Evidence as of November 1, 2026. Every figure below comes from the broadcast record of the Comuvia News video of November 3, 2026, and carries the label that video gave it. Where two sources disagree, both are shown and the disagreement is left standing. Where a conclusion is Comuvia's rather than the record's, it is marked as a judgement. This companion piece adds no evidence of its own. Not investment advice.
Clean electricity has never been cheaper to generate. Battery cells leave Chinese factories near forty dollars a kilowatt-hour; solar modules have never cost less. And yet electricity bills in three American states are up more than twenty percent in a year, new reactors are late almost everywhere, and the companies building small reactors have lost most of their market value.
Those two facts look contradictory only if you assume the hard part of the energy transition is still making clean energy cheap. It isn't. That problem is largely solved. The binding constraint has moved downstream of the generator.
The system, in five stages
The fights in this story each live at one stage, and keeping them apart is most of the work.
Primary energy — uranium, sunlight, wind, heat underground. The 2026 constraint is not the fuel but the fuel cycle: who mines, converts and enriches it.
Generation — reactors, panels, turbines. The cheapest it has ever been. The constraint here is financing the first plant of any new design.
Firming and storage — the boom segment, though multi-day storage is one company deep.
Grid delivery — the bottleneck of the decade: the queues, the transformers, the voltage engineering.
Consumption — where AI data centres arrived years ahead of new supply and turned absorption and affordability into the political front line.
Stages one to three got cheap. Stages four and five now set the price.
Fission: more money, no new order
The scoreboard is lopsided. On July 31, China approved eight more reactors in a single decision. The United States started construction on zero new large reactors in the period; its story is restarts, not starts.
The American money is real. The Department of Energy offered $17.5 billion in loans for up to ten large reactors, and in October added two further conditional commitments — $4.2 billion to Vistra on October 5, and up to $5 billion to Constellation on October 27 — taking federal nuclear support past $34 billion.
But notice what the money buys. Both October loans are uprates at plants already running: more output from existing reactors, not new ones. As of the evidence cut-off there is still no binding utility order for a new large reactor. That is the thesis in miniature — capital is not the scarce input any more.
In Europe the cost problem rules instead: Hinkley Point C near £48 billion, and France's six-reactor programme up 41% before construction, with first power targeted for 2038.
Fusion: its richest year, and its physics unchanged
The video sets the case for and against directly against each other, because this is where the gap between money and watts is widest. Commonwealth Fusion's chief executive makes the case for ARC. The physicist Sabine Hossenfelder makes the counter-case: power sold from reactors that do not exist.
Rather than adjudicate, the video grades both against five gates — tritium breeding, magnets, materials, power handling, and the wall-plug balance — because those are what either side's timeline has to clear. The honest position is that fusion's hardest gates moved in the laboratory this year, and none of them moved in a power station.
The delivery wall, including as a computing problem
The wall itself is prosaic: queues where historically only about 13% of requests ever get built, transformer lead times of two and a half years, and voltage engineering that most coverage never reaches.
One item from the laboratory is aimed at delivery rather than supply, and it is worth the attention precisely because its headline overstates it. In September the municipal utility in Chattanooga, Tennessee switched on a commercial quantum computer inside its own operation — around 200,000 customers, and the first American site to put a commercial quantum machine and a quantum network under one roof. The problem it chose is this section's problem seen from the inside: distribution network reconfiguration, deciding which switches to open and close so power keeps flowing one way down each line while data centres, rooftop solar and car chargers keep moving the load.
At the end of September the utility and its vendor published three papers, and the papers are more modest than the press release. The learning model reaches the right answer with far fewer parameters. The sampling method speeds up a conventional solver — but the comparison is that solver against itself, with and without guidance, not against a quantum machine. The third finds the known best configuration in six of ten test cases. The test grids are textbook networks, not Chattanooga's own feeders, and nothing runs on a live circuit.
So grade it honestly. The headline says grid complexity is being solved; the papers beneath it say the method may become competitive as hardware improves. What is genuinely new is that a utility now treats the delivery wall as a computing problem.
Why the market reads it the way it does
Capital now rewards contracted, firm, deliverable electrons and punishes uncontracted long-duration promises. That single sentence explains most of the share-price divergence in the sector, and it follows directly from where the constraint sits. If delivery is the scarce thing, then an offtake contract is worth more than a technology claim.
How to read this
Every number in the video is named and dated on screen, with its source stated at the moment the claim is made, so each can be checked against the original rather than against our summary of it. Ten recordings carry the argument, all from the official channel of the person or outlet named on screen, with original audio: Copernicus on the climate ledger; Energy Secretary Chris Wright against the American Petroleum Institute on what replaces fossil fuels; the DOE loan office on the money; Commonwealth Fusion against Sabine Hossenfelder on fusion; Rep. Bob Latta (R-OH) against Sen. Martin Heinrich (D-NM) on data centres and household bills; Fervo Energy on enhanced geothermal; and CSIS for the most careful read we found.
Comuvia's own complexity judgement is labelled as a judgement, not as a finding.
This is research and analysis, not investment advice.