TorontoGPU, 100% Canadian owned
Canadian-owned compute for AI, AI agents and models.
Three ways in, one company. Call an open model served from Toronto, rent an agent box that lives next to it, or reserve whole GPUs by the month.
- Owned in Canada, not only located here
- Hardware at one site in Toronto, run by a named operator
- Start by yourself: no call, no sales team
Questions first? Email Ali, the founder
One company, three steps up.
Start with a model call. Move to a box of your own when your agents need a home. Take whole GPUs when the work is steady.
-
Coming soon
Model API
Open models, served from our GPUs in Toronto. Call them the way you call any model API and pay for what you use.
- For
- Apps and agents that need a model, not a machine
- You pay
- Per use
- You start
- By yourself, with a sign-up
-
Coming soon
Agent box
A hosted agent that lives next to that model: a box of your own, set up with the harness you use, so prompts and files stay on hardware in Canada.
- For
- Teams that run agents every day
- You pay
- By the month
- You start
- By yourself; your setup is ready when you log in
-
Fully rented today, list open
Dedicated GPUs
Whole NVIDIA RTX PRO 6000 Blackwell GPUs, 96 GB each, yours alone. Bare metal or containers, direct SSH. This is what we run today.
- For
- Training, fine-tuning and steady inference
- You pay
- By the month, quoted in CAD
- You start
- Hold a place: 1, 2, 4 or 8 GPUs
Said plainly: all 12 of our GPUs are rented today. The first two steps open when capacity does, and the list tells us what to build first. Nothing here is a promise of a date or a price.
TorontoGPU is 100% Canadian owned. A Canadian address on a foreign company’s cloud is not the same thing.
The hardware is in Toronto, Canada, at one site that a Canadian company owns and runs.
Ali Hamadi owns and runs the hardware, and answers your email in Toronto business hours.
An agent box runs the harness you already use.
These are the setups we plan to have ready on day one. Name yours in the form’s note and we will know what to build first.
- Claude CodeAnthropic’s coding agent
- CodexOpenAI’s coding agent
- OpenClawAn open agent harness
- HermesAn open agent harness
- Your ownSSH in and install what you build with
Claude Code and Codex talk to their makers’ cloud models unless you point them elsewhere. If your data may not leave Canada, run your agents against the open model next to your box.
Straight answers.
When do the Model API and agent boxes open?
There is no date yet. Every GPU we run is rented today. The list tells us how much to build, and the people on it hear before anyone else.
What will they cost?
Prices are not set. You will get them in CAD, in writing, before any payment, and joining the list commits you to nothing.
What hardware is behind them?
The plan is the same cards as our fleet: NVIDIA RTX PRO 6000 Blackwell GPUs with 96 GB each, at our Toronto site. Which open models we serve first depends on what the list asks for.
Does my data stay in Canada?
A model served from our GPUs runs in Toronto, on hardware a Canadian company owns. A harness that calls another company’s cloud model sends your prompts to that company; that choice stays yours.
I need whole GPUs now.
That is step three and it exists today: dedicated RTX PRO 6000 Blackwell GPUs by the month. They are fully rented, and the list is served in order. Hold your place.
Get on the list, or email the founder.
Tell us what you are building and what must stay in Canada. It shapes what we open first.
Get on the list