Following up on a piece from 6 months ago, I thought it was timely to jot down some thoughts about sovereign AI factories.

In the digital age, compute is national power. As artificial intelligence shifts from software experiment to core utility, a new model is emerging: the sovereign AI factory. These domestically governed, GPU-rich facilities are vertically integrated platforms designed to train, produce, and deploy AI models at national scale—giving nations direct control over the economic, cultural, and security implications of AI.

Calling these facilities “factories” is deliberate. Unlike traditional passive data centres, they operate as active production lines for foundation models, autonomous agents, synthetic datasets, and safety frameworks—closer in strategic function to semiconductor fabs. Globally, governments are recognizing AI compute as a sovereign capability, aligning digital policy and alliances around domestic compute power and standards.

For Canada, the case is particularly strong for sovereign AI factories. While Canada possesses world-leading AI research talent, domestic compute capacity remains severely bottlenecked. This forces Canadian innovators and enterprises to rely on foreign hyperscalers, leaving pricing, availability, and regulatory oversight in foreign hands. Furthermore, sensitive datasets (such as healthcare, justice, and Indigenous knowledge, etc.) are often processed under external legal jurisdictions. A sovereign AI factory reverses this dependency, transforming Canada into a primary producer of intelligence rather than merely a consumer.

Beyond sovereign control, domestic compute accelerates productivity across manufacturing, resource management, and technology. It ensures critical data remains protected under Canadian legal safeguards while enabling secure, trusted deployments for the public sector. Crucially, it also protects cultural nuance: models trained predominantly on global English data risk erasing local context. Sovereign factories allow Canada to build systems tailored to French-language requirements, Indigenous language preservation, and distinct domestic values. Canadian identity by design, eh?

Canadian telecom operators are uniquely positioned to anchor this transition. Telecom carriers already manage the ultra-low-latency, highly secure digital backbone required to interconnect distributed compute nodes. As AI transforms network operations, spectrum management, and customer service, operators need trusted domestic environments to train and execute models safely.

Telecom policy has long promoted investment for Canadians to connect and communicate securely. It makes sense for telecom operators to extend that role to include sovereign compute capacity, as a logical next step to secure Canada’s national digital strategy.

Is the CRTC now regulating devices beyond reason? Have we reached a peak ‘nanny state’ state approach to telecommunications despite vibrant competition in that sector?

About 3 years ago, I wrote “Regulators regulate”, saying that it was a corollary to Maslow’s Hammer: “If the only tool you have is a hammer, it is tempting to treat everything as if it were a nail.”

How else can we explain some of the recent regulatory over-reach as part of the Commission’s implementation of legislative changes to the Telecom Act?

In early July, I wrote about cancellation fees over-reach, but there is another case that I find equally troubling. From the outset, let me say that I don’t think the CRTC should be regulating the way mobile devices are sold. The CRTC regulates mobile phone sales by service providers, but not sales by the manufacturers or by independent retailers such as Amazon. Financing plans are regulated if the financing is by the service provider, but not if the financing is provided by a bank card, even though the service providers are often providing zero-per cent financing versus annual financing charges by credit cards of 18-20% or more. These imbalances on their own should raise questions.

Today’s post was stimulated by a series of letters dealing with the CRTC intervening on robbery and loss-prevention strategies for mobile service provider retail outlets.

In April of 2025, Bell wrote to the CRTC to inform the Commission that the company was experiencing an increase in crime at its stores associated with the theft of high-end mobile devices. Given that smartphone prices are now running in the thousands of dollars, Bell said that initially locking the devices and then automatically unlocking them after the first 60 days would serve as a disincentive for thieves. In November, Commission staff told Bell that the company had not made its case. “Bell has not demonstrated that the practice of locking cellphones for up to 60 days after purchase is a necessary and proportionate response in this case. Specifically, Bell has not demonstrated that this practice is effective.”

In May, TELUS filed an application for authorization of the 60-day device locking to be formally approved on a temporary and permanent basis. As it explained in its application:

Canadian mobile devices were essentially transformed into the digital equivalent of [bearer] instruments, because of the following defining traits:

• Resale Liquidity: an unlocked iPhone 17 Pro Max or Samsung Galaxy S25 Ultra, retailing for amounts up to $3,000 CAD,23 functions on any GSM/LTE/5G network globally the moment it leaves the store.
• Anonymity of Transfer: unlike vehicles or real estate, which require registration upon transfer, an unlocked phone requires no handshake to transfer ownership. It can be sold for cash or crypto, leaving no auditable trail.
• High Value-to-Weight Ratio: a single backpack can hold devices worth tens of thousands of dollars, making them attractive targets for theft and trafficking across borders.

The prohibition on locking removed the only technical deterrent in the illicit device resale market.

By way of a letter, the CRTC asked the public to respond to respond for the temporary relief and suggested that the request for a permanent change be made as part of the CRTC’s public consultation on “Harmonizing the consumer protection codes”.

In its “Anticipated releases for the week of 3 to 7 August 2026”, the CRTC said it planned to issue a notice of consultation for a “Show cause and call for comments –Compliance with the prohibition on selling locked devices and other matters”. That was not released as planned.

In the meantime, Bell’s EVP and Chief Regulatory Officer had a post on LinkedIn that provided answers to some of the CRTC staff concerns from last November. Bell’s locking proposal was in response to a 500% increase in robberies and attempted robberies at its retail stores. In response, Bell “introduced a targeted safety practice: devices are automatically unlocked after 60 days and can be unlocked sooner upon request at no cost to customers.”

Since then, in-store robberies have been nearly eliminated, fraud targeting customer accounts has decreased by 80%, and thefts from warehouses and shipments have fallen by more than 35%. Most importantly, this practice is helping keep our employees, customers and communities safe.

As to the CRTC staff concern about proportionality, Bell says fewer than 1% of customers choose to switch providers within the first 60 days of acquiring a new device.

I don’t understand why the CRTC is regulating device pricing and financing at all. The CRTC doesn’t regulate hardware retailers, such as Costco, Best Buy or the Apple or Samsung stores selling the same devices. There is no regulation of credit card companies financing these same devices.

In its consultation for “Harmonizing the consumer protection codes”, the CRTC says 30% of consumers are “renting” their mobile devices with plans that allow customers to trade-in or return their device to the service provider at the end of their contract. The Consultation says “customers may not always be fully aware that they opted into a device rental plan, and that they may have to pay a balance to keep their device after their contract ends.”

Really? Isn’t this another case of regulating beyond reason? Can we treat consumers like adults? Do consumers who lease their cars not realize that they have a balance owing at the end of the lease?

I see ads for free high-end devices from US carriers, such as T-Mobile, Verizon, or AT&T, tied in to 2 or 3 year commitments on higher service plans.

I wax nostalgically. Canada used to have free phones available until the Wireless Code came around. I referred to the cost of regulation back in 2013. The increased cost of devices was seen as a short term cost in order to gain the longer term benefits of competition. An August 6 report from Scotiabank characterizes Canada’s mobile sector as a “competitive four player market”.

Right now, the way the CRTC is regulating devices has resulted in higher costs for consumers and carriers alike. Perhaps it is time for the CRTC to get out of the business of hardware regulation.

The next webinar from the International Telecommunications Society features Harvard Professor Cass R. Sunstein, with a presentation entitled “Imperfect Oracle: What AI Can and Cannot Do”. It will take place September 22, at 9:30 am (Eastern).

The promotional email caught my eye:

Human judgment is extraordinary, but it is also flawed. People are biased: our judgments can go systematically wrong in predictable ways. People are also noisy: our judgments can vary depending on mood, timing, context, pressure, or the decision-maker involved. Bias and noise can produce serious mistakes in government, business, law, health, finance, regulation, and everyday life.

AI offers a powerful response to both problems. It can identify patterns, reduce inconsistency, improve prediction, and support better decisions. For institutions trying to avoid errors, allocate resources, assess risks, serve citizens, or make complex judgments at scale, AI may become an extraordinary tool.

But AI is not an oracle. The world remains full of surprises. AI can help us know more, but it also forces us to understand what we cannot know.

This webinar will explore the practical and policy implications of that insight.

Professor Sunstein is currently the Robert Walmsley University Professor at Harvard. He is the founder and director of the Program on Behavioural Economics and Public Policy at Harvard Law School. In 2018, he received the Holberg Prize from the government of Norway, sometimes described as the equivalent of the Nobel Prize for law and the humanities. In 2020, the World Health Organization appointed him as Chair of its technical advisory group on Behavioural Insights and Sciences for Health.

There is no charge for the webinar. Register now to reserve your place.

Is the CRTC intentionally sabotaging its own prohibition on fees?

Before the Canada Day holiday, the Commission launched a consultation, “Show cause and call for comments – Compliance with the prohibition of fees that are a barrier to switching cellphone and Internet plans”.

I’ve already provided my opinion on the CRTC’s prohibition. No carrier charges activation fees that are a barrier to switching. What rational business would? Businesses are always trying to win over customers.

Now, if a service provider charged a fee when you want to terminate a subscription, that could be considered a barrier to switching. But the CRTC already has rules about early termination fees.

Take a look at the CRTC’s Notice of Consultation and you will see how the CRTC is setting itself up for ridicule. The Commission asked one of the carriers if it had ceased billing customers $25 for shipping a device ordered online. A shipping charge?

A recent article by Canadian Press highlights some of the mixed-up thinking at the Commission. A CRTC executive seemed to argue for greater transparency in the prices paid by consumers, while opposing shipping fees for precisely those kinds of charges. Wouldn’t such discrete costs provide such transparency? “If you have to increase your prices so be it, but do that through the front door. Charge a price, don’t surprise consumers with price increases in the middle of the contract, don’t have these special little fees that come out of nowhere.”

Isn’t that the approach being taken by the carriers? If you want a device shipped to you, is a shipping charge unreasonable? If you lose or destroy your SIM card, who should pay for that?

In its March Regulatory Policy (that took effect June 12), the CRTC observed that recent amendments to the Telecommunications Act broadly prohibit activation and modification fees, while recognizing “that prohibiting fees related to installation services at a customer’s premises could have a negative effect on future broadband Internet rollout because those installation services represent actual, necessary, and sometimes significant costs.” The CRTC also said “fees related to optional services and products do not fall under the category of “activation or modification fees” related to the telecommunications service itself.”

So the Policy set out a new definition, stating:

Activation or modification fee

Any fee incurred as a result of activating a new retail telecommunications service plan or modifying an existing one, except for reasonable fees related to the physical installation of a telecommunications service at a customer’s premises or fees related to additional products or services the customer has explicitly chosen to purchase.

If the rules need further clarification, then the CRTC should fix its self-inflicted ambiguity. Alternatively, it would be a good time for the Commission to back down and let the marketplace work.

Last week, The Hill Times published a Policy Briefing supplement looking at Rural & Remote Broadband.

I was asked to prepare an Op-Ed for that supplement.

Regular readers will notice that it was largely based on a piece I published last month.

Canada’s national broadband objective is defined as having high-speed (50 Mbps down / 10 Mbps up) connectivity available to all Canadians by the year 2030. I wonder if it may be time for us to declare victory and move on to setting a new objective: increasing adoption among those who still aren’t connected.

Various broadband funding programs (such as the Universal Broadband Fund – UBF, Connect to Innovate, provincial initiatives, the CRTC’s Broadband Fund, etc.) have collectively pushed high‑speed connectivity deeper into rural and remote regions than ever before. Fibre builds now reach thousands of communities once considered uneconomic, and latest generation wireless services have filled many mid‑density gaps. Yet despite billions of dollars of investment, a stubborn last 1–2% of households remain unserved, particularly in the North and in the most sparsely populated rural pockets.

This is where Low Earth Orbit (LEO) satellite networks should be added to the broadband connectivity toolkit. Indeed, we might consider whether direct satellite-to-device is a satisfactory mobile solution for those remote communities currently lacking terrestrial-based coverage.

LEO systems operate a few hundred kilometres above Earth, far closer than traditional geostationary satellites. This enables low‑latency, high‑throughput broadband rivalling terrestrial options. Starlink, the most mature LEO provider, already offers:

• High‑speed service with typical download speeds ranging from 45–280 Mbps.
• Low latency (25–60 ms), suitable for video calls, cloud apps, and real‑time services.
• Global availability, including remote and northern regions.

Other LEO constellations are literally on the horizon. Why isn’t LEO considered to be an obvious tool to fulfil Canada’s broadband ambition? For households beyond the economic reach of fibre or microwave backhaul, LEO solutions eliminate the need for towers, rights‑of‑way, or construction seasons. A dish, a clear view of the sky, and power are enough to provide connectivity.

Based on publicly available coverage maps and service availability data, existing LEO broadband constellations cover all populated regions of Canada.

Where availability issues arise, they are typically due to temporary local capacity constraints, obstructions due to trees, terrain, or building orientation, or weather‑related installation challenges. These are all easily solvable problems, at a cost far less than the $10-20,000 (and more) per household being spent for terrestrial solutions in some communities. Four years ago, the government contributed more than $46.6 million to connect 182 households in northern Ontario, more than $250,000 per household for broadband in an area where houses sell for less than that.

LEO solutions provide full national orbital coverage and can close the final connectivity gap quickly, affordably, and sustainably. One might say that we have walked the last mile of last mile connectivity.

Using LEO, we could (but shouldn’t) provide a permanent subsidy to equalize the prices paid by rural subscribers to those being paid in urban centres. We need to think carefully about subsidies for rural broadband broadband expansion. Subsidies should be based on financial need, not based on geography. There are people in urban centres who need lower cost everything, just as there are people in rural and remote communities who do not need financial aid. For example, a little over a year ago, I observed “Median household incomes in the north are considerably higher than in the rest of Canada.”

With technology now offering a reasonable option for broadband connectivity Canada’s broadband strategy needs to focus on getting the remaining unserved households to get online. This is no longer an engineering challenge that can be solved with money, but one of understanding the factors that inhibit increased adoption in both rural and urban settings.

Programs such as Internet for Good from TELUS, and Connected for Success from Rogers, and the national Connecting Families initiative have made broadband even more affordable for many disadvantaged households, fully funded by Canada’s telecommunications industry. But, we have also learned that there are issues beyond affordability inhibiting some people from connecting.

Integrating LEO into regulatory and policy frameworks, while preserving private sector investment incentives, will allow us to declare victory in meeting Canada’s national broadband objective. It is time to engage partnerships between service providers, government social service agencies, and training facilities to drive adoption, ensuring no Canadian household is left offline.