Own the brain.
Give it hands.
A private AI brain that runs in the customer's own infrastructure, plus proprietary hardware hands that operate their real computers, tablets and phones. Nothing installed on the controlled devices, no per-user or per-token fee.
128 / 192 GB unified-memory private brain running 100B-class open models on site (zero paid cloud AI needed). Each hand has two proprietary boards: an HDMI/USB I/O board and an Agentic board.
One private brain; the hands work the devices, at the office or on the move (LAN and Wi-Fi today; remote 5G link: roadmap).
SMEs want AI that does the work. They cannot send their data to it, and chatbots can't click.
Data leaves the company
Law firms, accountants, clinics, engineering and defence suppliers cannot hand documents and know-how to an AI provider's cloud.
A meter on every prompt
Seats, tokens and subscriptions grow with every user and every task, forever.
AI can't touch their tools
Chatbots talk; cloud agents need installs or APIs. Old and closed applications stay manual.
HDMI in, USB out: the hand sees the screen and types like a person
The private brain (128 / 192 GB of unified memory, 100B-class open models on site, zero paid cloud AI) watches how a person does a task, learns the steps and repeats them through the hands on every connected device. Each hand has two boards: the HDMI/USB I/O board sees the screen and types, and the onboard Agentic board lets the hand finish the task if the link drops. Sending, paying or publishing waits for approval, with a physical approval key.
Others sell a brain without hands, hands without a private brain, or agents in someone else's cloud
Three kinds of products come close. None of them does all of this.
Price of a private AI server for an SME (before any hands)
Prices from our business plan's comparison table (public list prices, VAT excluded where stated). None of these servers has hands.
There are also 128 GB boxes and NAS without hands: Minisforum N5 MAX (about €3,150–€3,840 for 128 GB) and Ghost Core ($3,499). PRIVAIA's €4,000 includes the initial implementation, the orchestrator and the coordination of the physical hands.
Sell the brain once, then hands and hours
| What the customer buys | Price (+VAT) | Hardware reference cost |
|---|---|---|
| 1-node brain, initial implementation included | €4,000 | ≈ €3,800 incl. VAT |
| 2-node brain, bag of hours included | €10,000 | being quoted |
| 3-node brain, bag of hours included | €13,000 | ≈ €7,500 + VAT |
| Each extra hand (one per machine) | €250 | proprietary device |
| Customisation, support, new workflows | €150 / hour | team time |
Also on the price list (+VAT): 1-node 192 GB brain €7,000 · extra 128 GB brain node €3,800 (hardware only, no bag of hours).
Gross margins are being validated against full direct costs in the first deployments; we do not present them as final.
What one customer can be worth (from the price list)
Examples computed from list prices, not a forecast.
Regulated and data-sensitive SMEs, starting in Seville
Law and accounting firms, private clinics, engineering and manufacturing companies, defence suppliers and SMEs already paying for SaaS AI, in Spain and the EU. Direct sales through our commercial lead to an initial network of Seville companies, plus AI and cybersecurity consultancies and business associations. Public sector and defence only after a technical proof.
Working prototype. Honestly pre-seed.
Product
Private AI brain plus a physical hand operating today; productisation in progress.
Team of 4 full-time (2 employees)
PRIVAIA is led by Florencia Malale Molico (commercial lead), with a team of four working full time in Bormujos (Seville), Spain, including engineers from Germany and Ukraine.
As of November 2026: pre-revenue, no customer deployments, first unpaid pilots agreed with Seville companies (defence, drones and 3D design), no paid pilot yet, company not yet incorporated.
Lean today, milestones that are measurable
12-month targets from the business plan
| Months | What we do | Target result |
|---|---|---|
| 0–3 | Lab running; administration interface, users and agents; 15–20 SME interviews | Reproducible benchmark; two letters of intent |
| 4–6 | Server v1; first paid pilot | One paid pilot; customer cost and performance measured |
| 7–9 | Package server, support and telemetry | 3–4 customers or partnerships; first support tickets and updates |
| 10–12 | Stabilise the product | 5 customer relationships; 2 deployments; evidence for the seed round |
A full revenue forecast will be built from the first pilot's real costs and sales cycle; until then we show targets, not projections.
Each step ends with evidence an investor can check
- Day 0–15Lab runningReproducible benchmark of capacity, energy use and compatibility (lab ≈ €7,500 + VAT).
- Day 15–30Administration, users and logsAn administrable demo with local traceability.
- Day 30–45Five workflowsExcel analysis, Word from sources, invoice extraction, offer comparison and a management report, run and measured.
- Day 45–60Test with 5–10 usersLatency, errors, concurrency and limits documented.
- Day 45–75Commercial validation15 company interviews and two non-binding letters of intent.
- Day 60–90Target: a first paid pilotAcceptance, cost before and after, and a first reference.
€200,000 pre-seed for 10% of PRIVAIA
Use of funds: 2–3 sale-ready units and a repeatable installation; five tested workflows with a measured benchmark and acceptance criteria; legal and IP setup; targeted specialist work; B2B outreach and working cash. The tranche is tied to four gates (product, evidence, market proof, legal readiness). The seed round comes after paid validation and repeatable unit economics.
Proof before you decide
Florencia arranges access to the technical benchmark methodology, the private evidence video, or a live demonstration.