The Counter — Year One

OP 010Thesis — why this winsSETUP

The gap is not machines.
It's coordination + priority.

10 days

MacroFab's "fast lane" — the leading US digital PCBA platform, headquartered in Houston. Protolabs / Xometry quote 1–3 days for CNC but ship from distant plants. Nobody sells a same-day combined loop.

Hours

Huaqiangbei's actual product is loop speed: components, PCBs, CNC, and assembly physically dense and culturally geared to small fast orders.

The gapHouston has thousands of oil-&-gas job shops. Small orders queue behind Raytheon and Halliburton. A broker who aggregates many small orders into one big priority customer, and runs the courier loop, can sell same/next-day today — with zero factories owned.
Anchor gravityTeraFab ($16.8B, Grimes County, announced 2026-08-06) + the SpaceX supplier network + Texas A&M (30 min) + Houston's machining base (60 min). Being early in this corridor is the land-and-network bet of the decade.
Year one, honestly scopedYear one is not the district. Year one is the brokerage that discovers the vendor graph, proves the speed promise, and funds the mission of building America's hardware market district.

OP 020The offer — SLA tiersRATE CARD

Every request gets a time and a price.

Tier	Promise	Scope	Price posture
Sprint	Same business day in by 9am · out by 6pm	3-axis CNC ≤ 5 parts, simple alu/steel/delrin; laser/waterjet; 3D print; components pulled from stock	3–5× market. Speed buyers don't price-shop.
Overnight	Next business day	Above + 2-layer PCB fab-and-populate (partner line), anodize/passivate via priority slot	2–3× market
72-Hour	3 business days	4–6 layer PCBA, 5-axis, small castings, harnesses	1.5–2× market
Retainer	Standing weekly capacity	Startups / R&D labs pre-buy N sprint-hours per month, get first-in-queue	$2–6k/mo range (validate)

A human answers. Phone, text, walk-in. Quotes in ≤ 30 minutes — AI-drafted, human-confirmed: the auto-quote agent reads the STEP/Gerber, drafts price + routing, a human presses go. This is where the AI stack is the moat.
We never say no; we say when. Every request gets a time and a price.
"5 units is tomorrow's job, not a nuisance." That sentence is the brand — it's what makes this Huaqiangbei and not Xometry.

OP 030Phase plan — 12 monthsROUTING

Four operations, in sequence.

PHASE 0

Vendor graph

Weeks 1–6 · can start remote

DELIVERABLE — a live database of 150+ corridor vendors, with a speed-tested top 25.

Build the target list from ThomasNet + a Google Maps sweep of Houston metro (CNC, sheet metal, anodizing, wire EDM, harnesses), the Houston Area Manufacturers directory, SpaceX/Tesla supplier registries, and A&M engineering vendor lists. This list-build is automatable — scraper + enrichment pipeline, reusing the founder's existing lead-gen stack.
Speed-test with real money: send the same benchmark part set (one bracket, one turned part, one 2-layer board) to 25 shops as a paying customer. Record quote latency, price, promised vs actual, and small-order attitude. This data IS the company. Budget: $3–6k.
Rank into the Fast 25 — shops that quoted < 4h and delivered on promise.

PHASE 1

First loop, boots on ground

Months 2–4

DELIVERABLE — 10 paying customers · ≥ 20 completed jobs · ≥ 80% on-time.

Physical presence required — the courier loop and shop relationships don't work remote. Pick one by end of Phase 0:
- (a) Relocate to the NW Houston corridor (Tomball / Magnolia / Cypress) — full commitment, fastest learning;
- (b) Founding ground partner in Houston (equity, not salary) while quoting/AI/ops run remote — slower trust-building, keeps the founder's current base intact;
- (c) Season on the ground — 8–12 weeks embedded, then hybrid; probably the honest minimum viable version.
Sign priority retainers with 6–10 of the Fast 25: "we buy your first machine-hour every morning, guaranteed monthly minimum, net-7 payment." Net-7 is the unlock — job shops live at net-60 from primes; fast cash buys priority nobody else gets.
PCBA partner: negotiate a priority lane with a small local board house (not MacroFab — they're volume/platform; find the 15-person shop that will run our panels first). Fallback: own a desktop line (~$40–70k used: printer, pick-and-place, reflow) for 2-layer overnight boards.
Courier: contract driver on AM/PM loops along a fixed route past the Fast 25; Roadie/Dropoff as overflow.

PHASE 2

Demand engine

Months 4–8

DELIVERABLE — 40+ active customers · $25–60k/mo gross (assumption to validate, not a promise).

Hardware startups: Greentown Labs Houston (climate-tech hardware), the Ion district, the A&M startup ecosystem — walk the floors, offer one free SPRINT job per team. One wow-loop converts.
SpaceX / TeraFab supplier constellation forming in the corridor — every supplier has fixture, bracket, and test-jig emergencies weekly.
Texas Medical Center device labs — prototype jigs and enclosures, chronically underserved.
Oil & gas R&D groups — they already pay expedite fees; give them a counter to call.
Content flywheel: film the loop ("sketch at 9am, part at 4pm") — build-in-public, the Gundo playbook. This also seeds the district's future culture.

PHASE 3

The Counter becomes a place

Months 8–12

DELIVERABLE — a walk-in storefront + the district thesis proven on paper.

Lease 3–6k sq ft flex-industrial on the 249/290 corridor. Front: counter, component shelves (buy a small broker's surplus stock — the "parts on shelves TODAY" experience no US city has). Back: print farm, one CNC mill, desktop SMT line, staging for courier loops.
Host monthly hardware build nights from day one of the lease.
The stall-tenant marketplace (the true Huaqiangbei model) is year 2–3, funded by brokerage cash flow — with the demand data as the pitch to Grimes County EDC and state incentive programs. Get named in the county's "chip city supplier ecosystem" story early: free land/tax leverage later.

OP 040Unit economicsUNVERIFIED

Assumptions — measure in Phase 0/1.
Do not trust these numbers yet.

Assumptionmeasure, don't trust

Sprint job

Charge $400–1,200; vendor cost 30–50%; courier ~$40–80/loop amortized across jobs; target 35–50% gross margin.

Break-even

Solo + contract driver + software: roughly 25–40 jobs/month at a SPRINT/OVERNIGHT mix. Validate against real Phase-1 costs.

Year-one capital

$60–120k: speed-testing $6k · working capital for net-7 float $20–40k · desktop SMT line $40–70k (optional) · travel / ground time · insurance. Bootstrappable if the SMT line waits.

Insurance / liability

General liability + errors on dispatched work. Jobs are "prototype grade, no cert" by default — aerospace/medical cert work is explicitly out of scope year one. That's a compliance swamp; do not step in it for revenue.

OP 050The AI moatOPS LAYER

Agents end-to-end.

Every competitor here is either a slow platform (upload → wait) or a single shop. The Counter's ops layer is the founder's actual edge:

Quote agent

STEP/Gerber in → feature recognition → routed vendor + price + SLA in minutes.

Dispatch agent

Picks the vendor from a live Fast-25 scorecard (speed history, current load), books the courier, opens the job thread.

Status agent

Proactive customer texts. No "any update?" emails, ever.

Vendor scorecard

Every job auto-updates the graph. Built as saved, versioned automations from job one — deterministic code, data as source-of-truth.

By month 12 this dataset — who in Houston is actually fast, at what, at what price — is worth more than the revenue. It is the district's founding asset.

OP 070Honest gaps & risksNONCONFORMANCE

What could kill it, stated plainly.

Geography

The founder currently lives out of state; this business's core loop is physical and in Houston. No version of year one works without meaningful ground time or a founding ground partner. This is the #1 open decision (Phase 1 options a/b/c).

Focus conflict

The founder's existing AI-services business is the current revenue mission; this is a second front. Mitigations: Phase 0 is automatable list-building/scoring that reuses the existing lead-gen stack, and the whole venture is an AI-ops showcase — but the CEO-desk question ("is this the week's needle-mover?") applies. Do not let this cannibalize the founder's current commitments. Phase 0 runs nights-and-weekends, with a go/no-go decision gate before any Phase 1 relocation money.

Numbers are unvalidated

Every dollar figure above is an assumption flagged for measurement, not evidence.

Priority is bought, not promised

If net-7 + monthly minimums don't actually move our jobs to the front of the queue, the model degrades to Xometry-with-a-phone. The Phase 0 speed test exists to find the shops where it does.

Anchor risk

TeraFab could slip or shrink (Musk announcements historically do). Mitigation: the brokerage works on Houston's existing base alone; TeraFab is upside, not dependency.

Sketch to part
in hours,
not weeks.

The gap is not machines.
It's coordination + priority.

Every request gets a time and a price.

Four operations, in sequence.

Vendor graph

First loop, boots on ground

Demand engine

The Counter becomes a place

Assumptions — measure in Phase 0/1.
Do not trust these numbers yet.

Agents end-to-end.

Quote agent

Dispatch agent

Status agent

Vendor scorecard

Six gauges. One that counts.

What could kill it, stated plainly.

Geography

Focus conflict

Numbers are unvalidated

Priority is bought, not promised

Anchor risk

The queue.

The gap is not machines.It's coordination + priority.

Every request gets a time and a price.

Four operations, in sequence.

Vendor graph

First loop, boots on ground

Demand engine

The Counter becomes a place

Assumptions — measure in Phase 0/1.Do not trust these numbers yet.

Agents end-to-end.

Quote agent

Dispatch agent

Status agent

Vendor scorecard

Six gauges. One that counts.

What could kill it, stated plainly.

Geography

Focus conflict

Numbers are unvalidated

Priority is bought, not promised

Anchor risk

The queue.

The gap is not machines.
It's coordination + priority.

Assumptions — measure in Phase 0/1.
Do not trust these numbers yet.