AI Data Center Water Consumption Concerns

Concerns are rising over the substantial water consumption of AI data centers in Canada, mirroring issues seen internationally in places like Spain, Uruguay, and parts of the United States. Citizens in these regions are protesting the construction of such facilities due to water scarcity. Google recently shelved a data center project in Indianapolis following community backlash, as announced last month.

Traditional data centers have historically been very water-intensive, but Alistair Spears, in charge of Microsoft's cloud infrastructure, stated that improvements in technology mean new facilities in Canada will use significantly less water. Microsoft’s data centers in and around Toronto have been cleared by city officials to use around 1 billion litres of water per year. However, Microsoft anticipates using only a fraction of that amount, utilizing large vents for cooling with outside air.

Canada’s cool climate and inexpensive electricity are positioning it as a desirable location for data center development. There are currently over 300 data centers in Canada, primarily located near Toronto and Montreal. Development is now expanding to drier regions, including plans for what will be the world’s largest AI data center in northwest Alberta, projected to be the “lowest cost data center in the world.” Even investors like Kevin O’Leary are expressing interest in the sector.

In Nanaimo, a proposed data center has sparked worry among residents regarding foreign tech companies utilizing local resources. A recent motion regarding the project was debated by city officials. Mayor Leonard Krogh stated that every community is seeking participation in “what we will call the future in terms of the economy” and the jobs created by such facilities are “the jobs of the future, to some extent.”

Catherine Barnwell, a retired English professor, is part of a citizen group opposing the project. She argued that “a few jobs aren’t worth the risk to the environment,” stating, “Life on this planet is sustained by water. It is not sustained by data. You know, we don't need data the way we need water. It's essential.” Barnwell expressed concern that Canada has been “pretty blithe about our natural resources” and is staking a longtime friendship with the mayor on changing his mind.

Currently, there is limited industry oversight regarding water usage in Canada, while officials in both Europe and the United States are attempting to regulate data center water consumption. David Mayer, a professor of municipal water engineering at the University of Toronto, warned that approving data centers reliant on treated city water reduces the water budget available for residential growth. He stated that Canadians living in dry areas “should care a lot about AI water usage” as the water is needed for agriculture and cities. He also noted that existing water pipes, some of which are 100 years old, were not designed to accommodate the demands of AI data centers.

Transparency is also an issue, as many data centers do not disclose their water consumption figures, and cities don’t always track usage. The Amazon data center in Varennes, a Montreal suburb, has been using municipal drinking water since 2018, but the building lacks a water meter, meaning the city does not know how much water is being used. Amazon declined to answer questions about its water use at the facility.

Nathan Wanguzi, formerly of Amazon’s water sustainability team, explained that “water is like blood to the data center industry.” He stated that there is a “tendency to deny or cover that that water consumption exists” within the industry, and particularly at Amazon, which operates dozens of data centers globally. Big tech companies, including Microsoft and Amazon, have pledged to achieve net zero water consumption by 2030, but Wanguzi is skeptical, stating it is “not possible” for them to both build massive AI data centers and reduce water intake. He believes they would need to go bankrupt to achieve this.

Wanguzi emphasized that Canadians “should care about it because that is the water that they should be consuming.” The broadcast concluded by stating that the AI data center boom is coming to Canada, and the question is whether the country has enough resources to meet the demand.

AI in Breast Cancer Detection

New studies indicate that artificial intelligence can detect breast cancer with comparable accuracy to radiologists in certain settings. Results from two trials exceeded expectations, showing the AI tool reduced radiologist workload by 44% and detected 20% more breast cancers in a Swedish trial involving over 80,000 women.

However, researchers found that a radiologist must remain in charge to prevent over-diagnosis and over-treatment, as the AI tool can generate too many false positives. The AI tool cannot currently function as a standalone screening tool. Radiologists can leverage their experience with patient history to rule out abnormalities, a capability the AI currently lacks.

The AI’s data primarily comes from scans of white women, raising concerns about its effectiveness for other demographics. It also struggles to detect cancer in women with dense breast tissue. Toronto researchers are working on AI that checks breast density to predict the chances of hidden cancer. Swedish researchers also noted that some cancers detected by AI may not be deadly and are monitoring patient outcomes to determine if early detection improves survival rates.

AI in Medical Transcription

Artificial intelligence is being used to assist doctors with medical transcription, freeing up time for patient care. Doctors can now use approved apps to record patient interactions, which are then transcribed and summarized by AI. James Tu, an emergency room doctor and co-founder of Plume AI, stated that the app saves him one to two hours of note-taking per day in the emergency room.

Approximately 10% of doctors in Quebec are already utilizing these AI tools. Dr. Félix Le Fat Ho reported a “huge impact” on his clinical practice, allowing him to see over 20 patients a day. Santé-Québec is launching a pilot project to roll out AI on a larger scale for medical transcription.

Concerns regarding accuracy, data security, and doctor oversight are being addressed. While the AI is not perfect and requires review, protocols are in place to ensure doctors verify the generated notes. The province is prioritizing data security and will only approve tools that guarantee it. The legal responsibility for the data remains with the doctors, and the AI is intended as a tool to assist, not replace, their judgment.