Private AI on your own hardware: an on-premises AI workstation.
An on-premises AI workstation runs language models inside your building. Prompts, documents, and outputs stay on a machine you own instead of going to an outside AI service, and the system keeps working without an internet connection. We build them from one GeForce RTX 5090 to four RTX PRO Blackwell GPUs.
- On-siteprompts and documents stay in the building
- 32–384GBVRAM
- 1–4GPUs
- 3-yearwarranty, lifetime support
Why run AI on-premises
Data stays in the building
Client files, contracts, records, and source code are processed on your own machine. Nothing is sent to a third party for inference.
Works offline
Once the models are downloaded, inference runs without an internet connection.
Hardware you own
The cost is the machine and its electricity, instead of per-seat or per-use fees that grow with every user.
Your choice of models
Run open-weight models from the families your team prefers, and change them as better ones are released.
An on-premises system keeps data out of third-party AI services. Compliance with HIPAA, legal, or financial rules also depends on your access controls, policies, and network, so involve whoever handles compliance in how the system will be used.
Sizing for a team
Two things decide the size: the model you want to run, and how many people use it at once.
Model weights occupy roughly parameters × bits per weight ÷ 8. A 70-billion-parameter model at 4-bit quantization is about 35GB of weights; with context and runtime overhead it requires approximately 48GB of VRAM. Context length and batch size add to every figure below.
One person, mid-size models
A 30B-class model at 4-bit quantization fits in 32GB: the JNCS AI 5090, on a standard outlet.
RTX 5090 AI workstation →A small team, 70B-class models
A 70B model needs approximately 48GB at 4-bit. One 72GB or 96GB RTX PRO card leaves room for several people's conversations at once.
RTX PRO Blackwell workstation →More users, longer documents
Each active user needs GPU memory for their own context, and long documents need more. Two or four GPUs provide 144GB to 384GB.
Multi-GPU AI workstation →Not sure yet
Tell us the model, the number of users, and the kind of documents, and we size the system with you before you request a quote. Call (585) 388-8780 or email sales@jncs.com.
Where it goes
A single-GPU system, or two RTX PRO 4500 or 5000 cards, runs on a standard 15A, 120V circuit. Two RTX PRO 6000 cards, or any four-GPU configuration, require a dedicated 20A, 110–120V circuit, with a 30A, 220–240V circuit recommended. Four RTX PRO 6000 cards draw approximately 2,850W and require a dedicated 30A, 220–240V circuit. The configurator shows the requirement for every configuration as you build it.
Plan any electrical work with your building before delivery. Every build is a tower; for rackmount servers, see AI servers and larger systems. Businesses in the Rochester area can pick the system up at our Fairport shop.
Assembly, burn-in, and delivery
Each system is assembled by hand at our shop in Fairport, NY, by a shop that has built computers for more than thirty years.
Fit and power validated
The configurator checks physical fit, memory, PCIe slots, and power on every selection. The power supply is sized automatically to the GPU count and estimated draw, with a 12V-2x6 cable for every GPU.
Burn-in under load
Every system is stress-tested under sustained full load and firmware-updated before it leaves the bench, so the first long job is not the first test.
Windows 11 Pro installed
Windows 11 Pro is installed and updated before shipment.
Warranty and support
Every build carries a 3-year warranty and lifetime technical support from the shop that built it.
Quoting and payment
The configurator shows a price for every part and an estimated total, including assembly, burn-in, and testing, as the build is configured. GPU and memory pricing changes frequently, so a technician confirms GPU availability and final pricing by email, with lead time, typically within 1-2 business days. There is no cost and no obligation to request a quote.
Payment for AI workstations is by ACH bank transfer; credit cards are not accepted for these systems. An invoice with bank details is issued once the quote is approved. Systems ship across the United States from Fairport, NY, with local pickup available.
On-premises AI questions
Is a local LLM private?
How many users can one AI workstation serve?
Does an on-premises AI workstation need an internet connection?
How does the cost compare with a cloud AI subscription?
Can it plug into a normal office outlet?
How is an AI workstation ordered and paid for?
Related AI workstation pages
Local LLM workstation guide
How much VRAM a model needs, and which build provides it.
Read →RTX PRO 6000 Blackwell workstation
One or two RTX PRO Blackwell GPUs, up to 192GB VRAM, Threadripper 9000.
Read →Multi-GPU AI workstation
Up to four RTX PRO Blackwell GPUs, up to 384GB VRAM, Threadripper PRO 9000.
Read →RTX 5090 AI workstation
One GeForce RTX 5090, 32GB VRAM, Ryzen 9000, standard 15A circuit.
Read →Dual RTX PRO 6000 workstation
192GB of VRAM on two 96GB GPUs, and the circuit it needs.
Read →For universities and labs
Spec sheets and RFQs quoted as written, NDAs, matched sets.
Read →Size a private AI system for your team.
Configure a build online, or tell us the model and the number of users and we size it with you. Quotes are free, with no obligation.
Have an exact parts list or a department spec sheet? Send the whole list. For more than four GPUs or rackmount systems, see AI servers and larger systems.