AmpDesk

1. One-liner

Tells an industrial broker which buildings on the tour list can actually power the tenant, before the LOI.

2. Trend signal — why now?

Power stopped being a utility bill and became the thing that decides whether an industrial deal happens at all.

The blunt version, from a broker writing about his own market in 2026: electrical capacity is “one of the most overlooked—and underestimated—factors in industrial real estate. But it’s also one of the most common deal killers.” A capital-projects consultant writing about LA industrial puts a clock on it — deals “die in due diligence. Usually around week six.”

Three numbers explain why week six is where it dies:

The upgrade timeline outruns the lease. LADWP’s typical timeline for an industrial service upgrade “has been running between 12 and 24 months.” In Southern California, if a substation upgrade is required, “timelines typically exceed 24 to 36 months.” A tenant who needs to be operating in nine months has just discovered, in week six, that the building they picked is a two-year project.

The cost is not a rounding error. Customer-side electrical work on these upgrades runs “$50K-$500K+.” Nobody budgets that as a line item on a lease they thought was turnkey.

Demand per square foot moved. Fully automated facilities “use 3−5x more power than the 2024 vintage baseline.” The building stock did not move with it — if a building “was built in the ’80s or ’90s for warehousing, your electrical infrastructure may no longer be a fit for these users without upgrades.” Cold storage, EV shops, robotics integrators, and CNC job shops are all touring 1990s warehouses wired for pallet racking and a forklift charger.

And the market has already repriced around it. CoStar’s 2026 read is that power-ready assets lease fast while “power-constrained assets face prolonged marketing periods” — occupiers now prioritise “power availability, automation-ready specifications, and skilled labor access over discounted rents.” Power is the amenity that beats price.

The tell that this is a real, priced problem rather than a talking point: Southern California Edison sells the answer for $3,000. Its Engineering Analysis Report “includes the site’s power availability and the time required to build new infrastructure, if needed.” One report. One site. One utility’s territory. A broker shortlisting eight buildings across two utilities cannot use that — the arithmetic is $24,000 and several weeks to screen a tour list, which is why nobody does it and everybody finds out in week six.

Meanwhile the raw data quietly went public. As of the DOE’s atlas, 58 utilities and state agencies have published hosting-capacity maps across 26 states, D.C. and Puerto Rico, many at individual distribution-circuit level. The grid told everyone where the capacity is. Nobody built the tool that reads it for a guy with eight addresses and a tenant on the phone.

Provenance:

• Signal 1 (demand): Electrical capacity is “one of the most overlooked—and underestimated—factors in industrial real estate. But it’s also one of the most common deal killers”; “Most people aren’t trained to recognize electrical limitations” — https://www.kennylaurealestate.com/blog/why-electrical-capacity-can-make-or-break-your-industrial-deal — observed 2026-08-23
• Signal 2 (feasibility): 58 utilities and state agencies publish public hosting-capacity maps across 26 states, D.C. and Puerto Rico, with circuit-level data available in the Northeast; SCE additionally runs a Power Site Search Tool showing capacity by address — https://www.energy.gov/cmei/vehicles/us-atlas-electric-distribution-system-hosting-capacity-maps — observed 2026-08-23
• Signal 3 (economic): SCE sells a $3,000 per-site Engineering Analysis Report covering power availability and build timeline; SoCal substation upgrades exceed 24–36 months; automated facilities draw 3–5× the 2024 baseline — https://voitco.com/access-to-power-the-new-priority-for-industrial-site-selection/ — observed 2026-08-23
• Supporting (cost/clock): LADWP industrial service upgrades run 12–24 months at “$50K-$500K+” customer-side; LA industrial deals “die in due diligence. Usually around week six” — https://emc4la.com/la-industrial-utility-constraints.html — observed 2026-08-23
• Supporting (market): Power-ready assets lease rapidly while power-constrained assets face prolonged marketing; occupiers prioritise power availability over discounted rents — https://www.costargroup.com/press-room/2026/costar-expects-us-industrial-vacancy-peak-year-rent-growth-remains-unchanged — observed 2026-08-23 Category: Underserved niche

3. The opportunity

Every tool that answers “is there power here?” is built for someone spending $200 million.

RadiusMapper maps substations for data centres, solar farms, BESS and crypto loads. Envelio sells an Intelligent Grid Platform to utilities. SCE’s PSST is a portal for “industrial developers.” The DOE atlas is a research artifact. All of them think in megawatts and substations, because their customer is siting a campus.

The tenant-rep broker thinks in amps at a service panel on a specific address he is touring Thursday. Nobody serves him. He does what the trade press tells him to do: “rely entirely on a contractor or electrician to confirm if a building can handle a tenant’s needs.” That means a favour-call to an electrician friend, per building, unpaid, slow, and only after the tenant has already fallen in love with the space.

This is the invert-the-customer gap in its cleanest form. A whole industry — utilities, grid-platform vendors, site-selection consultancies, EBI-class due-diligence firms — sells power intelligence to the party with a capital budget. The party who eats the loss when the answer is “no” is the broker and his 40-person manufacturing tenant, and they get nothing.

What the incumbents do badly, named plainly:

• SCE’s $3,000 EAR is priced per site and scoped to one utility. It is a closing document, not a screening tool. You buy it once you have already committed to a building — which is exactly too late to change your mind cheaply.
• Hosting-capacity maps are published per utility, in incompatible formats, on the utility’s own terms, and are built to answer a solar developer’s question (can I inject power here?) rather than a tenant’s (can I draw 1,200A here?). Reading them requires knowing which utility, which circuit, which map version, and how that utility words things.
• Site-selection consultants engage at $25K+ and think in states and counties. They are not screening eight addresses inside a two-week tour window.

The wedge is not new data. It is collapsing a $3,000, multi-week, one-site question into a $0-marginal-cost, same-day, eight-site answer — and doing it early enough that the answer can still change the tour list.

4. Target market

Primary customer: Industrial tenant-rep brokers and small industrial brokerage teams in the US — 2 to 25 agents, doing 20–150 industrial lease deals a year, in power-constrained metros (LA/Inland Empire, Phoenix, Dallas, Atlanta, Northern NJ, Chicago, Columbus, Nashville). Secondary buyer: the tenant directly — owner-operators of 20–200-person manufacturers, cold-storage operators, EV/battery service shops, robotics integrators and CNC job shops taking 20,000–150,000 sq ft.

Why they buy, in their words: The broker’s problem is that he cannot see the constraint he is most likely to lose the deal to. “Most people aren’t trained to recognize electrical limitations.” The tenant’s problem is the assumption the article calls out directly — the tenant looks at a panel and says “Great! We can just swap the panel and get more power,” not knowing that behind the panel is a utility queue measured in years. Both discover the truth in week six, when the tenant walks and the broker restarts a search he has already spent six weeks on.

Rough TAM reasoning: SIOR alone has 3,000+ members representing “roughly the top 5% of all licensed commercial real estate agents worldwide” — implying an order of magnitude more industrial-active agents below that tier. Conservatively, 15,000–30,000 US brokers touch industrial leasing regularly. At a 3–6% seat penetration and $150–400/mo, that is a $10M–$40M ceiling on the broker side alone before counting tenants, landlords repositioning vacant stock, and the industrial GC/electrical-contractor channel who want the same screen for a different reason.

Why now for them: Vacancy is rising (CoStar: 7.5% → 7.8% through 2026) while power-ready space stays tight. That is the worst possible combination for a broker — more buildings to show, fewer that actually work, and a tenant who now ranks power above rent. The screening problem got bigger and the reward for solving it got bigger at the same time.

5. Product sketch (MVP)

• Paste a tour list, get a power read. Drop in 3–12 addresses. Each comes back with a capacity verdict — Likely OK / Tight / Upgrade Territory / Utility Queue — against the tenant’s stated load.
• Load profile from plain English. The broker types “40-person CNC shop, twelve machines, two compressors, no cold storage” and gets an estimated service requirement in amps and kVA, with the assumptions shown so an electrician can argue with them.
• Utility identification and rulebook. Resolves each address to its serving utility, pulls that utility’s published hosting-capacity/circuit data where it exists, and surfaces that utility’s actual service-upgrade process, current published timelines, and how to request a will-serve letter.
• Days-to-power estimate, not just yes/no. The output the tenant actually needs: if this building needs an upgrade, roughly how many months, and does that fit the occupancy date.
• Existing-service inference from listing evidence. Reads the listing, brochure, and photos of the panel/switchgear the broker uploads to estimate what service the building has today.
• The week-one memo. A one-page, client-ready PDF per building — capacity read, risk grade, what to verify, and which questions to put to the utility in writing — so the broker looks like the only person in the deal who thought about this.
• Will-serve request drafting. Pre-fills the serving utility’s own service-availability/will-serve request for the shortlisted building, so the 2–6 week clock starts in week one instead of week six.
• Portfolio watch for landlords. Grade every building in a landlord’s vacant inventory so they know which ones to market as power-ready and which need capital before they will lease.

6. AI angle — what’s load-bearing

Remove the AI and this is a link to 58 incompatible utility websites. Nobody pays for that.

Three places the model does actual work:

Reading utility documents nobody has normalised. Each utility publishes its interconnection rules, service-upgrade process, line-extension policy and capacity map differently — different terms, different formats, different definitions of the same thing. Turning “this utility’s electric service requirements PDF” into a structured answer about timeline and process, for hundreds of utilities, is a document-understanding problem that used to require a consultant per territory.

Turning a tenant’s operation into an electrical load. “Twelve CNC machines, two 50HP compressors, a paint booth, forty people” → an estimated service size with a stated confidence band. This is the judgment an electrical engineer applies from experience, and it is the step that makes the product usable by someone who is not an engineer. It is also why the output must show its assumptions — the product’s job is to get the broker to the right question, not to stamp a design.

Extracting existing service from unstructured evidence. Listing text, brochure specs, and a phone photo of the switchgear nameplate. Vision plus domain priors, per building, in seconds.

If the product only did lookup, an intern could do it. The reason a broker will run it on eight buildings instead of one is that it costs him nothing per building — and that only holds if the judgment is automated.

7. Localization angle

N/A — this is a US-first play, and deliberately so. The wedge depends on two US-specific facts: the public hosting-capacity map disclosure regime driven by state PUCs (58 utilities, 26 states), and a fragmented investor-owned/municipal/co-op utility landscape that makes per-territory knowledge genuinely hard to assemble. In a market with one national grid operator this is a single web page, not a product. The natural second market is the UK, where DNO connection queues create the identical tenant-side problem, but it is a sequel, not a launch.

8. Business model — path to $1M–$5M ARR

Pricing:

• Broker seat — $149/mo (or $1,490/yr): unlimited screens, 20 client memos/mo.
• Team — $499/mo: 5 seats, shared deal history, landlord portfolio grading.
• Tenant/one-off — $299 per shortlist (up to 6 addresses): the direct-to-tenant impulse buy, and the funnel that recruits brokers.
• Landlord portfolio — $1,200/yr per 25 buildings: grade vacant stock for power-readiness.

The anchor is easy to hold. SCE charges $3,000 for one site. A $149/mo seat that screens every building a broker tours all year is a rounding error against a single lost deal, let alone a single EAR.

ACV: ~$2,200 blended (mix of $1,490 solo seats, $6,000 team accounts, and one-off tenant screens).

Rough math to $1M ARR: 455 accounts at $2,200 blended. Realistically: 350 broker seats ($521K) + 60 team accounts ($360K) + 100 landlord portfolios ($144K) = $1.02M. That is a low-hundreds customer count in an addressable pool of 15,000+ brokers.

Rough math to $5M ARR: needs two things to be true. First, coverage expands past the 26 hosting-capacity states into a workable national answer (including utilities that publish nothing, where the product’s value shifts from data lookup to process-and-timeline intelligence). Second, the product moves up the stack from screening into the transaction — will-serve request tracking, upgrade-cost estimating, and a paid handoff to electrical contractors. At that point team accounts carry the number: 600 teams × $6,000 + 1,200 seats × $1,490 = $5.4M.

Expansion path: seats within a brokerage → landlord portfolio grading (a different budget in the same building) → contractor/engineer referral revenue → data licensing of aggregate power-readiness by submarket, which is the thing CoStar-tier subscribers would actually pay for.

9. Go-to-market wedge — first 100 customers

1. The dead-deal autopsy, at scale. Public industrial listings that have sat on market 180+ days in constrained metros are enumerable from listing sites. Grade 500 of them for power-readiness, unasked. Send each listing broker a one-page memo: “Your building at X is on a circuit that reads constrained; here’s what a tenant’s engineer will find in week six, and here’s the sentence to put in your marketing if I’m wrong.” Half will be annoyed. The ones who reply have a building they cannot lease and now know why. Target 500 sends → 8% reply → 40 conversations → 15 paid.

2. SIOR chapter meetings, in person, with a live demo. SIOR has city-level chapters (LA, Northern California, KC, Inland Empire/OC, and dozens more) that meet regularly and take speakers. This is a room of exactly the buyer, assembled by someone else, for free. The demo is the whole pitch: take addresses from the audience, screen them live, show the one that fails. Six chapters × ~40 attendees × 10% = 24 paid seats, plus the referral tail that follows a room seeing its own buildings graded.

3. Ride the week-six failure while it’s hot. Utility service-upgrade applications and will-serve requests generate public trails in many jurisdictions, and permit data for industrial electrical upgrades is scrapeable in most large metros. Every one of those is a deal that discovered its power problem late. Contact the tenant’s broker of record: “You just found out the hard way. Next tour, find out in week one.” Small list, brutal relevance, high close.

4. Electrical contractors as a referral channel. Industrial electrical contractors get the favour-call today and hate it — it is unpaid pre-sales work. Give them a free co-branded tier: they run screens for their broker relationships, the output carries their logo, and qualified upgrade jobs route back to them. They become an unpaid salesforce with pre-existing broker trust. 30 contractors × 4 broker relationships each is a 120-broker pipeline that costs nothing.

5. One public artifact that does the arguing. A free “power-readiness by submarket” index for three constrained metros, updated quarterly. Not content marketing — a citable number that industrial brokers and local business press will pass around, with per-building answers behind the signup.

10. Build complexity — justification

Medium. The hard work is coverage, not cleverness. The AI pieces are off-the-shelf: document understanding on utility PDFs, vision on switchgear photos, structured extraction from listings. The stack is a standard web app with a geospatial join.

The real build is a utility-by-utility ingestion pipeline — 58 published hosting-capacity maps in incompatible formats, plus service-territory boundary resolution so an address maps to the right utility, plus per-utility process and timeline knowledge for the territories that publish nothing. That is grinding, unglamorous, and exactly why nobody has done it.

Sequence it: one metro (LA/Inland Empire — worst constraint, richest public data, SCE and LADWP as the two territories, and an active SIOR chapter to sell into) shipped in 10–12 weeks by two people. Then a new metro every 4–6 weeks. National coverage is a year-two problem and should not block revenue.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Public utility data, public listings. Output is a screening estimate, explicitly not a stamped engineering design — the disclaimer is load-bearing and must be honest.
Ethical — no harm / dark patterns	✅	Product tells buyers a truth sellers would rather delay. Risk is overconfidence, not deception; mitigated by showing confidence bands and assumptions.
Market exists (evidence above)	✅	SCE sells the one-site version at $3,000. Priced pain, named incumbent, quantified timelines.
1–5 person team can build this	✅	Two people, one metro, 10–12 weeks. Data grind, not research.
Launchable with <$50K / ₹40L	✅	Public data sources, standard APIs, no licensing. Main cost is founder time plus travel to SIOR chapters.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Deals die in week six over this. 12–36 month upgrade clocks, $50K–$500K+ costs, and a named “most common deal killer.” Falls short of 19–20 only because the broker feels it a few times a year, not daily.
Demand evidence	15	12/15	Strong: a utility literally sells this answer at $3,000/site, CoStar confirms the market has repriced around power-readiness, and trade press names the diligence gap. Held below 13 because I have industry-press statements of the pain, not brokers’ own verbatim complaints in a forum — that is the first thing to go validate.
Build feasibility	15	11/15	Off-the-shelf AI, standard stack, but 58 incompatible data sources and service-territory resolution make this a 10–12 week two-person build per the first metro, not a 4-week solo sprint.
Distribution clarity	15	12/15	SIOR chapters are a named, assembled, free room of exactly the buyer. Dead-listing autopsy is enumerable and specific. Contractor channel is real. Docked for unproven conversion in a relationship-driven industry that distrusts software from strangers.
Revenue mechanics	15	11/15	Pricing anchors cleanly against a $3,000 competitor artifact and the math to $1M needs only low-hundreds of accounts. Docked because brokers are notoriously bad software buyers — low tool budgets, high churn, and “the brokerage already pays for CoStar” is a real objection.
Time to first revenue	10	8/10	The one-off $299 tenant/broker shortlist can sell before full coverage exists — arguably before the product is automated. Revenue in 4–8 weeks is realistic; a metro of coverage first is the gate.
Defensibility	10	5/10	Honest read: the moat is the ingestion grind and per-utility process knowledge, which compounds but is copyable by a determined competitor in 6–9 months. No data exclusivity. What actually defends is accumulated verified outcomes — “we said upgrade territory, the utility agreed” — which nobody else can backfill.
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · sales-heavy

Needs someone who will grind 58 utility data formats and someone who will stand up in a SIOR chapter meeting and screen a stranger’s building live. The second person is harder to find and matters more. A former industrial broker or utility service-planning engineer as co-founder or first advisor changes the odds materially.

Key assumptions to validate

1. Assumption: Brokers will pay personally (not wait for brokerage IT) for a $149/mo tool. How to test: 40 cold calls to industrial brokers in LA/IE and Phoenix, pitch at price, count cards taken and pre-orders. Anything under 15% “yes at $149” means reprice to per-deal or sell to tenants instead.
2. Assumption: A screening-grade answer built from public data agrees with the utility’s real answer often enough to be trusted. How to test: run 25 buildings where the true outcome is known (via brokers who lived through the upgrade, or SCE EARs already purchased) and measure the hit rate on the four-way verdict. Below 70% agreement, the product is a liability.
3. Assumption: The pain lands on the broker, not only the tenant. How to test: 20 broker interviews — ask specifically how many deals in the last 24 months died or stalled on power, and who ate the cost. If it’s always the tenant, flip the go-to-market to direct-to-tenant and reprice as per-shortlist.
4. Assumption: Hosting-capacity maps are usable for load questions, not just generation interconnection. How to test: two weeks with the SCE, LADWP and Ameren datasets plus a consulting electrical engineer, confirming the load-side read is defensible. This is the single most likely place the idea breaks.

Risk flags

1. Data-fitness risk (the big one): Hosting-capacity maps were built to answer whether a solar array can inject power, not whether a shop can draw 1,200A. If the load-side inference is too weak, the product degrades into “here’s the utility’s process and phone number” — still useful, worth far less. Validate this before writing much code.
2. Liability and tone: A confident wrong answer that kills a good deal, or blesses a bad one, is reputational death in a referral industry. The product must be a screen that routes to verification, never a substitute for a will-serve letter or a stamped design. Under-claiming is the correct posture.
3. Buyer-quality risk: CRE brokers are a famously difficult software market — thin personal tool budgets, CoStar fatigue, and a habit of solving problems with relationships instead of subscriptions. The landlord and contractor channels exist partly as insurance against this.
4. Incumbent adjacency: CoStar could add a power-readiness field to listings. It would be shallow (a self-reported number, not a utility-side read) and slow, but it would compress the marketing story. Speed to a defensible outcome record matters.
5. Coverage cliff: 26 states publish; 24 do not. In non-publishing territories the product’s answer is materially weaker, and a broker in Nashville who gets a thin answer churns and tells his chapter. Better to refuse coverage than ship a bad read.

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder who will grind utility data formats, paired with
                        someone credible in a room of industrial brokers — ideally an
                        ex-broker or an ex-utility service-planning engineer.
Time to revenue:        6–10 weeks (one-off shortlist sales can precede full coverage)
Capital to launch:      $15–25K (founder time, API costs, SIOR chapter travel)
Top 3 assumptions to validate first:
  1. Public hosting-capacity data supports a defensible LOAD-side read — two weeks with
     SCE/LADWP/Ameren datasets and a consulting electrical engineer.
  2. Screening verdicts agree with real utility outcomes ≥70% on 25 known-outcome buildings.
  3. Industrial brokers pay $149/mo personally — 40 cold calls, count pre-orders, need ≥15%.
Kill criteria:
  - Abandon if verdict agreement against known outcomes is below 70% on the 25-building test.
  - Abandon if fewer than 6 of 40 cold-called brokers will pre-order at $149/mo AND
    direct-to-tenant one-off sales fail to clear 10 units in the first 60 days.
  - Abandon if hosting-capacity data proves generation-only in practice and the load read
    collapses to "call your utility" — at that point this is a directory, not a product.

15. Next step — 1-week validation sprint

• Day 1–2 — Kill the data question first. Pull the SCE, LADWP and Ameren hosting-capacity datasets. Sit with a consulting electrical engineer for three hours and answer one question: can this data support a four-way load-side verdict on a specific address, or is it generation-only? This is the assumption most likely to be fatal and it is answerable in two days for the cost of an engineer’s afternoon.
• Day 3–4 — Build the 25-building truth set by hand. Get 25 LA/Inland Empire industrial buildings where the power outcome is known — from brokers who lived the upgrade, or from EARs already purchased. Produce a manual verdict for each using only public data and no automation. Measure agreement.
• Day 5 — Sell it before it exists. 40 cold calls to industrial tenant reps in LA, IE and Phoenix. Pitch the tour-list screen at $149/mo and the one-off shortlist at $299. Take pre-orders, not compliments.

Go/no-go, falsifiable: proceed only if the manual verdicts agree with known outcomes on ≥18 of 25 buildings AND ≥6 of 40 brokers put money down or verbally pre-commit at price. Miss either and the answer is no — the first miss means the data cannot carry the product, the second means the buyer is the tenant, not the broker, and the whole go-to-market needs rebuilding before a line of code gets written.