Reading the Timeline Honestly: Filings, Deadlines, and the Constraint

Let me close the course by teaching you to read a nuclear timeline the way a skeptical analyst does, because the gap between what is announced and what is delivered is where most of the money and credibility in this sector will be won or lost.

Start with a discipline: separate the press release from the docket. When a developer announces that it will deliver power to a data center by, say, 2030, that date is a hope until it is anchored to regulatory reality. The anchor you look for is the filing. Has the developer actually submitted its application to the NRC, and if so, when? The clock on a multi-year review does not start when the press release goes out; it starts when the application is docketed and the agency begins its work. A company promising power in five years that has not yet filed is promising something the calendar may not permit, because the review alone can consume a large share of that window before a shovel moves. The filing date is the honest starting gun, and I would train you to ask for it every time.

Recognize, too, that the review is not the only step, merely the first hard one. After the NRC approves, you still have to build, and building a first-of-a-kind reactor carries its own schedule risk on top of the licensing risk. So a realistic timeline for a novel SMR stacks several multi-year phases: finish the design, secure the fuel, complete the licensing review, construct the unit, commission and test it, and only then deliver power. Each phase can slip, and in a first-of-a-kind project they often do. This is why sober observers treat the late-2020s delivery dates attached to some SMR announcements with caution and view the 2030s as more plausible for meaningful commercial volume. The restart timelines are shorter and more credible precisely because they compress or skip several of those phases, which is the throughline of this entire course.

Now, the policy layer, because it is genuinely changing and you should track it without overrating it. There has been real legislative and administrative effort to accelerate nuclear licensing, including provisions directing the NRC to modernize and speed its processes and to prepare for the wave of advanced-reactor applications. The intent is clear and the direction is real. But intent is not delivery. An agency built for certainty does not become fast by statute overnight; it adapts gradually, hires and trains reviewers, and works through a growing queue. So when you hear that licensing has been streamlined, treat it as a favorable trend to verify in the actual review durations, not as an accomplished fact. Watch how long real applications actually take under the new framework. That evidence, accumulating over the next few years, will tell you whether the acceleration is genuine or aspirational.

Here is the synthesis I want you to carry out of this course. Power is the first-order constraint on AI infrastructure. That is the thesis, and everything we have covered is downstream of it. The data-center buildout needs firm, clean, dense power at a scale and on a timeline that collapses the option space toward nuclear. Nuclear responds along two paths: restarts, which are fast, finite, and nearly certain because they exploit assets and a regulatory understanding that already exist; and SMRs, which are larger, slower, and higher-variance because they depend on unproven manufacturing economics and first-of-a-kind licensing. The economics of both are made financeable by the long, firm offtake contract that a creditworthy hyperscaler can provide, which is the ingredient the last nuclear cycle lacked. And the binding constraint on how fast any of it arrives is the regulator and its clock, which favors the known over the novel and moves at the speed of certainty rather than the speed of demand.

So when the next announcement crosses your desk, run it through the frame this course has built. Is it a restart or a new build? Who is the offtaker, and is the contract long and firm enough to make the project financeable? What does the NRC have to approve, is the technology familiar or novel, and has the developer actually filed? Does the fuel exist? Answer those questions and you will know, better than the headline does, when and whether the electrons will actually flow. In a sector crowded with optimistic dates and grand ambitions, that discipline, of reading the constraint rather than the promise, is the most valuable thing you can take away. The power constraint is real, nuclear is one serious answer to it, and the honest measure of any nuclear claim is how it fares against the clock and the regulator, not how it reads in the release.