What If Ethical AI Isn't Actually Ethical?

As Canada accelerates investment in AI data centres, the conversation must expand beyond algorithms to include the energy, water, and communities that power artificial intelligence. 

When we talk about ethical AI, the conversation usually focuses on privacy, bias, transparency, accountability and human oversight.

Those are important conversations.

But at PredictEdge, we've been asking another question.

What about the infrastructure that powers AI?

Every AI model depends on data centres operating around the clock. They consume electricity, generate heat and, depending on their design, can require significant amounts of water. As AI adoption accelerates, so does the demand for computing infrastructure.

For years, much of the debate around the environmental impact of data centres has been happening in the United States. Communities have questioned electricity demand, water use and whether the benefits outweigh the costs.

Canada can no longer treat this as someone else's conversation.

Over the past year, we've seen significant investments in Canadian AI infrastructure. Bell's AI Fabric is building sovereign AI compute capacity across the country. TELUS is expanding its Canadian AI facilities, powered largely by renewable energy and designed with waste-heat recovery and advanced cooling. Meta has announced a multibillion-dollar AI data centre in Alberta, highlighting both the opportunity and the enormous scale of infrastructure required to power next-generation AI.

Canada needs AI infrastructure. 

It strengthens our digital sovereignty, supports research, creates jobs and reduces reliance on foreign computing resources. The question isn't whether we should build data centres.

The question is how we build them.

The United Nations Environment Programme (UNEP) argues that sustainability should be treated as a foundation of digital infrastructure, not a constraint on innovation. That means improving energy efficiency, reducing water consumption, adopting smarter cooling technologies, integrating renewable energy and increasing transparency around environmental performance.

Around the world, leading data centres are already demonstrating that sustainability and performance can go hand in hand. Renewable energy, waste heat recovery, innovative cooling systems and smarter energy management are proving that responsible infrastructure is achievable.

But voluntary commitments alone aren't enough.

Without clear expectations, the race to build AI infrastructure could become a race to build it as quickly and cheaply as possible. Short-term economic gains should never come at the expense of long-term environmental stewardship.

The current wave of investment gives us an opportunity to establish those expectations now, while much of the infrastructure is still being planned. 

That's where governments, businesses and users all have a role to play.

Governments should establish clear standards for sustainable AI infrastructure.

Businesses should look beyond quarterly returns and consider the lifetime environmental impact of their investments.

That includes engaging municipalities, local residents and Indigenous communities early and meaningfully, not because consultation is a regulatory hurdle, but because these communities are partners in shaping the future of Canada's digital infrastructure. Projects that are built with communities, rather than simply within them, are more likely to earn trust, address local concerns and deliver lasting economic and social benefits.

These are considerations that we find ourselves asking at PredictEdge as we evaluate data centre partners.

• Where is the infrastructure located?

• How is it powered?

• How much water does it consume?

• Does it recover waste heat?

• Does it strengthen or strain the local electricity grid?

• How resilient is it to future climate and energy challenges?

• Has the project meaningfully engaged municipalities and Indigenous communities?

• Does it create lasting value for the communities where it operates?

• Is the operator transparent about its environmental impact?

Organizations such as Ceres have argued that data centre development should be fair, affordable, efficient and clean. 

We would add one more word: Responsible.

We don't need less AI infrastructure. What we need is better AI infrastructure.

This isn't an argument against innovation. But we must acknowledge that there is a moral tension at the centre of AI development.

We are building tools that promise to help solve some of society’s most complex problems. AI could improve medical research, strengthen public services, reduce waste, manage energy more effectively and help communities prepare for climate-related risks.

Yet the infrastructure supporting these tools may also increase emissions, strain electrical systems, draw on local water supplies and shift infrastructure costs onto the public.

The contradiction does not mean we should stop building AI. It means we have a responsibility to build it differently.

It's an argument for responsible innovation.

Canada has a unique opportunity to become a global leader in ethical AI, not just because of the models we develop, but because of the standards we set for the infrastructure that powers them.

At PredictEdge, we've always believed that ethical AI is more than responsible algorithms.

It includes responsible data.

Responsible governance.

Responsible deployment.

And increasingly, responsible infrastructure.

Ethics cannot end at the software layer.

It must extend to the electricity powering the servers, the water used to cool them, the communities hosting them and the public systems supporting their development.

The future of AI will not be judged solely by the intelligence it creates.

It will also be judged by the footprint it leaves behind and whether we had the courage to demand better foundations while there was still time to shape them.

The question is no longer whether Canada will build AI data centres.

We are already building them.

The question is whether governments, businesses and users will insist that what we build advances innovation without adversely affecting the world it is meant to improve.