues to deliver power directly to a campus, such as on-site generation, co-location beside a plant, or flexible load. The reports state that Google and Constellation signed a PPA to bring 890 MW of nuclear uprates across 11 units directly onto the PJM grid. Financing grid-connected generation capacity via a PPA does not bypass grid interconnection for a data centre campus or provide a way around the grid as defined by M5.

    Its ground: what the reports state is not a change in this causal factor.
    • Report 1, quoted: “Global tech giant Google and utility power producer Constellation are bringing 890 MW of new, emissions-free generation onto the PJM Interconnection grid through a 20-year power purchase agreement (PPA).”
    • Report 1, quoted: “enable investments in new equipment and technology at 11 Constellation-owned nuclear units in Illinois, Pennsylvania and New Jersey”
    • Report 2, quoted: “adding 890 MW of new nuclear energy to the U.S.’ PJM grid.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “FERC Accepts and Suspends PJM’s Reliability Backstop Procurement Proposal, Establishes Paper Hearing Procedures, and Encourages a New Section 205 Filing | Troutman Pepper Locke - JDSupra” · jdsupra.com · 7 October 2026 · Open the report · archived copy
    • “FERC Suspends PJM's Reliability Backstop. The September 30 Window Does Not Open — Hyperscape Digital” · hyperscapedigital.com · 5 October 2026 · Open the report · archived copy
    • “FERC Accepts and Suspends PJM’s Reliability Backstop Procurement Proposal, Establishes Paper Hearing Procedures, and Encourages a New Section 205 Filing” · lexblog.com · 6 October 2026 · Open the report · archived copy
    And 1 more report of the same development. First reported 5 October 2026. Reported by 4 outlets when it was read. Outlet grade: unknown, the best of them.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): FERC's acceptance and suspension of PJM's Reliability Backstop Procurement proposal, with a paper hearing specifically on LSE collateral requirements and cost allocation, represents a regulatory proceeding that directly bears on the special financial terms (collateral, deposits) utilities and regulators impose on large new loads in PJM territory. The suspension delays implementation until at least February 2027, prolonging uncertainty around how large-load interconnection costs and collateral will be structured, which tightens the power-as-binding-constraint dynamic. None of the reports cite chip supply, financing costs, local permitting, or demand shortfalls.

    Disputed by the challenge step, written by Gemini 4 Argon (Google): The reading claims that the RBP Proposal bears on special financial terms utilities and regulators impose on large new loads. However, the reports only mention 'Load-Serving Entity (“LSE”) collateral requirements' and 'Transmission Owner (“TO”) exit provisions' within a Reliability Backstop Procurement proposal, without stating that this proposal applies to or sets terms for large new loads or data centers. Because the link to large new loads is not stated in the reports, the move is unsupported.

    Its ground: the reports do not state the development the reading relies on.
    • Report 1, quoted: “establishing a paper hearing on three elements of the RBP Proposal: (1) cost allocation; (2) Transmission Owner (“TO”) exit provisions; and (3) Load-Serving Entity (“LSE”) collateral requirements”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · The alternative
    • “Oracle's $Multi-Billion AI Data Center in Wisconsin Faces Years of Delays Over Power Grid Approvals” · webpronews.com · 2 October 2026 · Open the report · archived copy
    • “PSC wipes out 10 months of progress on Oracle's 1.3 GW Wisconsin data center — pushing partial 310 MW power to late 2027 and full load to 2029” · techradar.com · 4 October 2026 · Open the report · archived copy
    • “Oracle's Wisconsin AI data center hits a Wisconsin power line roadblock” · startupfortune.com · 5 October 2026 · Open the report · archived copy
    First reported 2 October 2026. Reported by 3 outlets when it was read. Outlet grade: unknown, the best of them.
    Strengthens factor C5 (Power still given as the reason for delays). Read as material.
    • On the assessment: less likely.
    • On alternative CX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative CX2: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Wisconsin regulators restarted the transmission review for Oracle's 1.3 GW campus, pushing partial power to late 2027 and full load to 2029. This is a concrete regulatory action causing multi-year delay explicitly due to power grid approvals, directly strengthening C5 (power still given as the reason for delays) and making the demand-paced scenario less likely.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports explicitly state that Oracle's 1.3 GW AI data center in Wisconsin faces multi-year delays due to a restarted regulatory review of its transmission infrastructure and power grid approvals, pushing partial power to late 2027 and full load to 2029. This directly matches factor C5 (projects slipping because of power), which undercuts the assessment that demand rather than grid access sets the pace of buildout.

    • Report 1, quoted: “Oracles $Multi-Billion AI Data Center in Wisconsin Faces Years of Delays Over Power Grid Approvals”
    • Report 2, quoted: “PSC wipes out 10 months of progress on Oracles 1.3 GW Wisconsin data center — pushing partial 310 MW power to late 2027 and full load to 2029”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “Oracle's $Multi-Billion AI Data Center in Wisconsin Faces Years of Delays Over Power Grid Approvals” · webpronews.com · 2 October 2026 · Open the report · archived copy
    • “PSC wipes out 10 months of progress on Oracle's 1.3 GW Wisconsin data center — pushing partial 310 MW power to late 2027 and full load to 2029” · techradar.com · 4 October 2026 · Open the report · archived copy
    • “Oracle's Wisconsin AI data center hits a Wisconsin power line roadblock” · startupfortune.com · 5 October 2026 · Open the report · archived copy
    First reported 2 October 2026. Reported by 3 outlets when it was read. Outlet grade: unknown, the best of them.
    Strengthens factor M3 (Projects delayed, reduced or moved because of power). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): The PSC's decision to wipe out 10 months of regulatory progress on Oracle's 1.3 GW Wisconsin data centre, delaying initial 310 MW delivery to late 2027 and full capacity to 2029, is a direct instance of a major project delayed because of power grid approvals (M3). No mention of chip shortages, financing problems, or demand concerns as the cause, making accelerator supply, capital, and demand-paced alternatives less likely.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports explicitly state that Oracle's 1.3 GW data center in Wisconsin is delayed due to a regulatory decision on power grid and transmission infrastructure approvals, pushing partial power to late 2027 and full load to 2029. This directly matches factor M3, which covers projects delayed because of the availability or timing of power. None of the four challenge grounds apply.

    • Report 1, quoted: “Oracles $Multi-Billion AI Data Center in Wisconsin Faces Years of Delays Over Power Grid Approvals”
    • Report 2, quoted: “PSC wipes out 10 months of progress on Oracles 1.3 GW Wisconsin data center — pushing partial 310 MW power to late 2027 and full load to 2029”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “FERC asks grid operator PJM to revise plan to shield homes from data center costs” · reuters.com · 30 September 2026 · Open the report · archived copy
    • “Biggest US Grid's Data-Center Plan Accepted With 5-Month Pause” · bloomberg.com · 30 September 2026 · Open the report · archived copy
    • “Grid manager delays data center power auction over concern on who gets the bill” · cleveland.com · 1 October 2026 · Open the report · archived copy
    And 12 more reports of the same development. First reported 30 September 2026. Reported by 5 outlets when it was read. Outlet grade: high, the best of them.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): FERC's suspension of PJM's procurement plan and its order that large new loads (data centers) bear their attributable costs—combined with the identified 6.8-GW capacity shortfall driven by data center demand—are regulatory actions that impose special terms on large loads and confirm grid capacity as a binding constraint. These are concrete regulatory orders and findings, not mere commentary, directly strengthening M2 and the thesis that grid power is the binding constraint.

    Disputed by the challenge step, written by Gemini 4 Argon (Google): These reports cover FERC's September 30 order suspending PJM's special reliability backstop capacity auction for five months. This is the same development already counted as item 8 ('FERC Suspends PJMs Reliability Backstop. The September 30 Window Does Not Open'). Counting it again would double-count the same regulatory action.

    Its ground: it restates a development already counted.
    • Report 2, quoted: “FERC accepted PJM’s proposal to accelerate new capacity procurement but suspended it until February 28 while opening proceedings on unresolved issues.”
    • Report 3, quoted: “FERC suspended PJM’s planned emergency-power auction for up to five months, saying PJM had not adequately ensured that large new loads driving costs would pay their share”
    • Report 5, quoted: “PJM Targets 6.8-GW Reliability Gap With Special Backstop Auction”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “Amazon deal will enable $3 billion Calvert Cliffs investment, power data centers” · baltimoresun.com · 30 September 2026 · Open the report · archived copy
    • “Amazon and Constellation sign 20-year nuclear power purchase agreement for 690 MW at Calvert Cliffs — Private Infrastructure Daily” · privateinfrastructuredaily.com · 5 October 2026 · Open the report · archived copy
    First reported 30 September 2026. Reported by 2 outlets when it was read. Outlet grade: unknown, the best of them.
    Strengthens factor M5 (Ways around the grid: on-site power, co-location and flexible load). Read as material.
    • On the assessment: less likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Amazon has signed a 20-year, 690 MW nuclear PPA with Constellation at Calvert Cliffs, including backing an uprate, which is a direct on-site/co-location-adjacent arrangement that lets Amazon secure large-scale power outside conventional grid interconnection queues. Per the M5 definition, when such an arrangement is approved and delivers power to a large campus, grid power binds less as a constraint, making the main thesis less likely.

    Disputed by the challenge step, written by Gemini 4 Argon (Google): The reading claims this PPA is an on-site or co-location-adjacent arrangement that lets Amazon secure power outside conventional grid interconnection queues, which is required to strengthen Factor M5 (ways around the grid). However, neither report states that any data center is co-located at Calvert Cliffs or bypassing the grid; instead, the reports describe a power purchase agreement supporting "additional nuclear output for the PJM region" and a "retail supply deal for Amazons PJM operations." Because the reports do not state that the arrangement bypasses conventional grid connection via on-site power, co-location, or flexible load, the reading relies on an unstated fact.

    Its ground: the reports do not state the development the reading relies on.
    • Report 1, quoted: “Amazon agreed to a 20-year power purchase arrangement for up to 690 megawatts from Constellation, supporting a $3 billion Calvert Cliffs investment and additional nuclear output for the PJM region.”
    • Report 2, quoted: “The agreement covers 690 MW from the 1,790-MW Maryland plant, including 190 MW from an uprate the contract is backing, plus a separate retail supply deal for Amazons PJM operations.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “PJM launches annual review of large load requests for 2027 forecast - Solmar Capital” · solmarcapital.co · 30 September 2026 · Open the report · archived copy
    • “PJM begins deciding which AI data centers count in its 2027 forecast” · computelaw.blog · 2 October 2026 · Open the report · archived copy
    First reported 30 September 2026. Reported by 2 outlets when it was read. Outlet grade: unknown, the best of them.
    Strengthens factor M1 (Waits for large grid connections). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): PJM's annual review process for large load requests directly governs whether new data centre demand gets incorporated into the grid's long-term forecast and capacity planning, meaning projects left out of the forecast will not have transmission built to serve them—extending effective wait times for grid connection. The tightened guidelines emphasizing project completion likelihood further ration scarce grid capacity, consistent with power being the binding constraint.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports confirm that PJM has tightened the guidelines for including large load requests, such as AI data centers, in its long-term load forecast. Because PJM plans new transmission lines against that forecast, projects excluded by the stricter screening will not have grid infrastructure built to serve them, which directly bears on waits and studies for large grid connections under factor M1. None of the four challenge grounds apply.

    • Report 1, quoted: “This process, occurring amid heightened regional power demand largely due to data center growth, now incorporates tightened guidelines emphasizing the likelihood that projects will actually reach completion.”
    • Report 2, quoted: “PJM buys power plant capacity and plans new transmission lines against that estimate ... left out of it is a project nobody is building the grid to serve.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “FERC rejects Oklo complaint seeking to reinstate project in PJM interconnection study cycle” · utilitydive.com · 25 September 2026 · Open the report · archived copy
    • “FERC rejects Oklo's bid to put a 750 MW project back in PJM's queue — Grid Brief” · gridbrief.org · 26 September 2026 · Open the report · archived copy
    • “FERC Rejects Oklo Bid to Restore 750 MW Project to PJM Queue – American Buildout” · americanbuildout.com · 25 September 2026 · Open the report · archived copy
    And 1 more report of the same development. First reported 25 September 2026. Reported by 4 outlets when it was read. Outlet grade: medium, the best of them.
    Strengthens factor M1 (Waits for large grid connections). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): FERC's rejection of Oklo's complaint means a 750 MW generation project in Virginia—a high-demand data-centre corridor—is removed from PJM's active interconnection study cycle, forcing Oklo to restart the queue process and extending the timeline for new power supply to reach the grid. This is a concrete regulatory outcome that lengthens interconnection waits (M1), reinforcing that grid interconnection processes are a binding constraint, without bearing on chip supply, capital availability, permitting opposition, or demand-side dynamics.

    Disputed by the challenge step, written by Gemini 4 Argon (Google): Factor M1 measures how long a large new load waits between its request for a grid connection and energization. The reports describe a nuclear power company's 750-megawatt generation project being removed from PJM's generator interconnection study cycle for failing to provide a workable application model, rather than a wait time for a large new load requesting connection.

    Its ground: what the reports state is not a change in this causal factor.
    • Report 3, quoted: “The Federal Energy Regulatory Commission (FERC) has turned thumbs down on Oklo Inc.’s bid to get back in line for a 750-megawatt power project to PJM Interconnection’s first reformed study cycle.”
    • Report 3, quoted: “FERC’s said it found that that the company failed to provide a workable model for the full facility after receiving a chance to correct its application.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    And 4 more reports of the same development. First reported 23 September 2026. Reported by 3 outlets when it was read. Outlet grade: medium, the best of them.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: more likely.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): Texas Governor Abbott directed TCEQ to halt all data-center permits pending ERCOT's audit of a 474-GW interconnection queue, constituting a regulatory pause that explicitly cites grid capacity and water/environmental concerns. This strengthens M2 (special terms/pauses for large loads citing grid capacity) and thus the main thesis, while also partially favouring MX3 since the freeze extends to environmental permitting beyond grid interconnection alone.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports confirm that Texas paused data-center permits pending an ERCOT audit of a 474-GW interconnection queue and impacts on the electric grid. This state-level pause tied to grid capacity directly supports factor M2, while the environmental permitting aspect is already accounted for in the reading's assessment of MX3.

    • Report 4, quoted: “On September 21, 2026, Texas Governor Greg Abbott directed the Texas Commission on Environmental Quality (TCEQ) to halt issuance of permits to data centers until the Electric Reliability Council of Texas (ERCOT) completes an audit of impacts on the Texas electric grid and water resources.”
    • Report 6, quoted: “Texas Puts Data-Center Permits on Hold as ERCOT Considers 474 GW in Requests”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
  • Read 15 October 2026, 03:55 UTC · Main assessment
    • “FERC rejects ComEd notice tied to data-center transmission deal; contract dispute moves to court” · diariobitcoin.com · 24 September 2026 · Open the report · archived copy
    • “FERC rejects ComEd data center contract cancellation | LYU LLC” · lyullc.com · 24 September 2026 · Open the report · archived copy
    • “FERC Rejects ComEd $20B Data Center TSA Cancellation · TFTC” · tftc.io · 24 September 2026 · Open the report · archived copy
    And 1 more report of the same development. First reported 23 September 2026. Reported by 4 outlets when it was read. Outlet grade: medium, the best of them.
    Strengthens factor M2 (Special terms for large loads: tariffs, deposits and pauses). Read as material.
    • On the assessment: more likely.
    • On alternative MX1: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX2: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MX3: neither more nor less likely. This tells the alternative apart from the assessment.
    • On alternative MXC: neither more nor less likely. This tells the alternative apart from the assessment.

    Its reason, written by Claude Sonnet 4.6 (Anthropic): FERC's rejection of ComEd's cancellation of the 1.8-GW transmission security agreement, combined with its directive requiring RTOs/ISOs to submit new pro forma cost-recovery and credit rules by mid-November, directly illustrates utilities and regulators setting special terms (collateral, credit instruments) for large data-centre loads and the active regulatory effort to ration and govern large-load interconnection capacity. This strengthens the case that grid interconnection terms are a binding constraint, with no evidence bearing on chip supply, capital costs, local permitting, or demand-side limits.

    Upheld by the challenge step, written by Gemini 4 Argon (Google): The reports confirm a new regulatory decision on September 22 where FERC rejected ComEd's attempt to cancel a 1.8-GW data center transmission security agreement over a minimal credit instrument and tied the ruling to upcoming requirements for RTOs and ISOs to submit new large-load cost-recovery and credit rules. This directly involves utilities and regulators establishing and disputing collateral and cost-recovery terms for large data center loads, fitting Factor M2. None of the alternative explanations are supported by these reports.

    • Report 2, quoted: “A $1 letter of credit on a $20-billion, 1.8-GW project became the focal point of the decision and drew a direct rebuke from Commissioner David LaCerte.”
    • Report 3, quoted: “FERC tied the ruling directly to its broader large-load interconnection reform push, with RTOs and ISOs required to submit new pro forma cost-recovery rules by mid-November, rules that will govern Bitcoin miners and data centers alike.”
    Outcome: The record already knew it when it began, so the rule does not weigh it.
The record

Where the sealed record is

The numbers, the changes and their fingerprints are on the judgment’s page and the changes page. How the starting point was read, and the rule the loop applies to developments first reported before the record began, are on the method page.