# Offers — overview
> **Audience.** Buyers. This page maps *everything* Bilbs AI sells onto a
> single table so you can skim it in 60 seconds. Jump into any line for
> detail.
Bilbs AI sells two things:
1. **The Bilbs Box** — a physical AI appliance you own outright. Four sizes.
2. **Engagements** — fixed-scope services, with or without a Box.
Both are governed by the same MSA + DPA ([engagements/msa-dpa.md](./engagements/msa-dpa.md)).
---
## 01 · The Bilbs Box (flagship)
The Box ships pre-flashed with your fine-tuned model, orchestration
layer, admin UI, eval harness and runbook. Plug it in, point your VPN at
it, done. **You own it outright — no subscription required.**
| Tier | Concurrent users | GPU | One-time | Detail |
|------|-----------------:|-----|---------:|--------|
| [Box Mini](./tiers/box-mini.md) | 5 | 1× RTX 6000 Ada 48 GB | **$14,000** | 1U · solo / pilots |
| [Box](./tiers/box.md) | 25 | 1× RTX PRO 6000 Blackwell 96 GB | **$28,000** | Desktop · small–mid teams · ★ most popular |
| [Box Pro](./tiers/box-pro.md) | 200 | 2× H100 NVL 188 GB | **$78,000** | 4U · growth / mid-market |
| [Box Cluster](./tiers/box-cluster.md) | 1,000+ | 8× H100 / H200 | **from $280,000** | Half-rack · enterprise / regulated |
> **Included in every tier:** the Box, pre-burned weights, admin UI,
> orchestration + eval runtime, a printed + PDF runbook, a spare NVMe,
> and a 30-day post-launch Slack channel. Full list →
> [setup/whats-in-the-box.md](./setup/whats-in-the-box.md).
### Optional monthly plans
Both plans are **opt-in**. Skip both and your recurring bill is **$0 / mo**.
| Plan | What it covers | Mini | Box | Pro | Cluster |
|------|---------------|-----:|----:|----:|--------:|
| [Updates](./plans/updates-plan.md) | New base-model weights (Gemma 5, Llama 5, Qwen 3…), security patches, dependency refresh, eval-harness CI | $99/mo | $149/mo | $299/mo | from $699/mo |
| [Support](./plans/support-plan.md) | Private Slack, < 4 h response, 24/7 Sev-1 pager, incident triage, OEM RMA shepherding | $249/mo | $399/mo | $799/mo | custom SLA |
**Warranty.** Hardware ships with the OEM's standard 3-year warranty
(NVIDIA / Dell / Supermicro). The Support plan will shepherd the RMA for
you; without it you claim directly with the OEM — instructions in
[operations/troubleshooting.md](./operations/troubleshooting.md).
---
## 02 · Engagements
### Audit — 2 weeks · **$12,000** fixed
A written assessment of your existing AI stack or proposed stack.
**You get:** an architecture + cost review, an eval-suite draft against
your real data, a prioritised roadmap doc, and one follow-up call. Good
for teams who aren't sure whether they need a Box, a Build, or nothing
at all.
**Does not include:** any code shipped to production.
### Build — 6 weeks · **$48,000** fixed
Our flagship engagement. Ship a production private-AI system in six
weeks, or your money back.
Week-by-week process: [engagements/process.md](./engagements/process.md).
Exact deliverables: [engagements/deliverables.md](./engagements/deliverables.md).
**Covered under the ship-or-refund guarantee.** If the week-1 spike
doesn't hit the signed acceptance criteria, you keep the audit doc and
get a 100% refund. In 40+ engagements since 2024 the refund clause has
never been invoked.
**Does not include:** the hardware — the Box is sold separately and
added as a line item to the Statement of Work. A Build priced against
a Box already in your rack is $8k cheaper because the runtime is
fixed.
### Embedded — monthly · **$18,000 / month**
A rolling retainer. A single senior operator embedded into your team at
~30 hours / week — shared Slack, PR review authority, on-call for
production AI.
- Two weeks' notice either side.
- Minimum term: one month.
- No access to the "most popular quarter slot" limit — this is for teams
who've already bought a Box or done a Build and want ongoing depth.
### Custom engagements
For anything that doesn't fit the three shapes above — typically
regulated-industry deployments, cluster-scale rollouts, or work on
behalf of a venture fund — we'll quote against a written RFP.
Timeline: 72 hours from brief to fixed price.
---
## 03 · How to decide
| You are… | Start with |
|---------|-----------|
| A solo founder prototyping on OpenAI and nervous about your bill | [Box Mini](./tiers/box-mini.md) |
| A growth-stage SaaS with enterprise deals stuck in legal | [Build](#build) + [Box](./tiers/box.md) |
| An enterprise IT team that already has a GPU fleet | [Audit](#audit) first |
| A regulated org (healthtech, fintech, defence) that needs airgap | [Box Pro](./tiers/box-pro.md) or [Cluster](./tiers/box-cluster.md) |
| Unsure | [Book a 20-min sizing call](https://cal.com/bilbsgroup/bilbs-ai) |
## 04 · What we do not sell
- **Token-metered API access.** We do not run a public inference API.
- **Seats in a multi-tenant SaaS.** Every deployment is single-tenant.
- **Training-data brokerage.** We never resell, aggregate, or use
customer data to train anything beyond that customer's own model.
- **Ad-based anything.**
## 05 · Ordering & payment
- **Currency:** USD for the Box, CAD available on request.
- **Payment milestones:**
- Box: 50% on PO, 40% on shipment, 10% on UAT sign-off.
- Build: 33% on SOW, 33% on week-3 demo, 34% on handover.
- Audit / Embedded: Net-15 monthly.
- **Method:** wire / ACH / Stripe. Crypto (USDC) for clients that
require it.
- **Taxes:** QC HST / GST added for Canadian clients. Not added for
foreign invoices, reverse charge applies.
Questions on commercial terms: [admin@groupebilbs.com](mailto:admin@groupebilbs.com).