The Ledger
Dated predictions related to the future of energy. Resolution dates set in advance. Outcomes recorded whether they were correct or not.
Forecasting in the energy and data center infrastructure sectors is rarely recorded and scored. Claims are made in passing at conferences or in trade articles. The timeline stays vague enough that nothing is ever “wrong” and the same people are invited back to make the next set of projections. The incentive runs against precision, so precision is what disappears first.
This page will be transparent, with every prediction carrying the date is was made, the date when it becomes true or not and the specific source / sources that will settle it. Entries are not edited after publication. When one becomes complete, the outcome is recorded, correct or incorrect and the entry stays on the page.
The reason for doing this is not accuracy for its own sake. An argument that cannot lose is an argument that is not worth as much. The frameworks presented elsewhere on this site, the Structured Transition Model, the Temporal Trilemma, Infrastructure Half-Life, Speed to Power, are only useful if they generate claims that can fail. These are those claims.
How entries work
Every prediction states a specific outcome, a resolution date, and a named source that settles it. No entry is added without all three.
Resolution sources are fixed when the prediction is made, not selected afterward.
Entries are never revised. Where circumstances change materially, or where resolution criteria are added to an earlier claim, a dated note is appended and the original text remains visible.
Predictions originate in published articles. This page records them. It does not replace the argument they came from.
Where a prediction resolves partially, or the resolving source becomes unavailable, it is marked unresolved rather than removed.
| # | Prediction | Made | Resolves | Resolution source | Status |
|---|---|---|---|---|---|
| 8 | At least one operating data center site will be served by reciprocating gas engine generation of 5 MW or greater nameplate, fueled by renewable natural gas, with post-combustion carbon capture installed on the engine exhaust, and the operator or developer will have publicly claimed net negative emissions for that generation. Test: capture equipment commissioned and operating on engines totaling 5 MW or more at a single site, generation serving data center load on site or behind the meter, and a net negative claim published by the operator or developer. Worldwide. The test is whether the plant operates and the claim is made, not whether the claim is independently verified. Captured CO2 must be routed to geologic storage or mineralization. Sale for utilization does not satisfy the test. Entered as a standalone prediction, August 2026 | 11 Aug 2026 | 31 Dec 2029 | Operator or developer announcement; CO2 injection or storage permit record; reporting in Data Center Dynamics, POWER, or Carbon Capture Journal | Open |
| 7 | Flexible, dispatchable computing load will be treated as an underwritten asset in at least ONE publicly disclosed data center financing. Stated as an observation in the original article and entered here as a scored prediction on August 8 2026. The Grid Is Learning to Dispatch Demand | 6 Aug 2026 | 31 Dec 2029 | Deal disclosure and trade reporting | Open |
| 6 | Residual value and redeployment assumptions will become headline underwriting criteria for data center power transactions. Test: stated as a criterion in the published underwriting or investment criteria of at least TWO named infrastructure lenders or funds. Infrastructure Half-Life | 6 Jul 2026 | 31 Dec 2029 | Lender and fund published criteria, investor materials | Open |
| 5 | A visible secondary market for data center bridge-power assets, including published pricing, will exist. Infrastructure Half-Life | 6 Jul 2026 | 31 Dec 2029 | Published transaction pricing from a broker, exchange, or trade index | Open |
| 4 | Where data center projects are delayed or curtailed during 2026 and 2027 under strong El Niño conditions, the principal cause will be constrained firm power and water availability rather than compute or chip supply. Test: of publicly reported delays or curtailments at named NORTH AMERICAN AND EUROPEAN sites in the period, a majority cite power or water. Resolution criteria added Aug 8 2026. A Super El Niño Tests the Transition | 18 Jun 2026 | 31 Dec 2027 | Operator disclosure and trade reporting | Open |
| 3 | If hyperscaler AI capital expenditure falls by more than [30] percent peak to trough before 2030, US data center electricity consumption in the year following the trough will still exceed consumption in the year preceding the peak, and no sustained surplus of power capacity will result. Resolution criteria added Aug 8 2026. The AI Bubble May Burst | 11 Jun 2026 | 31 Dec 2030 | EIA electricity data and Lawrence Berkeley National Laboratory data center energy reporting | Open |
| 2 | Power availability will constrain where data centers can be built, rather than data center demand determining where power is built. Published commentary, 2017 | 2017 | Resolved | Public record | Correct |
| 1 | Electricity cost, grid reliability, and cooling, rather than processor performance, will constrain data center development, and onsite generation will become more attractive as a result. See the note on wording below. Published article, March 2016 | Mar 2016 | Resolved | Public record | Correct |
Note on entry 1. The original March 2016 article argued that electricity cost, grid reliability, and cooling would constrain data centers, and that onsite generation would become more attractive. It did not use the formulation "power, not compute." That phrasing is a later compression of the same argument. The original wording is recorded above so the claim can be checked against the source rather than against my restatement of it.
Personal predictions, publicly sourced, never revised after publication. Entries are personal opinions stated as of the date shown, formed entirely from publicly available information. They are not guidance from any organization and are not investment advice. Some will be wrong.
Disclaimer
About these predictions
The predictions on this page are my personal opinions, held in a personal capacity, and stated as of the date shown against each entry. They are not forecasts, guidance, or statements made by or on behalf of any organization with which I am associated.
Every prediction on this page is formed from publicly available information. No confidential, proprietary, or non-public information from any employer, client, or third party informs any entry, and none is disclosed here.
Nothing on this page is investment advice, financial advice, or a recommendation to buy, sell, or hold any security, asset, or instrument. Nothing here should be relied on as the basis for any commercial, investment, or procurement decision.
These are predictions about market structure, technology sequencing, and infrastructure development. They are not predictions about the performance of any specific company, and no entry should be read as a statement about the prospects of any organization with which I am associated.
Entries are not revised after publication. Where circumstances change materially, a dated note is appended below the original text and the original remains visible. Some predictions will be wrong. That is the point of publishing them.