Skip to content
Small team, full backlog, zero orders dropped. Support replies are slower than we’d like. Read our status update → Zero orders dropped. Status → 📬 Check your spam folder — most of our replies land there. We do answer. Status update → 📬 Check your spam folder. Status →
1325 Rue Bergar, Laval, QC support@d-central.tech 1-855-753-9997 Mon-Fri: 8:30 - 18:00

AI Consulting Quebec City — On-Premise AI and Local LLM for Capitale-Nationale

D-Central Technologies — decentralized computing powering your digital sovereignty.

D-Central provides on-premise AI and local-LLM consulting for organizations in Quebec City and the greater Capitale-Nationale region — scoped remotely, hardware configured at our Montreal facility, and deployed on your premises anywhere in the region, in French and English. If your organization works with the provincial government, operates under Quebec’s Law 25, or simply wants AI that runs on hardware you own and cannot be suspended by a foreign vendor’s policy change, this is where to start.

Quebec City is where Law 25 is enforced. The Commission d’accès à l’information (CAI) — the provincial body that administers the Act and investigates breaches — has its offices on the Grande Allée. If you are a government contractor, a civil-service supplier, or a professional practice in the capital, your data-residency posture is not abstract risk management: it is visible to the institution with the mandate to audit it.

Why Quebec City organizations are choosing on-premise AI now

  • Law 25 and the CAI’s home city. Quebec’s Act to modernize legislative provisions as regards the protection of personal information (S.Q. 2021, c. 25) requires organizations to assess and justify cross-border transfers of personal information and to ensure adequate protection before those transfers happen. Every prompt containing personal information sent to a US-hosted AI platform is, in law, a transfer outside Quebec — one that the CAI has the authority to investigate. A fully local deployment, where data never leaves your premises, removes the transfer at the source. Our technical guide is at Quebec Law 25 and on-premise LLM. (This is technical orientation, not legal advice — confirm specifics with qualified counsel.)
  • The CLOUD Act and Canadian hosting. Storing data on a server located in Canada does not, by itself, eliminate US-government reach. The Clarifying Lawful Overseas Use of Data Act (2018) permits US authorities to compel US-headquartered providers to produce data they control, wherever it physically sits. If your cloud AI provider is a US firm — and the major ones are — that reach applies to your queries and documents. Owning the hardware and running your own inference removes the US firm from the picture entirely. See CLOUD Act and Canadian AI for a full breakdown.
  • Vendor reliability after June 2026. A US export-control directive in June 2026 required a major AI provider to disable two frontier models for all foreign nationals without warning. Organizations that had embedded those models in internal workflows lost access with no appeal process. A model running on hardware you own cannot be suspended by a provider’s policy change, an export-control classification, or a platform outage. See Sovereign AI in Canada for the argument.
  • French-first operations. Quebec City is the most francophone major city in Canada. Bill 96 — the Act respecting French, the official and common language of Québec (2022) — reinforces the right of workers to use French in the workplace and creates obligations for employers. Local open-weight models can be deployed with French-first interfaces, French prompt defaults, and French documentation. Cloud providers configure this at their discretion; you configure yours.
  • Government procurement and supply-chain posture. Organizations that supply services to the Quebec provincial government or its agencies face increasing scrutiny of their data-handling practices. A documented on-premise AI posture — no cross-border inference traffic, no third-party data processing — is a defensible answer in a supplier questionnaire or a CAI audit.

Remote-deploy model: how we serve Quebec City

Our facility is at 1325 Rue Bergar in Montreal — approximately 250 kilometres from Quebec City, or two and a half hours by road. On-site arrival in the capital is possible for larger engagements, but most Quebec City deployments follow our remote-deploy model: we scope the right hardware for your workload, build and test the unit at our Montreal bench, ship it to your office, and complete the setup with you over a call. You receive a machine that has been run through its paces before it leaves us, not assembled on your desk.

Remote deployment is not a lesser version of our service — it is how we handle all engagements outside Greater Montreal, and the outcome is identical: a running local AI stack, on hardware sized for your team, with your model loaded and tuned before handover. The full tiers and engagement process, including what on-site visits to the capital look like for larger builds, are on the AI Sovereignty Consulting page.

Who we serve in Quebec City and Capitale-Nationale

  • Government contractors and professional services firms. Law firms, engineering consultancies, and IT services organizations that hold government contracts and need a clean, auditable data-handling posture for personal information processed in the course of that work.
  • Notarial and legal practices. Notaires and lawyers in Quebec City operate under strict professional confidentiality obligations. Using cloud AI to draft acts, review contracts, or summarize client files creates a data-handling exposure their professional codes were not designed to contemplate. On-premise AI handles the same drafting and review work without the exposure.
  • Healthcare and social services. The CHU de Québec-Université Montreal network and affiliated community health organizations handle highly sensitive personal health information subject to both Law 25 and the Act respecting health and social services. On-premise AI allows clinical documentation and administrative support without personal health information leaving the facility.
  • Technology companies and scale-ups. Quebec City has a growing technology sector — Coveo, Absolunet, and a cluster of software firms and agencies. Organizations that want to embed AI in their products or operations without creating a dependency on a cloud provider’s API terms are natural clients for on-premise consulting.
  • Université Montreal and affiliated research units. Research teams that work with participant data, confidential research protocols, or sensitive datasets — and want AI assistance without processing that data on third-party infrastructure — benefit from isolated local deployments.
  • SMBs and professional practices in the broader region. Charlevoix, Chaudière-Appalaches, and the surrounding région administrative. We can ship hardware to any address in the region; setup follows the same remote-deploy process.

Hardware sizing for a Quebec City team

The right hardware depends entirely on your team size, the model class you need, and the workload pattern. The table below is a planning reference — not a quote. Verify current model availability, VRAM requirements, and compatibility at source (as of mid-2026; the model landscape moves quickly). All pricing is quote-only.

Use case Typical VRAM needed Example GPU (representative, not exhaustive) Model class it supports
Solo professional or small notarial practice (1–3 users) 16 GB RTX 4070 Ti, RTX 3090 (16/24 GB) 7B–13B at Q4 quantization; fast inference for drafting and summarization
Small office or professional practice (4–15 users, shared server) 24 GB RTX 4090, RTX 3090 (24 GB) 13B–34B at Q4; handles concurrent sessions for a team
Department or mid-size firm (15–50 users, with RAG over document library) 48 GB+ Dual RTX 3090, NVIDIA RTX 6000 Ada, RTX A6000 70B at Q4/Q5; private RAG over large document sets; French-language models
Government contractor or agency (50+ users, high availability) 80 GB+ NVIDIA A100 80 GB, H100 (if available), multi-GPU rack 70B+ in full precision or near-full; specialized domain fine-tunes; multi-model deployment

All VRAM figures are approximate and depend on the specific model, quantization level, context window, and concurrent-user load. The model landscape changes rapidly — verify current requirements at the respective model repository or runtime documentation before purchasing. We size hardware to your actual workload in a Sovereignty Briefing.

For a complete hardware and model guide, see Local LLM Canada.

Four consulting tiers

All engagements are scoped individually and priced by quote — no published rate card, no off-the-shelf package, nothing pushed toward more hardware than the job requires. Full scope, tier descriptions, and the booking form are on the AI Sovereignty Consulting page. A brief map:

1 · Sovereignty Briefing

A focused session — remote or by video call — where you describe your use case, your data obligations, and your budget. You leave with a written, plain-English recommendation: is local AI right for your situation, which model class fits, and what hardware that implies. Yours to keep whether or not you engage further. For Quebec City clients, this is almost always the right first step before committing to hardware.

2 · Pleb Sovereign Build

For the individual. We size the hardware to your budget, build and test the unit at our Montreal bench, ship it, and walk you through setup. The goal is a machine you own, understand, and can maintain without depending on us. Well-suited to the sovereign professional, the privacy-conscious researcher, or the technologist who wants to stop paying per-month for a capability they can run at home.

3 · Business On-Prem AI

For the Quebec City office or professional practice. A full needs assessment, hardware sizing and sourcing, bench build and test in Montreal, remote setup with your team, private RAG over your document library, a staff-facing interface in French or English, and a written operator runbook so your team runs it without depending on us. Designed with Law 25 in mind and structured to withstand a supplier audit. The tier most directly relevant to notarial, legal, and professional practices in the capital.

4 · Hashcenter Sovereign Compute

For government agencies, large enterprises, research institutions, or organizations with serious power and cooling infrastructure. Multi-GPU or high-VRAM design, power and cooling engineering, control-plane independence via mesh and Nostr fallback where it matters, and optional colocation at a D-Central Quebec facility. See Distributed Compute for the broader architecture.

Law 25 in Quebec City — what the proximity of the CAI means

The Commission d’accès à l’information operates out of offices in Quebec City and Montreal. It has the power to investigate breaches on its own initiative, not just on complaint. For organizations that process personal information in the capital — particularly those that interact with the public sector — awareness of the CAI as a proximity risk, not a distant regulator, is reasonable.

Law 25’s cross-border transfer provisions apply to every personal-information transfer to a system outside Quebec. A cloud AI tool that processes a document containing a client’s name, address, or health information on US infrastructure is executing exactly the kind of transfer the Act anticipates. The assessment and contractual protections required before that transfer is legally defensible fall on the organization sending the data — not the platform.

On-premise AI avoids the transfer entirely. The model runs locally; the data never leaves the server. That is the strongest technical posture available under Law 25’s framework, and it is why legal and professional practices in Quebec City — where personal information is the core work product — are the most natural early adopters. Full guide at Quebec Law 25 and on-premise LLM. Confirm specifics with qualified counsel.

Where the cloud still wins — an honest answer

We would rather lose a client to a clear answer than sell them a system they do not need. The cloud AI tools have genuine advantages that local deployments do not fully match as of 2026:

  • Absolute frontier capability. The most capable models available today run on cloud infrastructure. For research requiring the absolute edge of reasoning or long-context analysis, cloud tools currently hold an advantage — and that gap narrows with each model generation but has not closed.
  • Zero infrastructure overhead. If you handle no personal information, have no data-residency obligation, and want to move immediately, a cloud subscription has lower friction than a hardware deployment. We will tell you that plainly.
  • Specialized multi-modal tasks. Image analysis, video understanding, and real-time speech at scale require dedicated hardware that a small on-premise deployment may not justify. Cloud tools cover these more flexibly at the current moment.

For the work most Quebec City professional practices and organizations actually do — drafting in French and English, summarizing meetings and documents, reviewing contracts for context, answering questions over a private document library, classifying correspondence — well-chosen local models are already highly capable. Our honest comparison is at CLOUD Act and Canadian AI and Local LLM Canada.

French-language model support

Open-weight models with strong French-language support have improved substantially. Models built by Mistral AI (based in Paris), Meta AI’s Llama family, and the Qwen project (Alibaba Cloud research, MIT-licensed) include multilingual training data and perform well on French drafting and comprehension tasks as of 2026 — though English capability remains stronger across the field. For Quebec City organizations where French-first operations are a legal and cultural requirement, we configure the interface, system prompts, and default behaviour in French and test performance on representative work samples before handover. Verify current model French-language benchmarks at source before committing to a specific model family.

Related resources

Frequently asked questions

Do you provide on-site service in Quebec City?

Our facility is in Montreal; most Quebec City engagements use our remote-deploy model — we build and test the hardware at our bench, ship it to your office, and complete setup over a call. On-site visits to Quebec City are available for larger enterprise engagements where in-person assessment and installation are warranted. We discuss this during the Sovereignty Briefing so you know exactly what to expect before committing.

What is the difference between this page and /ai-consulting-quebec/?

The Quebec page covers our province-wide offering — a high-level overview of who we serve and how, for anyone searching across the province. This page is for organizations specifically in Quebec City or the Capitale-Nationale region who want to understand how remote deployment works for their context, the CAI proximity, the French-first angle, and the government/professional-practice landscape of the capital. Different city, different intent.

What is the Commission d’accès à l’information and why does its location matter?

The CAI is the Quebec independent body responsible for overseeing access to public documents and the protection of personal information in the private sector. It administers and enforces Law 25, with the authority to investigate organizations, issue orders, and impose administrative monetary penalties. Its main office is in Quebec City. For organizations operating in the capital — especially those with government contracts or high-profile professional practices — awareness of the CAI as a local regulator rather than a distant one is practical risk-management.

Does on-premise AI satisfy Quebec Law 25 for our firm?

A fully local deployment — where personal information never leaves your premises and no inference traffic passes through a third party — eliminates the cross-border transfer and third-party-access exposures that Law 25 is designed to govern. That is the strongest technical posture available. Whether it satisfies your complete compliance picture depends on your full set of data practices, your sector’s specific obligations, and qualified legal interpretation. This is technical orientation; confirm with Quebec counsel.

Can you configure the AI interface in French?

Yes. We configure the interface language, system prompts, and default behaviour in French for Quebec City clients where that is the operational requirement. The open-source runtimes we install — built and maintained by their respective communities — support French-language configuration. We test French performance on representative tasks before handover and give you an honest assessment of where the model performs well and where it has limits.

Do you work with government contractors or suppliers to the provincial government?

Yes. A documented on-premise AI posture — no cross-border inference traffic, no third-party data processing, local hardware you own — is a defensible answer in a supplier data-handling questionnaire, a Law 25 audit, or a government client’s due-diligence process. We can help you articulate the architecture for that purpose and document it in terms that make sense to a procurement or privacy review.

What models work well for French-language Quebec legal or notarial work?

As of mid-2026, models from Mistral AI’s families and the Llama and Qwen lineages show strong French-language capability for drafting and document review tasks. The right choice depends on your VRAM budget, the workload type, and current benchmark data — which changes with each model release. We assess this in a Sovereignty Briefing using your actual sample documents. Verify current French-language benchmark results at source before committing to a specific model; we do not publish static recommendations because the model landscape changes faster than web pages get updated.

Is there a minimum engagement size?

No. The Sovereignty Briefing is designed to be genuinely useful for any organization, regardless of size or budget — whether you are a solo notary, a government department, or an SMB. You leave with a written recommendation you keep whether or not you proceed. No minimum spend, no sales funnel obligation.

Can you help an organization outside Quebec City that is in the Capitale-Nationale region?

Yes. Our remote-deploy model covers any address in Quebec — Charlevoix, Chaudière-Appalaches, Portneuf, or anywhere in the broader region. We ship the hardware; setup is completed with you remotely. Same process, same outcome.

Book a Sovereignty Briefing for your Quebec City organization

Tell us what your organization does, what you would like AI to help with, and what your data-handling situation looks like — Law 25 obligations, professional confidentiality, government contracts, or simply a desire to stop depending on a foreign platform. Mention whether you want French-first configuration, remote deploy, or an on-site assessment for a larger build. We reply on business days, in French or English, and work from there. No newsletter, no list — this message reaches our team directly.

Prefer a direct line? Call +1 855-753-9997 or email support@d-central.tech. We are at 1325 Rue Bergar, Laval QC H7L 4Z7 — Monday to Friday, in French and English.

Related products, repair, and setup paths

Last reviewed July 22, 2026.