Strategic position
chels.ai is the sovereign AI infrastructure partner for regulated enterprises in Canada and the US. Law firms, healthcare networks, financial institutions, and defence suppliers need frontier-class AI that never leaves their jurisdiction — and almost none of them can build it alone. We design, build, and operate it for them, and we aggregate their demand into capacity sited where power actually exists.
Corporate information — not an offer of securities. See the notice at the foot of this page.
The thesis
Six observations, each attributed to the third-party analysts and public records that made them. Together they describe the gap chels.ai occupies.
Analysts at the Futurum Group estimate combined hyperscaler AI capital expenditure at roughly US$630–690 billion in 2026. Yet in its public earnings commentary, Microsoft has pointed to a backlog of roughly US$80 billion in Azure demand it attributes to one constraint: the availability of power. In the analysts’ framing, the AI buildout is no longer gated by chips or customers — it is gated by megawatts.
Researchers at Lawrence Berkeley National Laboratory put the US interconnection queue at roughly 2,600 GW of projects waiting to connect, with data-center power waits of 4–7 years reported in Northern Virginia, Phoenix, and Dallas. Goldman Sachs analysts estimate a US data-center power shortfall of 11+ GW today; Morgan Stanley projects it reaching roughly 49 GW by 2028.
Canada’s grid is roughly 85% non-emitting (public federal energy data), with cold-climate cooling advantages that US Sun Belt markets cannot match. The Government of Alberta has set a target of C$100 billion in data-centre investment, and publicly announced pipelines across Ontario and Alberta total roughly 26 GW of planned capacity. For Canadian regulated enterprises, siting AI capacity in Canada is not just an energy decision — it is a compliance posture.
Innovation, Science and Economic Development Canada (ISED) operates the C$2 billion Sovereign AI Compute Strategy, including a C$700 million commercial AI Compute Challenge and the C$300 million AI Compute Access Fund — a demand-side program that subsidizes eligible Canadian organizations’ purchases of sovereign compute. That funding flows to customers — the buyers of exactly the kind of capacity chels.ai deploys — not to chels.ai. The pattern is global: the EU’s InvestAI program targets €200 billion, and Gulf sovereign programs are building national AI capacity on the same premise.
In its FY2026 results, NVIDIA reported sovereign-AI revenue that roughly tripled year-over-year to more than US$30 billion — about 14% of company revenue — naming Canada among the sovereign buildouts. McKinsey sizes sovereign AI at US$500–600 billion by 2030 and estimates 30–40% of AI workloads will be sovereignty-influenced. Third-party enterprise surveys point the same direction inside regulated sectors: roughly 59% of LLM deployments in privacy-driven sectors are on-premise, 77% of companies weigh an AI vendor’s country of origin, and 53% rank data privacy as the #1 barrier to AI adoption.
CoreWeave reported US$5.1 billion in 2025 revenue (public filings). Crusoe has been valued above US$10 billion in press reports. TELUS announced its Rimouski sovereign AI facility sold out at opening, and Bell launched its AI Fabric with a Cohere software stack. Those players prove the economics at hyperscale and national-carrier scale. chels.ai is early-stage and deliberately narrower: the regulated mid-market enterprise — the 120-lawyer firm, the clinic network, the credit union — that the giants do not serve directly, reached through an end-to-end practice rather than a raw-capacity product.
The numbers, attributed
Every figure below is a third-party analyst estimate, public filing, or public government program — cited to its source. None is a chels.ai projection.
| Signal | Figure | Source |
|---|---|---|
| Hyperscaler AI capex, 2026 | ~US$630–690B | Futurum Group estimates |
| Azure backlog attributed to power availability | ~US$80B | Microsoft earnings commentary |
| US interconnection queue | ~2,600 GW awaiting connection | Lawrence Berkeley National Laboratory |
| US data-center power shortfall | 11+ GW now · ~49 GW by 2028 | Goldman Sachs · Morgan Stanley |
| Canadian grid, non-emitting share | ~85% | Public federal energy data |
| Ontario + Alberta announced DC pipeline | ~26 GW planned · C$100B Alberta target | Provincial announcements |
| Canada Sovereign AI Compute Strategy | C$2B, incl. C$300M customer-side Compute Access Fund | ISED (Government of Canada) |
| NVIDIA sovereign-AI revenue, FY2026 | US$30B+ · ~3× YoY · ~14% of revenue | NVIDIA FY2026 results |
| Sovereign AI market by 2030 | US$500–600B · 30–40% of workloads sovereignty-influenced | McKinsey |
| On-prem LLM share, privacy-driven sectors | ~59% of deployments | Third-party enterprise surveys |
| Comparable: CoreWeave 2025 revenue | US$5.1B | Public filings |
Information for accredited investors (NI 45-106 in Canada / Regulation D in the United States). All terms and detailed materials are shared privately, through direct conversation — never on this website.
This page is corporate information about Chels.ai and the market in which it operates. It is not an offer to sell, or a solicitation of an offer to buy, any security in any jurisdiction, and nothing on this page constitutes investment, legal, or tax advice or a forecast of Chels.ai’s financial performance.
Market figures on this page are third-party analyst estimates, public filings, and public government-program descriptions, cited to their sources. Chels.ai has not independently verified them, and they are not projections by Chels.ai. References to government programs — including Canada’s Sovereign AI Compute Strategy and the AI Compute Access Fund — describe funding available to eligible purchasers of compute; they are not funding awarded to Chels.ai. References to third-party companies are for market context only and imply no affiliation, partnership, or endorsement.
Any discussion of investment is limited to persons qualifying as accredited investors under National Instrument 45-106 (Canada) or Regulation D under the U.S. Securities Act of 1933, and occurs only through private channels.