Blackwell B300s.
On Canadian soil.
Free to start.
Up to $5,000 in compute credits for NVIDIA B300 Blackwell GPUs. 275 GB HBM3e per GPU, NVLink-5, native FP4. Your data, weights, and metadata never leave Canada. No US hyperscaler in the path. No obligation.
Takes 60 seconds · No credit card · No sales call · Reviewed within 1 business day
NVLink-5, native FP4
2.2 TB per node
No US CLOUD Act exposure
Not a ticket queue
Three reasons builders pick us over the hyperscalers.
We are not cheaper AWS. We are a different jurisdiction, different hardware access, and a different support model.
Sovereign by architecture, not by policy
The workload, the network, and the operators are all in Canada. There is no US parent company that a subpoena can reach. PIPEDA-aligned, operationally aligned to SOC 2, PCI-DSS, and ISO 27001 controls. No reselling, no sub-processors you did not agree to.
B300s you can actually get on
Blackwell B300 SXM6 with NVLink-5 mesh and InfiniBand-ready fabric, plus B200, RTX 5090, RTX 4090, and custom configurations available on request. No reservation queue, no twelve-month commit.
Your data stays yours, always
Prompts and outputs are processed in memory and never stored. We never train on your data, never sell it, never share it. Usage metering tracks token counts only, never message content. Stripe is our only subprocessor. No inference subprocessors at all.
Three tracks, one fleet.
Prototyping a startup, running research in a lab, or building with us at a hackathon, pick the track that fits and claim it below.
Student / Researcher
Enrolled at a Canadian institution
- B300 access*
- Weekly engineering office hours
- Valid until credits run out
- Best for thesis, coursework, self-paced learning
Founder / Startup
Canadian or Canada-based startup
- B300 and B200 access*
- Engineering support on your project
- Valid until credits run out
- Best for prototyping and shipping products
Event Participant
Registered SPUR event attendee
- B300 access*
- On-floor mentors during the event
- Valid until credits run out
- Best for hackathons, workshops, demo days
*GPU access is available through our open source models. Bring your own weights or use ours.
What people ship with these credits.
Real projects from founders, students, and hackathon teams using SPUR credits.
RaceMind ↗
A real-time predictive analytics platform that turns OBD-II CAN bus telemetry into racing intelligence. Combines TimesFM for unsupervised anomaly detection with Gemma to translate technical alerts into clear guidance for drivers and engineers. Covers tread wear, fuel forecasting, and rules-based analysis.
Janus ↗
An offline edge-AI system that helps spacecraft decide which sensor to trust during a critical failure, when Earth is too far away to help. Compares conflicting IMU and camera readings, identifies the likely fault, and recommends a safe response. Runs entirely on a laptop using Gemma through Ollama. Inspired by the ESA Schiaparelli Mars lander failure.
Claim your credits
Four fields. We review within one business day and email your access.
Request received.
Check your inbox for a confirmation. Your credits will be activated within one business day and you'll receive your access details.
Your quote could be here next.
We are collecting testimonials from builders right now. Build something great with SPUR credits and yours could be the next one featured here.
Claim your credits →Form to first job in one business day.
Tell us what you're building
Pick your track, four fields, one paragraph about the project. No pitch deck, no call, no procurement process.
60 secondsAn engineer reviews it
A SPUR engineer reads it, sizes what you need, activates your credits, and emails your access details.
Within 1 business dayStart operating
Get started, pick your configuration, and launch. Our team helps you tune the stack and stays with you through to production.
Same day you're approvedSPUR vs. the alternatives
| Capability | SPUR | US hyperscaler | Own hardware |
|---|---|---|---|
| Data stays on Canadian soil | ✓ | ✕ | ✓ |
| No US CLOUD Act exposure | ✓ | ✕ | ✓ |
| Engineer on your project | ✓ | ✕ | ✕ |
| Free credits to start | ✓ | ✓ | ✕ |
| No capital outlay | ✓ | ✓ | ✕ |
| Nothing to rack, cool, or staff | ✓ | ✓ | ✕ |
Hosted in Canada.
No US CLOUD Act exposure.
SPUR's compute is operated end-to-end inside Canadian jurisdiction. No US CLOUD Act exposure. No cross-border transfers. PIPEDA-aligned by default and engineered to SOC 2, PCI-DSS, and ISO 27001 controls. Real engineers, no reselling.
- Hardware
NVIDIA B300 SXM6, 275 GB HBM3e per GPU, 2.2 TB per node. NVLink-5 mesh, InfiniBand-ready. B200, RTX 5090, RTX 4090 also available on request.
- Compliance
PIPEDA-aligned. Operationally aligned to SOC 2, PCI-DSS, and ISO 27001 controls. No CLOUD Act exposure. No cross-border transfers.
- Support
Every credit recipient gets a SPUR engineer. We help you size the cluster, tune the stack, and ship to production.
Where your data lives. And doesn't.
Where your data is processed
All inference runs on hardware in our Waterloo, Ontario data centre. Your requests are processed entirely within Canada. No US-headquartered cloud provider sits in the serving path, so your data carries no US CLOUD Act exposure. SPUR operates under PIPEDA.
Your prompts and completions
- Prompts and model outputs are processed in memory to serve your request only, and are not stored.
- We never train models on your data. The models we serve are open-weight models hosted by SPUR directly; no third-party model provider receives your requests or has visibility into your traffic.
- We never sell or share your content with anyone.
What we do store
| Data | Retention |
|---|---|
| Account information | Email and account settings, retained for the life of your account. |
| Billing records | Processed by Stripe. Invoices retained as required under Canadian accounting rules (7 years). |
| Usage metering | Model ID, token counts, timestamps, API-key ID per request, numbers only, never message content. Used for billing, capacity planning, and abuse prevention. |
| Support correspondence | Emails you send us, retained per our standard mailbox retention policy. |
GPU rentals and dedicated compute
Tenant data on rented GPU systems lives on the rented machine only. When a rental ends, tenant storage is wiped before the machine is reused.
Deletion and questions
You may request deletion of your account and all associated data at any time via spuric.com/contact. We will confirm completion within 30 days. Questions about this policy reach a human at the same address.
Subprocessors
Stripe (payments) is our only subprocessor. There are no inference subprocessors, all model serving is done on SPUR infrastructure.
Last updated: July 2026 · SPUR Compute (ai.spuric.com)
Questions people ask before claiming.
Is this actually free? What's the catch?
It is actually free. No credit card, no auto-conversion to a paid plan, no obligation to continue. We fund credits because builders who use SPUR credits tend to stay with us. If your credits run out and you want more, you talk to your engineer.
Do I have to be Canadian?
You need to be Canadian or Canada-based for the founder track, and enrolled at a Canadian institution for the student track. Event participants just need to be registered for a SPUR event.
What does $5,000 in credits actually get me?
Your engineer will size it against what you described in the form and tell you honestly whether it is enough for what you are trying to do. The credits don't expire until you've used them. No clock ticking in the background.
What happens when my credits run out?
Nothing breaks and nothing auto-charges. Your engineer will have flagged it before you get there. You can move to a paid plan, apply for an extension, or export everything and walk away. No egress fees either way.
Can I run my own models?
If you selected API credits + compute, yes. SSH, Jupyter, VS Code, ComfyUI, Ollama, and vLLM templates available. Standard PyTorch, JAX, and the usual training stack. Bring your own weights, containers, and data pipeline. If you selected API credits only, you get access to our hosted models as a service through an API endpoint at ai.spuric.com. OpenAI-compatible, no infrastructure to manage.
What's the difference between API credits and compute credits?
API credits give you access to our hosted models as a service. You call them through an API endpoint, we handle the infrastructure. Compute credits give you GPU access to run your own models, training jobs, and workloads. You can claim either or both in the form above.
How is this different from just using AWS or Azure credits?
Jurisdiction and support. Their credits run on infrastructure owned by a US company subject to the CLOUD Act, and their support is a ticket queue. Ours is operated in Canada, and you get an engineer on your project. If neither of those matters for your project, their credits are fine and we will tell you so.
What happens to my data and prompts?
Prompts and model outputs are processed in memory to serve your request only, and are not stored. We never train on your data, never sell or share your content, and no third-party model provider ever sees your traffic. Usage metering stores token counts and timestamps only, never message content. See our full data retention statement below.
Claim your compute credits
Up to $5,000 on NVIDIA B300s, activated within one business day, with engineering support. No credit card, no obligation.
Claim your credits →🍁 Sovereign by design · Waterloo, Ontario