OTTAWA — Anthropic, a company that has warned of the potentially catastrophic consequences of artificial intelligence, has been named as the AI superpower behind a massive data centre project in Nova Scotia.
The company’s involvement was quietly announced early Friday evening in a statement from Anthropic’s head of public policy in Canada, Jacob Glick.
The deal ties one of Canada’s largest-ever foreign AI infrastructure investments to a company whose executives have simultaneously warned about the risks posed by increasingly powerful AI systems while pursuing the computing capacity needed to build them.
“Canadians are using Claude in growing numbers, and we want our infrastructure to be closer to the people who rely on it. Canada has set a high bar for responsible data centre development, as has the premier of Nova Scotia, and we are committing to meeting it,” Glick’s statement read.
“We will work together to achieve shared goals and earn our place as a good neighbour,” he wrote.
Nova Scotia energy company EverWind Fuels will own and develop the facility, which it billed as the largest private investment in the province’s history. The company is putting more than $10 billion into the project, with more than $8 billion going toward a data centre campus in Cape Breton and $2 billion to expand EverWind’s onshore wind portfolio.
Anthropic has signed onto a 20-year lease for the first phase of the project, which has an initial capacity of 95 megawatts and is set to be operational by late 2027, provided the proposal meets regulatory, environmental and permitting conditions. A 325-megawatt expansion is planned for mid-2028, with full capacity hitting 420 megawatts later that year.
The site will be used to power the large language models behind Claude, the chatbot for which Canadian users represent 2.6 per cent of the AI assistant’s global traffic. Canada’s “usage per capita is more than four times higher than would be expected given the size of its population,” the company has said.
Earlier in the day, EverWind officials, providing a technical briefing to media on a background basis, repeatedly dodged questions from reporters about the identity of the company, how it planned to use the facility and the total value of the lease.
The project is expected to adhere to the federal government’s Responsible Data Centre Development Principles, which Anthropic and other AI giants signed on to last month. The principles, which are voluntary, ensure that data centres benefit the communities in which they’re located, don’t offload electricity costs to residents, minimize environmental impacts, and boost Canadian security and compute capacity.
Sean Fraser, a Nova Scotia cabinet minister, was asked Friday about what the announcement portended for Canadian data sovereignty due to the U.S.-based company’s involvement.
“AI is a part of the world today and it’s going to be an increasing part of the world tomorrow,” Fraser said.
“My own view is that we need to recognize that as that change is happening, there’s very real concerns that need to be accommodated. This to me says, let’s put rules in place that are designed to protect Canadians and recognize that if we don’t welcome investment opportunities with new technology, including but not limited to AI, then the technology is going to be based elsewhere.”
The minister also said building the project on home turf means American companies could be forced to abide by Canadian rules, not foreign ones.
Friday’s announcement comes several months after Meta unveiled a $13-billion plan to build a one-gigawatt AI data centre north of Edmonton.
Artificial Intelligence Minister Evan Solomon has said that because data centres fall under provincial and municipal jurisdiction, it is not up to Ottawa to set legally-binding standards for such facilities.
Nova Scotia Premier Tim Houston said Friday there will be “a number of checks and balances” tied to the project’s development and operation.
“We need more compute power in this country. The data we use every day, in our health care system, for all of our banking needs, everything on our phones, businesses run on this data. Government runs on it,” Houston said. “We need more secure infrastructure, more secure data infrastructure. This is part of that puzzle. This is really important for Nova Scotia, but also for our country.”
In early September, Anthropic found itself swept up in concerns that advanced forms of its technology could wipe out human life in a matter of years. Jacob Coxon — a researcher who had also worked for OpenAI — announced he was leaving Anthropic because he believed neither company was “acting responsibly” and that they were “gambling with our lives.”
Coxon’s departure sparked a wave of admissions from AI leaders, including Anthropic’s Dario Amodei, that the technology could become unbridled, misused for dangerous purposes and spur significant economic disruption.
Amodei told the United Nations Security Council last month that he believed AI models may pose “a risk to humanity as a whole” and called for a global effort to slow down the pace at which models are released.
Solomon has said AI companies should not set the terms for their own governance, while Prime Minister Mark Carney has floated the idea of a “technology stability board” to regulate AI deployment.
Solomon confirmed earlier this year that Canada had been given access to Anthropic’s powerful Claude Mythos model for cybersecurity purposes. The company also invested $10 million into Canadian research institutions in July.
