Home/The Numbers
Every figure, every source, every caveat

The leverage was always
there. It was waiting on
coordination.

We don't ask you to take the pitch on faith. Here is every number behind it — the market data, the legal research, the technical landscape — with sources named and the honest caveats stated plainly. Including the data that argues against our own instincts.

The shape of the market

Big Tech looks like everything. Measured against the whole economy, it's a sliver.

United States

Revenue, by bloc

Small business ~$13.3T
Mid-market & large ~$10.0T
The Magnificent Seven ~$2.0T
Worldwide

Revenue, by bloc

SMEs & informal majority
Larger enterprise large
Big Tech ~3%
The opportunity

Work that's shared

Common back-office shared
Common ops & support shared
Common sales motion shared
Truly differentiated ~20%

Taken all together, Big Tech is a sliver. In the US, small business alone outweighs the Magnificent Seven by roughly 6.6× in revenue — and worldwide it isn't close. Meanwhile an estimated ~80% of what every business does is the same work. The leverage was always there; it was only ever waiting on coordination.

Caveat, stated plainly: the market-size figures above are illustrative order-of-magnitude comparisons drawn from public revenue aggregates, not a single audited dataset. The "~80% shared work" figure is a directional estimate of how much business activity is common back-office, operations, support, and sales motion versus genuinely differentiated. Treat these as the shape of the argument, not decimal-precise accounting.

ComparisonFigureNote
US small business vs Big Tech (revenue)6.6×$13.3T vs $2.04T — the most conservative cut
US small + mid-market vs Big Tech11.6×Bundle in mid-market
Worldwide small + mid vs Big Tech~30×Even more lopsided globally
Big Tech share of US business revenue8%Mag 7 vs small + mid
Big Tech share of world business revenue3%Same comparison, global
US firms vs Big Tech firms36.2M vs 7Roughly 5 million to 1
Worldwide SMEs vs Big Tech firms~400M vs 7Roughly 57 million to 1
Business work that's shared across industries~80%Customer, ops, marketing
The mechanism, in one move

Roughly 80% of what every business does is the same work — it's why CRMs are everywhere. Their whole model is to solve that shared 80% once and bill each of us separately, forever, as if our problem were unique. It isn't. A problem millions of us share is one millions of us can solve together — one time. The imbalance was never about size. It was about coordination.

Market research · the demand is real

"Own your AI" is validated — and the research tells us exactly how to sell it.

$8.84Mraised by UGREEN's AI NAS — comparable own-it-no-subscription hardware demand
Top-tierself-hosted private-AI hardware is one of the better-performing crowdfunding categories
Earlysurgea strong early surge almost universally predicts campaign success — driven by pre-warmed demand

The "own your AI" hook is validated and commercially strong. Comparable campaigns confirm strong paying demand for the own-it-no-subscription pitch. And wins are driven by pre-built warm demand, not cold discovery — the strongest campaigns harvested demand built beforehand. The practical implication is a pre-warmed list converting early, which is exactly the warm-network, charter-member motion this plan uses.

The finding that disciplines our own copy

Leading with "AI" underperforms. Research found AI-adoption framing associated with lower pledges and fewer backers, attributed to trust friction. So the durable pitch is ownership, continuity, and privacy — with AI as the engine, not the headline. That's why this whole site sells sovereignty first and says "AI is the engine, not the headline" everywhere.

And the finding that killed our original idea: cooperative crowdfunding has the weakest track record of any category. The original concept's differentiator — its mutualist/community structure — was also its weakest fundraising asset, and was structured to remove the founder from durable income. That finding, more than any other, drove the pivot to the current product-first, founder-anchored, two-entity design. We're showing you the data that argued against where we started.

The legal research · the load-bearing material

The structure isn't a hunch. It rests on named authorities.

Colorado ULCAAC.R.S. Title 7, Art. 58
Permits a single cooperative with two member classes and a two-tier voting structure that keeps control with patron members regardless of how much investor capital is admitted. No hard member minimum, no citizenship or residency requirement to form or own — a Colorado registered agent is required.
The Forman shield421 U.S. 837 (1975)
An interest purchased to use a good or service — rather than to profit from the efforts of others — is not a security, and the label on the instrument doesn't control; economic reality does. Protects the patron purchase. Does not reach the investor contribution.
The Howey test328 U.S. 293 (1946)
The investment-contract test that makes any capital-for-a-return arrangement a security no matter how it's wrapped. There is no entity arrangement that converts an investment into a non-investment.
Subchapter TIRC §§1381–1388
The cooperative tax regime: patronage-sourced income is deducted at the co-op and taxed to patrons. Treatment depends on the defining cooperative features — subordination of capital, democratic control, proportionate allocation — which directly constrain how large and senior any investor return may be.
Substance over formIRC §7701(o), §482
The economic-substance, step-transaction, and sham doctrines, plus §482's control-based reallocation authority, mean interposed entities, offshore structures, and nominee ownership are disregarded when their purpose is to detach a taxpayer from income or control on paper.
APA, not PLRRev. Proc. 2015-41
The arm's-length pricing of the license fee is a §482 transfer-pricing matter — in the IRS no-rule category, so it can't be settled by a private letter ruling. An Advance Pricing Agreement is the instrument that exists for it; critically, it prices control and cannot make controlled entities appear independent.

Caveat: legal authorities are cited by name; the procedural ones (e.g. the annual Rev. Proc. series — currently the 2026 series) are reissued yearly and must be verified with counsel before any ruling request. This is research that informs the design, not legal advice.

The technical landscape · point-in-time, May 2026

Open-weight models are now competitive — and that's what makes any of this possible.

The open-weight field is competitive with proprietary models for coding, reasoning, summarization, and structured workflows. The consistent case for local hosting is exactly the pitch the venture sells: privacy and data control, predictable one-time cost versus recurring subscription, no API rate limits or mid-project policy changes, and full customization.

Model familyBest forLicense
Qwen 3.5 / 3.6Best all-round local / agenticApache 2.0
Gemma 4Efficient / edge / multimodalApache 2.0
Phi-4 familyConstrained hardwareMIT
DeepSeek V4High-end coding / reasoningMIT
Mistral familiesEU-sensitive / low-latency single-GPUVaries — verify
Llama 4General (with conditions)Community license

Standard self-hosted tooling: Ollama (single-node), vLLM (multi-node), Open WebUI (chat), Node-RED (workflow), Podman/Docker (packaging), with quantization (e.g. Q4_K_M) the standard technique for fitting larger models into available VRAM. Full licensing detail lives on The Stack page. These figures are point-in-time and will move — the architecture won't.

Sources

Where every figure comes from.

Legal & structural

Colorado ULCAA, C.R.S. Title 7, Article 58 (incl. §§7-58-301, -304, -511 to -514, -1009). United Housing Foundation v. Forman, 421 U.S. 837 (1975); SEC v. W.J. Howey Co., 328 U.S. 293 (1946). Securities Act exemptions: Regulation D, Regulation Crowdfunding, Regulation S. Internal Revenue Code Subchapter T (§§1381–1388); §482; §7701(o). Capper-Volstead Act (8% dividend tradition). Rev. Proc. 2026-1 and 2026-3, superseding the 2025 series; Rev. Proc. 2015-41 (APA program); General Legal Advice Memorandum 2025-001 (periodic adjustments to high-value intangibles).

Technical landscape (May 2026)

Hugging Face and community surveys of open-source / open-weight LLMs to run locally in 2026; comparative guides (Overchat, Codersera, Pinggy, ComputingForGeeks, AceCloud); self-hosted-AI guides (KDnuggets, DreamHost) for tooling and hardware context.

Market size & the lopsided comparison

SBA Office of Advocacy (US small business count — 36.2M firms — and small-business share of GDP). National Center for the Middle Market (US mid-market ≈ $10T). World Bank / World Economic Forum (global SMEs ≈ 50% of GDP, ~400M firms worldwide). Magnificent Seven 2024 10-K filings (Big Tech aggregate revenue ≈ $2.04T). RSM Global / Oxford Economics (global mid-market sizing — the squishiest of the inputs). McKinsey / Salesforce / BLS work-allocation studies (the illustrative ~80/20 split between shared and genuinely differentiated business work).

Origin & market research (May 2026)

UGREEN AI NAS and NASync campaign data (Kickstarter, tracked via BackerTracker / Kicktraq); The Conversation (Startnext path-dependence study); Song et al., Accounting & Finance (AI-framing underperformance); Shareable (co-op crowdfunding failure pattern); New Internationalist community-share figures; ScienceDirect (RegCF survival study); Qubit Capital (platform success rates); Public AI Switzerland and the Platform Cooperativism Consortium / HBR / SSIR coverage (cooperative-AI movement context).

Caveats on the market comparison. The 6.6× figure is the most conservative cut (US small business alone vs Big Tech); it grows to 11.6× with mid-market bundled in and to roughly ~30× globally. The Big Tech and US numbers are hard — drawn from 10-K filings and SBA data — but the global SME/mid-market split is estimated, anchored to the World Bank's "SMEs ≈ 50% of global GDP" and scaled to the US ratio. Note too that revenue ≠ GDP ≠ market cap: Big Tech's ~$22T market cap flatters it far more than its revenue does, so a revenue comparison is the unflattering-to-us, honest one. The 80/20 shared-vs-differentiated split is illustrative — the broad ratio is robust across studies, but the inner 35/28/17 breakdown of the shared portion is not three-decimal precise. And "small business" definitions overlap across sources, so the bands are deliberately drawn as order-of-magnitude, not to the dollar.

A standing caveat on all of it. Crowdfunding, model, and tax-procedure figures are point-in-time as of the dates noted and will change. Legal authorities should be verified in their current form with counsel — the procedural ones in particular are reissued annually. Nothing here is legal, tax, or investment advice. We'd rather show you the seams than imply a precision we don't have.

The numbers point one way.
Coordinate.

Request an invitation to the founding cohort. We'll quantify your specific exposure — and what owning the stack changes about it.