Research library
Energy systems
View the four-date completion graphs and reconstruction notes.
Promise and evidence
AI could improve energy-system operation, equipment design, storage, and research into new generation methods. DeepMind and collaborators' plasma-control work demonstrates learned control in fusion research. Controlling plasma does not establish a commercially useful fusion power plant.
Apply the shared measurement protocol. Proposed metrics cover separate subareas; do not combine them into one undifferentiated score.
Dated baseline coverage
All linked measurable-outcome files below contain an evidence baseline reviewed November 7, 2026, with source dates, findings, limitations and follow-up measurements. These distinguish measured results from proxies and unknown values; they do not assert that proposed milestones are achieved. See the baseline coverage notes.
Progress estimates
Judgment-based estimates as of November 7, 2026, using the same 0–100% scale and five-part rubric as mathematics. 0% anchors the period before ChatGPT's public release in 2022; 100% requires AI to manage this field's continuing advancement without human participation. Confidence is low. These are evidence-based judgments, not empirical completion measurements, forecasts or percentages of jobs replaced.
| Measurable outcome | Estimated progress | Execution / breadth / agenda / validation / improvement (each out of 20) |
|---|---|---|
| Fusion control | 35% | 10 / 5 / 5 / 10 / 5 |
| Grid reliability | 0% | 0 / 0 / 0 / 0 / 0 |
| Net energy | 0% | 0 / 0 / 0 / 0 / 0 |
| Operational efficiency | 0% | 0 / 0 / 0 / 0 / 0 |
| Research productivity | 30% | 10 / 5 / 5 / 5 / 5 |
| Storage performance | 35% | 10 / 5 / 5 / 10 / 5 |
Each linked outcome explains its score, cites its dated evidence baseline and identifies the next measurement. Zero indicates no qualifying post-anchor progress established by the reviewed evidence for that outcome, not proof of no relevant capability. Do not average these overlapping indicators into a field-wide percentage or interpret their differences as a precise ranking of fields. All outcomes reaching 100% would still require a separate field-wide demonstration of sustained autonomous advancement.
Measurable features
| Feature | Measurement and acceptance rule |
|---|---|
| Operational efficiency | Delivered kWh per unit input, operating cost per MWh, and losses under matched demand, weather, equipment, and reliability conditions. |
| Grid reliability | Unserved energy, outage duration, constraint violations, and renewable curtailment compared with established control methods. |
| Storage performance | Measured round-trip efficiency, useful cycle life, and lifetime cost per delivered kWh under the same duty cycle. |
| Fusion control | Plasma tracking error, stable operating duration, disruption rate, and experimental reproducibility on specified hardware. |
| Net energy | Report plasma gain, facility energy balance, and net electricity delivered separately. Count all relevant energy inputs and auxiliary systems. |
| Research productivity | Experiments and elapsed time per independently verified advance, with total human and compute effort. |
Evidence requirements
Separate simulation, controlled experiment, pilot installation, and operational deployment. Use metered outcomes and matched or randomized comparisons where feasible. Include hardware changes, maintenance costs, model computation, and rebound effects when relevant. Evaluate reliability constraints alongside efficiency; simulated savings or plasma gain must not be relabeled as delivered commercial electricity.
Proposed milestones
Reduce a prespecified operating loss or cost by at least 5% over 90 days under comparable conditions, with no deterioration beyond registered reliability limits.
Independently reproduce the improvement at a second facility across a different operating season or load regime.
Demonstrate positive net economic value over a full year, including integration and maintenance costs. For fusion, use a separate milestone requiring independently measured net facility electricity rather than a control benchmark.
Quantify how energy improvements reduce the cost or resource constraints of subsequent AI research. General energy gains are valuable without necessarily demonstrating this connection.