A new database tracking over 400 local instruments across 17 states finds that the density of data center restrictions and the bite of those restrictions are not the same number — and conflating them is where most regulatory analysis goes wrong.
On episode 32 of the SemiAnalysis podcast — titled “300 Datacenter Bans, 6 Projects Delayed: Moratoriums Explained” — analyst Maya Barkin, who covers data centers, energy, and industrials at the firm, described a database she built tracking over 400 local instruments across 17 states. Everything that follows is SemiAnalysis’s own research as described on a podcast — unaudited, and not independently verified here. The database follows each instrument through its full lifecycle: proposed, enacted, rejected, expired, or lifted. Crucially, on her account, permanent zoning bans were deliberately included alongside time-limited moratoriums, on the reasoning that they constrain development identically — they simply never expire. That methodological choice is worth holding: it makes the restriction count larger, not smaller.
Barkin then overlaid that database against SemiAnalysis’s US project pipeline — which she hedged in her own words as “I think 3,500 projects.” That hedge tends to fall off in retelling, so it is worth preserving here. The overlay is the analytical step that separates a count of rules from a measurement of exposure: a restriction only binds where somebody actually wanted to build. The title figure of six projects delayed is the episode’s framing; the underlying explanation is not available in the material reviewed here, and no argument below rests on it.
Two states anchor the analysis and point in opposite directions. Michigan, on Barkin’s account, has 45 moratoriums currently enacted — most adopted late last year, following a governor-signed data center tax exemption she describes as triggering a rush of proposals across the state. Yet overlaid against the pipeline, she finds no pipeline exposure: the largest projects kept moving. New York presents the inverse: a single July measure pausing state environmental permits for new data centers of 50 megawatts or more until July 2027 touches about 1.4 gigawatts of planned capacity and, on her estimates, delays roughly 800 megawatts by six to ten months. Three words that get used interchangeably are doing different work there. About 1.4 gigawatts is touched; roughly 800 megawatts is delayed, on her estimate; and nothing is cancelled. The affected capacity was scheduled for 2028 or later in any case, so a six-to-ten-month slip lands well inside its own horizon.
The key insight: Counting restrictions is public work that anyone with patience can do. Measuring exposure requires a project pipeline dataset that almost nobody has. The cheap number circulates; the expensive one does not. That gap is where most regulatory commentary lives — and it consistently flatters whichever conclusion the headline prefers.
One state, one instrument, and a real number. Most of the other 400 touch nothing.
The Structural Read
This is a clean application of what the Permission Layer framework identifies as the gap between regulatory surface area and regulatory bite. Surface area — the count of instruments — is observable and quotable. Bite requires a second dataset: where the activity actually is. A ban in a county with nothing in the pipeline constrains precisely nothing, however forcefully worded, and it still counts as one in any tally.
What makes the SemiAnalysis approach analytically distinct is that it performs the overlay explicitly, and the two states it surfaces produce opposite answers from the same method. Michigan: 45 instruments, near-zero exposure. New York: one instrument, roughly 800 megawatts estimated delayed. That asymmetry is not a flaw in the method — it is the method working. A measurement that can return either answer depending on the underlying data is a measurement. One that always returns the same answer regardless of the data is a prior wearing a lab coat.
There is a further precision point that matters more than it first appears. Three words in the New York figures are not synonyms: touched, delayed, and cancelled. About 1.4 gigawatts is touched by the measure. Roughly 800 megawatts is Barkin’s estimate of what is actually delayed. And delayed means six to ten months — not cancelled — against capacity that was scheduled for 2028 or later in any case. A headline will reach for the widest of the three because it is the largest available number. The qualification is available in the source to anyone who opens it.
Permission Layer — Structural Pattern
The Regulation Surface-Area Fallacy
Regulatory commentary systematically overstates impact when it reports the count of restrictions without overlaying a map of where the activity is. The bias runs in both directions: a high count with low exposure reads as a crisis that isn’t; a low count with concentrated exposure reads as a non-event that isn’t. The corrective is always the second dataset — and access to the second dataset is the actual competitive moat, not the analysis itself.
SemiAnalysis — Maya Barkin, Podcast Ep. 32 (paraphrased from description)
“Yet we see no zero, we see no pipeline exposure. So the largest projects keep moving forward.” — Barkin’s account of Michigan, where 45 enacted moratoriums show near-zero overlap with the active project pipeline.
Three Implications
IMPLICATION 1 — The Pipeline Dataset Is the Moat
The restriction count is public and replicable. The exposure calculation requires a proprietary project pipeline — in this case, SemiAnalysis’s own ~3,500-project US database. That asymmetry means the analytical value is almost entirely in the second dataset, not the first. Organizations without that pipeline can count rules; they cannot measure whether the rules bind. This is a durable information advantage, not a one-time scoop.
IMPLICATION 2 — Touched / Delayed / Cancelled Is a Discipline, Not a Detail
The New York figures illustrate a precision problem that recurs across infrastructure and technology reporting: the most quotable version of a number is reliably the least qualified one. ~1.4 GW touched is a larger number than ~800 MW estimated delayed, which is a larger number than zero cancellations. Collapsing these into a single headline figure — whichever direction — misrepresents the underlying estimate. The qualification is not a footnote; it is the claim.
IMPLICATION 3 — The Michigan / New York Pair Resists Simple Narrative
The temptation is to read 45 instruments and near-zero exposure as evidence that moratoriums are noise. The New York data refuses that conclusion: one instrument, one gigawatt touched. The two cases together demonstrate that the answer depends entirely on where you are in the pipeline map — which means the structural question for any restriction is always local, not aggregate. Both the “moratoriums don’t matter” and the “moratoriums are a crisis” readings are available from selective citation of this same dataset.
Business Engineer Framework
The Permission Layer
The Permission Layer framework maps how regulatory and policy constraints shape which infrastructure gets built, where, and when — and why the surface area of regulation consistently overstates its bite until someone overlays it against a real activity map. The Michigan-New York contrast is a textbook illustration: same method, opposite conclusions, because the pipeline data underneath is different. Understanding where the layer is thin and where it actually binds is the analytical step most commentary skips.
SemiAnalysis’s moratorium database is a useful reminder that regulatory analysis has two steps, and most commentary only completes the first: counting instruments is not measuring exposure, and the distance between those two operations is exactly where confident-sounding conclusions go wrong in both directions. One state with 45 restrictions and near-zero pipeline overlap; one state with a single measure and roughly a gigawatt touched — both findings come from the same method, applied to different underlying realities. The number that circulates will be whichever is largest and least qualified; the number that matters is whichever one someone has actually overlaid against where the activity is. Nothing here is investment advice, and nothing here predicts how any of this develops; the figures are one firm’s research, reported as such.
Every figure above is SemiAnalysis’s own research as described by Maya Barkin on the firm’s podcast. It is unaudited and not independently verified here. Her own hedge on the pipeline size — “I think 3,500 projects” — is carried above. The “6 projects delayed” figure comes from the episode title. The underlying explanation is not in the material available here, so it is cited as the episode’s framing and no argument above rests on it. Nothing above concludes that moratoriums do not matter — New York is her own counter-example and carries equal weight. Touched, delayed and cancelled are three different things. About 1.4 gigawatts is described as touched; roughly 800 megawatts is an estimated delay of six to ten months, not a cancellation; and the affected capacity was scheduled for 2028 or later in any case. The account of a Michigan tax exemption triggering a rush of proposals is hers. No position is taken above on datacentres, moratoriums, tax exemptions or any state’s policy, and no community’s or government’s motives are characterised. Specific projects, counties and companies, the pipeline methodology, how exposure is defined, figures for other states, total US capacity, grid detail and any developer, operator or government comment are not established and do not appear above. Nothing above is investment advice or predicts anything.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
