DEMAND-SCAN ORIGINATION · EDGE & PROXIMITY COMPUTE

Every edge data-center deal starts with the same question. Where should we even be looking?

A crop from a real Proximity Compute Scout Market Scan: all 48 scored markets drawn as their own counties on the Texas map and shaded by index, the top five labelled by rank, over the report's own caption saying the 206 unshaded counties were NOT scanned and that every shade is a ceiling while friction is unassessed.
Illustrative, dated. Every figure is sourced. One crop from the sample Market Scan — every scored market ranked on the map, above the report's own line about the counties this scan did not cover. The whole report also shows the evidence it did not rank on: a market whose momentum evidence did not clear the floor is drawn as a dashed spoke marked MEASURED — we looked, on a date the page prints, and what we found was not enough — and a market no run has covered is marked ABSENT. Neither is ever plotted at zero, because a triangle drawn to zero would be a claim about the market rather than about the scan. Read the whole report →
There is a reliability contract, with a refund attached. It is written out in full further down this page, in the same words the report itself carries. Read the contract →
WHAT IT SAVES YOU

Don't fly a county that will say no.

We rank where demand, grid/fiber, and entitlement evidence actually coexist — and we show the work. A site search burns weeks of somebody senior's time before it returns a single answer, and most of those weeks go to markets that were never going to work. This is the document that decides which week is worth taking.

Screen, don't decide. It is a ranking with its evidence attached, not a recommendation: you can argue with every number in it, which is the only kind of ranking worth taking to an investment committee.

Two scored legs, one visible friction dial
Demand and supply gravity are combined as a geometric mean, then multiplied by a friction factor that can only bring the score down. One weak leg drags the whole number — by construction, not by judgement.
Every figure carries its source
Dataset, vintage, pull timestamp, grain, confidence tier, record reference, and two ways to check it yourself: a path you can click and the exact programmatic query.
An absent source is a named gap
Never a zero, never a guess, never silence. The report says which dataset is missing, what that costs the answer, and which phone call would close it.
THE TWO REPORTS

Two deliverables. Both graphics-led. Both end with who to call on Monday morning.

One finds the market. The other finds the ground inside it. You can read both in full, further down this page, before you speak to anyone.

PCS · MARKET SCAN

Where to look

For teams choosing among regions, states, or metros
  • Ranked markets with the index and every leg sub-score shown — never a black-box number
  • One-paragraph "why" per finalist, every figure cited inline
  • Friction watch-list with dated citations — what the county has actually done, and what it has only been reported to have done
  • Sensitivity envelope — how far the weights must move before the #1 changes
  • Monday Morning Actions — who to call and what to ask, derived from the findings
PCS · SCOUT SHORTLIST

Where to stand

For teams ready to evaluate ground in a chosen market
  • Everything in the Market Scan, plus a parcel-level screen against your stated gates
  • Analyst-verified candidate parcels, one card each: acreage from the county roll, slope, floodplain zone, nearest transmission line and its voltage, fibre availability at the grain the data actually has
  • Serving utility named per county from federal filings — and the report states plainly that the utility's load letter remains the definitive step
  • The questions no dataset answers, written out as calls you can make, with who holds each answer
  • Your constraints taken at order time as must-haves vs. preferences — and a constraint no public dataset can evaluate is offered as a preference only, never as a gate we would pretend to check
Mid-market edge developersTowercos entering edgeRegional colo operatorsMEC & carrier partnersIndustrial & healthcare compute buyers
READ ONE BEFORE YOU BUY ONE

Both reports, in full, redacted only where a name or an internal code appeared.

Not a teaser, not an excerpt, not a template with the numbers taken out. These are real renders from our committed data snapshots. Every figure, badge, source, limit and freshness date is the genuine output. Internal decision references and staff attributions were generalised, and every page is stamped as a sample.

SAMPLE · MARKET SCAN

A ranked market scan

Ranked markets with their leg sub-scores and confidence badges, the friction watch-list with its dated citations, the sensitivity envelope, and the provenance appendix every figure points into.

Illustrative, and dated. It is a real scan of one region on the date stamped in its footer — not a current answer about your market, and not advice about any market.

SAMPLE · SCOUT SHORTLIST

A shortlist, down to the parcel

A parcel-led shortlist inside Harris County — the market our published scan ranked #1. It leads with the parcel cards: acreage from the county appraisal roll, a slope sample, the mapped floodplain zone, transmission distance and voltage, fibre at the grain the data has — each with the instrument that measured it and the limit that applies to it. Then the named gaps, routed to phone calls. The scan says which market; this says which parcels.

Illustrative, and dated. The parcels are real and public, proposed by our own analysts to exercise the screen; the page says so. Nothing here says a parcel is buildable.

TWO PAGES · IC BRIEF

The two pages you forward to whoever signs off

The problem, the method in twelve lines, both deliverables, and the reliability contract in the report's own words. No email, no form, nothing to sign up for.

WHAT WE WON'T DO

The honest gaps, stated up front. The Aruon way.

NO GUESSED QUEUE POSITIONS

Interconnection queue position is unpublished nationwide. We report queue activity where public — never a project's own position.

NO PRETEND PRECISION

Friction signals are news- and legislation-based by nature. They're tiered, dated, and linked — and the report says so plainly.

NO PURCHASE ORDERS

Demand is a proxy stack, not a signed customer. The scan finds where the signals converge; your BD team still sells the capacity. Screen, don't decide.

NO SUBSTITUTE FOR THE LOAD LETTER

We name the serving utility from its own federal filing and hand you the call list. The utility's load letter remains the definitive step — exactly as our reports state.

WHO THIS IS FOR

Three buyers, three different first questions. One instrument.

TOWERCO & CARRIER
  • You already hold sites. The question is which ring is worth a build, not where the map is empty.
  • We rank which rings have demand and a county whose rules we scored — friction is measured and dated, never an entitlement we promise.
  • The latency ring is chosen at order time, so the radius you get is the one your workload needs.
  • You leave with ranked markets, every figure sourced, and the calls no dataset answers.
REGIONAL COLO & EDGE DEVELOPER
  • You are choosing between metros before anyone will fund the study.
  • Demand is measured at county grain against comparable counties nationwide, so a weak market cannot look strong in weak company.
  • Supply gravity says whether the grid and the network can serve it; the gap ratio says whether the white space is real.
  • You leave with a sensitivity envelope: how far the weights must move before the #1 changes.
INDUSTRIAL & HEALTHCARE SELF-SITER
  • You are siting your own compute beside your own operations, not selling capacity.
  • Your constraints are taken as hard gates and preferences, and a constraint no public dataset can evaluate is offered as a preference only.
  • Where nothing public answers, you get the named gap and the phone call — never a filled-in guess.
  • You leave with parcels inside the market you already care about, acreage read from the county's own roll.
HOW IT WORKS

Independent evidence, combined so the weakest part still counts. No black boxes.

Every dataset feeds exactly one leg. If two legs shared a source, "convergence" would be the same fact counted twice — so leg independence is checked on every scoring run, and a violation blocks release. Not advisory. Blocking.

SCORED · DEMAND

Who needs compute here

How much latency-sensitive activity actually sits in the ring: federal jobs data at block grain, industry mix, population and growth, named anchor institutions. Percentiles are taken against comparable counties nationwide, never against the handful in your request — so a weak market cannot look strong by being asked about in weak company.

SCORED · SUPPLY GRAVITY

Can the grid and network serve it

Interconnection and peering presence, fibre depth, transmission headroom and voltage, serving utilities named from their own federal filings — and the gap ratio: demand per unit of capacity already there. White space is the product.

SHOWN, NOT RANKED · MOMENTUM

Is it growing right now

Dated, quoted, linked evidence of expansion. Every claim is re-fetched from the publisher's own page and the quoted sentence confirmed to be on it; what fails is discarded and counted. A market clears the floor at three qualifying signals inside eighteen months and its reading shows as measured; below that it is named as thin. The page prints the date we looked on either way.

⚠ Momentum does not move your rank, and that is deliberate. We have researched momentum in a small number of markets and not in the rest. Ranking on it would order the report by where we have looked rather than by where the market is moving — the markets we paid to research would be marked down for having been researched. So every market is ranked on the two legs every market has, demand and supply gravity, and momentum is reported beside the ranking rather than inside it. When coverage is broad enough that being researched is no longer the deciding variable, that changes — and it changes in public, with a version on it.

× FRICTION — DOES THE COUNTY SAY YES

Moratoria, blocked projects, incentive posture — a multiplier on the whole score, shown as its own dial. A county under moratorium can't rank high no matter how loud the demand, because that's how a developer experiences reality. Friction facts never hide inside an evidence leg, they are sized against your build (a 50 MW threshold does not bind a 2 MW room, and the report shows the number rather than deciding for you), and they don't expire until repealed. A jurisdiction whose act cannot reach your market is still listed — under an explicit verdict saying it does not reach you. Disclosed, never deleted.

Index = GeoMean(the legs that are present) × Friction   // geometric mean: 70·70 beats 90·20 by construction
Every market is ranked on the same two legs — demand and supply gravity — whatever its momentum evidence says, so no market is moved up or down for having been researched. Momentum is shown beside the ranking: a market that clears the floor — three dated, linked, quote-verified signals inside eighteen months — reads as measured, and one below the floor is named as thin. It is never counted as zero. On the radar a leg with no qualifying reading is a dashed spoke: marked MEASURED where a completed run came back below the floor, and ABSENT where no run has covered the market at all.
THE WORDS THE REPORT USES

A short glossary, so page one needs no translator.

PCDI

The Proximity Compute Demand Index — the single ranking number. A geometric mean of the scored legs, multiplied by friction. Geometric, so a market that is excellent at one thing and poor at another cannot average its way to the top.

THE DASHED SPOKE

On each market's radar, a leg we did not score is drawn as a dashed line labelled ABSENT, and the shape stops short of it. Drawing it at zero would claim we looked and found nothing. We didn't look — and those are different documents.

CONFIDENCE TIERS

Every figure is badged with how we know it. MEASURED is read from a named dataset. Weaker tiers are stated as such, and the report's overall confidence is the lowest tier of the legs it used — it cannot be talked up by the strong parts.

GAP RATIO

Demand divided by the serving capacity already measured in the ring. Above 1.0 is white space; below it, the market is already served.

SCREENED, NOT SCORED

A required leg was absent, so the market is listed separately with its named gap and left unranked. Scoring it zero would rank it — and rank it last — which is a claim we did not measure.

NEAR TIE

Two markets within a couple of index points, or which swap places inside the sensitivity envelope. Read them as a pair, not as a ranking. The report names them rather than letting the sort order imply a winner.

FRICTION NOT ASSESSED

No friction research has been run for that market, so its index is an upper bound. Friction can only bring a number down, so "at most X" is exactly true and "X" would not be.

NAMED GAP

A question we asked and no dataset we hold could answer — by-right zoning status, distance to the nearest substation, lit fibre at a specific address. It is printed with the call that would close it and who holds the answer, never quietly dropped.

THE ARUON RELIABILITY CONTRACT · THE ERROR BOUNTY

What we promise on every report. In writing. With a refund attached.

Most market intelligence asks for your trust. Ours ships with the tools to check it — because the instrument that finds your next market should be held to the same standard as the interconnection study that follows it.

Every number is reconstructible.

Dataset, exact vintage, pull date, how-to-verify — per figure. Find one you can't reconstruct and the scan is free. A paid order is refunded in full, by us, to the card that paid.

Every claim is dated and linked.

Momentum evidence comes from dated sources with links verified to render — our own checker re-fetches every page and finds the quoted sentence on it. Failures are discarded and counted.

We score only what we measured.

Thin public momentum is labeled exactly that — badged, capped, never dressed up. An absent source is a named gap, not a guess.

The math is published.

A geometric mean of independently-sourced legs times a visible friction dial. The Gap Ratio formula is versioned and printed in your appendix.

The ranking admits its wobble.

Every report states how far the weights must move to change the #1. Near-ties are called near-ties, not verdicts.

A person signs every release.

Nothing auto-sends. An analyst reviews each report, approves it personally, and the release is recorded in a ledger.

Our failures are ours.

If a source goes down, your report says which one and when — and the credit comes back. It never renders as a fact about your market.

The instrument has a health record.

Leg-independence checks on every scoring run — a violation blocks release, it does not warn — plus a data-freshness board and a calibration harness. Inspectable by founding clients on request.

HONEST IS THE FLOOR. PROUD IS THE BAR.
NOT A COMPUTE BRIEF?

The same discipline on one parcel, one ZIP, or one tower.

Property Screen, Market Report, Site Hunt, Tower Sheet and Tower Target List are ordered on the ARUON report page. Every one is analyst-released; the order page checks whether we hold your county before you pay.

Open the report page →

REQUEST A SCAN

Tell us what you deploy and where you're looking. A person reads it.

This creates an order request — it is not a checkout. There is no card field on this page and no payment processor behind it. ARUON confirms scope and invoices separately, and work begins on your written go-ahead, not before.

Loading the current options and pricing from our order system…

Or go straight to the order page at reir.aruon.ai/pcs →

1 · CHOOSE YOUR DELIVERABLE
2 · IN ONE SENTENCE

I'm at , we're looking at , and you can reply to .

Plain words are fine for the region — "within a day's drive of Dallas" is an answer. A person reads this before anything is quoted.
3 · DESCRIBE THE DEPLOYMENT
State names or two-letter codes — either reads.
The area you described above is enough — we'll resolve the counties with you. One county is enough for a shortlist.
Comma separated. One is enough.
4 · YOUR CONSTRAINTS

Mark what is a hard gate and what is merely preferred. The report tells you which gate killed what.

5 · SEND IT

Request received.

WHAT HAPPENS NEXT