● Pre-seed · 2026 · Now raising · Product in production
A dedicated AI computer for accounting practices, physician groups and other regulated firms. No outbound path. US-developed models. Fully managed for a flat monthly fee.
$23K
Monthly recurring revenue
$276K
Annualized run-rate
Live
In production at a paying customer
$3.0B
US TAM in annual service fees
Raising
$10M pre-seed
Pre-money
$15M
Post-money
$25M
Runway
36 months to Series A
Problem
01
Staff at accounting practices and physician groups are already pasting work into chatbots. Firms need a sanctioned option that keeps client and patient data in-house.
02
ChatGPT Enterprise reportedly needs ~150 seats (~$108K/yr). Microsoft 365 Copilot processes data in Microsoft's cloud, and its web searches fall outside the HIPAA agreement.
03
On Oct 26, 2026, Cohere reported a popular Chinese open-weight model tried to upload a file to a Beijing server during a sandboxed test.
Solution
Installed on-site by default; Tier 3 datacenter hosting optional.
The system can't reach the internet. Data never leaves the network.
OpenAI gpt-oss-120b today; model-agnostic platform.
Monitored, patched, backed up and audit-ready, 24/7.
Staff
Existing company sign-in. Office network or VPN only.
Zenthral system
Department workspaces · approved document collections · gpt-oss-120b + fast companion model · permissions mirror the firm's own
Internet: blocked
No outbound path. Tools and web access only if the client approves.
Operated by Zenthral: 24/7 monitoring, monthly tested patches, nightly backups, quarterly access reviews and model upgrades, monthly no-outbound verification.
Product · tiers
| Tier | Hardware | Users / at once | Hardware price | Monthly |
|---|---|---|---|---|
| Essentials | 128GB unified memory, ~256GB/s | 10 / ~4 | $4,860 | $595 |
| + GPU option | Nvidia RTX Pro 4000 SFF Blackwell, 24GB | 10 / 4 | $4,860 | $595 |
| Professional | 128GB unified memory, 614GB/s | 25 / 6 | $6,480 | $1,095 |
| Enterprise | 96GB Blackwell GPU, 1,792GB/s | 75 / 40+ | $30,480 | $1,995 |
"At once" = users getting answers simultaneously at 15+ tokens/sec (~3× reading speed). Monthly shown on-premises; hosted adds $200–$950. Hardware sold at cost + 20%.
Traction
$23K
Monthly recurring revenue
$276K
Annualized run-rate
Live
Working product sold to a paying customer
Lab reference build in place: every capacity figure is confirmed on our own hardware before it's quoted. NVIDIA, Meta, Google and IBM models in qualification.
Why now
$11M
Seed raised Oct 5, 2026 by Ghost (led by Andreessen Horowitz) for Core, a $3,499 personal AI computer that runs agents locally — before shipping a single unit.
Zenthral is already live at a paying customer, with $23K MRR.
US-developed, Apache-licensed models now run on a single office machine.
The Qwen exfiltration report makes "which model, and can it phone home?" a board question.
59% of US workers use unapproved AI tools; 75% of them have shared possibly sensitive data.
Market size · US
| Vertical (10–99 staff) | Firms | ARR |
|---|---|---|
| Physician practices | 56.5K | $909M |
| Accounting firms | 12.1K | $190M |
| Dental practices | 36.3K | $515M |
| Home health, therapy, behavioral | 34.9K | $586M |
| Law firms | 20.7K | $327M |
| RIAs, insurance, banks, credit unions | 14.8K | $298M |
| Police agencies (CJIS) | 9.2K | $151M |
| Upside: CMMC defense contractors | ~76K | not sized |
TAM · 7 regulated verticals
$3.0B ARR
$13.4B incl. hardware + implementation over 3 years
SAM · accounting + physician
$1.1B ARR
68,600 firms · $4.9B over 3 years
Landscape
| Buyer | Where data goes | Pricing | |
|---|---|---|---|
| Zenthral | Regulated firms, 10–75 users | Client's office, no outbound | $4.9K–$30.5K hardware + $595–$1,995/mo |
| Ghost Core | Individuals | On device | $3,499 one-time |
| M365 Copilot | Microsoft 365 shops | Microsoft cloud | $21/user/mo |
| ChatGPT Enterprise | Large orgs | OpenAI cloud | ~$60/user/mo, ~150-seat minimum (reported) |
We don't beat Copilot on price or raw model strength. We win where regulators, insurers or clients require data off third-party infrastructure — with a fixed cost that doesn't grow per seat.
Business model
| Essentials | Professional | Enterprise | |
|---|---|---|---|
| Hardware (cost + 20%) | $4,860 | $6,480 | $30,480 |
| Implementation, fixed fee | $7,500 | $9,800 | $14,300 |
| ARR per client (on-prem) | $7,140 | $13,140 | $23,940 |
| 3-year revenue per client | $33,780 | $55,700 | $116,600 |
60–80%
Operating margins
Priced against the alternatives, not billed by the hour.
Go-to-market
Beachhead
Firms with 10–75 users, where data must stay off third-party servers.
Wedge
Credited in full if the client proceeds within 90 days.
Deployment
Sign-off at each: discovery, build and harden, pilot, rollout.
The ask · use of funds
By month 24
600
clients deployed
$10M
ARR
SOC 2
complete; HIPAA clients in our datacenter
5
verticals
Data room
Includes the full pre-seed deck, financial model, customer references, lab benchmark results and security architecture.
Or reach the founder directly: Riess@Zenthral.ai