As generative AI rapidly finds applications across industries, the data centers that power these models are no longer back-office infrastructure, they are becoming strategic national assets. In tandem, the question of digital sovereignty – not just where data resides but who controls the infrastructure, governance and ecosystems around it – is arising too.

Governments and institutions are increasingly moving toward more localized and secure data environments. But doing so will require enormous capital investment, expanded power generation and the ability to convince communities that hosting this infrastructure delivers meaningful long-term benefits.

Even with hundreds of billions committed to this effort in 2026 alone, there are questions as to whether data center capacity can keep pace with AI demand. That was the focus of the “Digital sovereignty: Scaling data centres for an AI world” panel to close out day two of the Conference de Montréal. The session featured:

• 

  Kristin Milchanowski, Chief AI and Quantum Officer, BMO

• 

  Chris Crosby, CEO, Compass Datacenters

• 

  Simon Ahdoot, CEO, Hypertec

• 

  Jean Boivin, Managing Director and Head, BlackRock Investment Institute

• 

  Martin Durkin, Partner, Holland & Knight (Moderator)

Listen to the conversation (9 mins):

Markets Plus is listed on all major channels, including Apple and Spotify

Here are the highlights:

Jean Boivin said the pipeline of spending for AI data centers is incomparable to any other private sector industry. Global spending through 2030 is estimated in the range of roughly US$5 trillion to US$8 trillion, with forecasts continuing to trend upward as demand accelerates. “This will make it the fastest and biggest capital-expenditure buildout in the history of humanity,” he said.

While such investment is needed, Kristin Milchanowski noted spending is outpacing AI company usage with compute use among global companies remaining low at about three percent. To ensure infrastructure spending is valuable, companies need to move beyond automation and embed AI into every decision that shapes the business, she explained. “That’s where you’re going to see a direct tie into ROI,” she said.

Chris Crosby described data centers as long-term infrastructure rather than temporary facilities built around a single generation of technology. They sit in the same category as railroads, airports and water treatment plants, he said, built to last rather than to flip or sell. He drew a parallel to electricity, which was initially viewed as having a narrow application before becoming foundational. “This is the greatest enabling technology since electricity,” he said.

Milchanowski noted that investors and lenders have grown more selective about selecting data center projects, with those attached to more speculative profiles, coming from the crypto world, subject to a higher bar. The challenge right now is that some jurisdictions have unclear or inconsistent policies that could disrupt infrastructure build outs.

In some instances, there have been zoning reversals and shifting energy or infrastructure rules. “Capital craves certainty,” she said. “Capital is going to follow the predictability of things, and I do fear that there’s some unpredictability around the policy trajectory.”

For the largest developers, Crosby said that policy risk is playing a larger role in where projects get built. Compass recently walked away from a large-scale Northern Virginia data center project, part of a broader development valued in the tens of billions after a court overturned its zoning approval on a technicality.”

He said projects that take years to generate a return become harder to justify when approvals can change after money has already been committed. “We just leave; we don’t go to those places,” he said.

Milchanowski said the data sovereignty issue goes beyond location. A data center can house sensitive information, but institutions still need to govern access, encryption and the decisions made with that data. Data sovereignty must be built into the infrastructure from the start. “Sovereignty isn’t just the data center; it’s the moat around it,” she said.

An important part of that moat is ensuring data centers are protected from cyberattacks. As Simon Ahdoot pointed out, countries need to expand their cyber security capabilities. In his eyes, this type of investment will ultimately help speed the development of AI rather than act as a speed bump. “If you look at the fastest cars, they have the best brakes; they have the best security systems,” he said. “If we want to go as fast as possible, we have to have the right security measures in place.”

Powering these data centers and winning over communities is another challenge to preserving data sovereignty. Crosby said access to power is already narrowing the list of places where AI data centres can be built. The North American power grid has expanded far more slowly than demand over the past few decades, and in many regions, fragmented planning and regulatory barriers make it difficult to invest in large-scale system upgrades.

When asked what could stall the data infrastructure buildout, the panelists said execution. Crosby said developers need to get better at showing communities how projects can help them, especially when local debates focus on bans, water use or power costs. Ahdoot said governments need to start seeing data centers as infrastructure, instead of judging them only by the number of jobs they create. Boivin pointed to the need for permitting reform, especially around energy.

Milchanowski closed on the workforce needed to build it all, arguing that AI infrastructure will depend as much on trades and technical talent as it does on capital and compute. “We’re already short half a million electricians, right this second, let alone with this trajectory we’re on,” she said.