Difference The Node Pricing Included Security FAQ Request access

AI for law & finance firms

Your private AI chat
Your keys

A secure, ChatGPT-style workspace on a dedicated GPU in Johannesburg. Upload contracts and financials, ask questions, get answers. We deploy it, you own it — your firm holds every password and key, and nothing is ever sent to OpenAI or Anthropic.

0
To OpenAI
100%
In SA
PDF
& Excel
JHB
Teraco

The chat box is the tip.
We run the secure AI underneath.

Public AI tools send your contracts and client data to servers abroad. Building a private AI in-house means GPUs, models, vector databases and security work no law or finance firm wants to run. We build all of it, hand your firm the keys and step back — your data never leaves a single-tenant GPU in Johannesburg, and after handover we keep no access to it.

Iceberg: a small tip above the waterline, a much larger mass below
What you seeYour team's chat
What we run for youPrivate language modelDocument Q&A (RAG)Secure chat workspaceUser accounts & adminSingle-tenant GPU in JHBKeys held by your firm
01
Tell us how your team works
Contracts, due diligence, financial reports — we set the workspace up around the documents and questions your firm actually handles.
02
We build the private AI below the waterline
Model, document Q&A, secure chat and user accounts — installed and tested on your single-tenant GPU before handover.
03
Your team logs in on day one
At handover you change every password and keep the keys — from that point even we cannot open your files. Your team just signs in and asks.

One secure workspace. Built around your documents.

Every deployment is a single-tenant RTX PRO 6000 with 96 GB of VRAM at Teraco Johannesburg, running a private language model with document Q&A over your own files. Here is what your firm gets.

96 GB
VRAM single-tenant
Day 1
Team logs in
0
Data sent abroad
99.9%
Uptime target*

Ask your documents

Upload a 500-page contract or a stack of financial reports and ask plain questions — risks, clauses, numbers, dates. Answers are drawn from your files, on your server.

You hold the keys

Your model and data sit on one GPU reserved for your firm. At handover you set every password and encryption key — we keep no access to your documents, and nothing is sent to OpenAI, Anthropic or any third party.

Predictable monthly cost

One fixed monthly fee per firm — not per query, not per token. Budget it like any other professional subscription, with no surprise usage bills.

* Pricing in USD, excl. VAT (15%). Uptime target reflects Teraco facility availability; a formal SLA is available for custom setups on request.

Simple, fixed monthly pricing

Dedicated nodes billed monthly on transparent commercial terms — no usage metering or surprise bills. Prices in USD, excl. 15% VAT; indicative ZAR shown as a guide and may move with FX.

Private AI API
$2,199
/ month · excl. VAT
Approx. ZAR: ~R36,200 / month

For firms with a developer or IT partner. We deploy the model on your dedicated GPU and hand over the keys; your developer builds on a private, OpenAI-compatible endpoint.

  • Dedicated RTX PRO 6000 · 96 GB, single-tenant
  • Private OpenAI-compatible API endpoint
  • 70B-class open model (Llama), ready to call
  • Your data stays on the node in Johannesburg
  • Host & hardware monitoring included
Request access
Early access
Private AI Workspace
$2,799
/ month · excl. VAT
Approx. ZAR: ~R46,000 / month

For partners and finance teams. We deploy a ready ChatGPT-style workspace on your node, hand over the keys and step back — your team logs in on day one.

  • Everything in Private AI API
  • Browser chat workspace, no setup for your team
  • Upload contracts & reports (PDF, Word, Excel)
  • Document Q&A over your own files (RAG)
  • User accounts & admin panel for your firm
  • Sized for teams of up to 30 users
  • Handover: you set every password — we keep no access
Request access

YSBERG vs public AI tools

The model is similar everywhere — the difference is where your contracts and client data end up.

Swipe to compare →

Feature Buy your own Hourly cloud YSBERG Bare Metal

Built for South African data, end to end

Your AI runs on one GPU reserved for your firm at Teraco Johannesburg. Documents and questions stay on that node — never sent to OpenAI, Anthropic or any third party — which supports POPIA data-residency expectations. At handover your firm sets every password and encryption key — we keep no access, and your firm controls who logs in.

Customers remain responsible for their own application security, data governance and internal access management.

POPIA-aware hosting

Data resident in South Africa at Teraco JHB

You hold the keys

At handover you set every password and key — we keep none

Single-tenant

One GPU reserved for your firm — never shared with other clients

Tier III facility

Teraco JHB — carrier-neutral, redundant power, 24/7 security

Secure server node connected to VPN access, firewall, data residency in South Africa and a security shield

Common questions

Start a conversation

Tell us about your workload and we'll come back with the right tier and a short pilot plan for your firm.

Prefer email? sales@ysberg.co.za

Please enter your name.

Please enter a valid email.

By sending this enquiry you agree to our privacy notice. In line with POPIA, your details are used only to respond to your enquiry, are kept confidential, and are never sold or used for marketing without your consent.

Request access →