The Physical Economy Doesn’t Care What You Announced
Every physical infrastructure narrative right now runs on the same currency: announcements. A gigawatt target. A funding round. A production goal. A partnership. These are the numbers that move into headlines, into pitch decks, into the slide an investment committee sees before a decision gets made.
We think that currency is broken, and we built Stratum Atlas because of it.
An announcement is a stated intention. It is not a delivery. The gap between the two is not a rounding error, it is where almost all of the risk in physical infrastructure investing actually lives, and right now almost nobody is pricing that gap correctly, because almost nobody is measuring it.
Why the gap exists, and why it’s growing
Physical infrastructure investing has inherited its information habits from software investing, where an announcement and a delivery are often close enough in time that the distinction barely matters. A software company announces a feature and ships it in a quarter. The market has learned to treat the announcement as a reasonable proxy for the outcome.
Physical infrastructure does not work that way, and the gap between announcement and delivery is getting wider, not narrower, as the buildout accelerates. A transformer ordered today arrives in 128 to 174 weeks, not next quarter. A gigawatt of interconnection queue capacity has a roughly one in seven chance of actually reaching commercial operation, based on two decades of withdrawal data across US regional grid operators. A humanoid robotics developer’s stated production target depends on a gearbox supply chain that, as of this year, has zero disclosed allocation agreements behind it for any of the eight largest developers combined.
None of this is secret. It is disclosed, in filings, in dockets, in supply agreements, in the same documents everyone has access to. The information asymmetry is not that the data doesn’t exist. It’s that almost nobody does the unglamorous work of pulling it out of a hundred separate primary sources, cross-referencing it, and updating it every time a new filing lands.
What we actually track, and why it’s different from research
We do not publish opinions about where infrastructure investing is headed. We track what has been disclosed, and we build the constraint math from there. A valve, the specific physical or financial choke point where a buildout actually stalls. A gatekeeper, the entity capturing margin downstream of that choke point. A scarcity type, whether the constraint is structural and multi-year or cyclical and likely to resolve on its own. A resolution condition, the specific, dated, checkable thing that would have to happen for the constraint to ease.
That is a different object than a research report. A research report tells you what a team of analysts believes. A constraint registry tells you what is disclosed, cited to its source, updated when the source updates, and corrected in public when we get it wrong. We publish that correction log permanently, next to the number it corrects, because a track record only means something if the mistakes are visible along with the calls that were right.
Why this matters more now than it did two years ago
The physical economy is absorbing more capital, faster, than at almost any point in the last several decades. Hyperscaler capital expenditure alone is projected to exceed a trillion dollars over three years. Utility capital expenditure is running at a pace not seen since post-war electrification. Humanoid robotics has raised billions of dollars against a production base still measured in the thousands of units. Quantum computing carries a captive reservation on cooling infrastructure worth more than ten billion dollars from a single buyer, while everyone else waits months for the same equipment despite an apparent surplus.
At this scale, the gap between what is announced and what is disclosed is not a research curiosity. It is the difference between a fund correctly sizing a position and a fund discovering, eighteen months into a hold, that the gigawatt it underwrote was never going to reach commercial operation on the timeline it was sold on.
What we believe, stated plainly
We believe the physical economy is investable, at scale, right now, and we believe most of the capital moving into it is being sized against the wrong numbers. We believe the correction is not more forecasting. It is better measurement of what has already happened and what has already been disclosed, tracked consistently enough that a pattern becomes visible before it becomes a headline.
We are not against the enthusiasm behind this buildout. The physical constraints we track exist precisely because the demand is real, not because the thesis is wrong. A transformer lead time of 151 weeks is not evidence that AI power demand is overstated. It is evidence that the demand is real enough to outrun the supply chain that has to deliver it, which is a different and more useful thing to know before you write a check.
That is the work. Not a forecast of where infrastructure investing is going, but a constantly updated, source-by-source account of where it actually is, so that whoever is making the decision is making it against what has been delivered, not what has been announced.
The Stratum Atlas terminal is where that account lives, updated monthly, corrected in public, and sourced to something you can go check yourself.