Who Makes Money From AI After the Chips Are Sold?

AI spending doesn’t stop with processors as data centres also need electricity, grids, substations, and engineering.

Amy Legate-Wolfe

Author Bio Amy became interested in investing in 2018 after having her first daughter. After receiving a masters degree in journalism from Western University, she became frustrated that the finance industry remained a confusing place for Canadians like her: new parents, millennials, and other young people who needed to understand their finances. Now, Amy focuses on tech companies and renewable energy for growth opportunities, coupling that with long-term investing strategies and equities. She also provides investors with a daily newsletter, Canada Stock Digest, for daily market updates. Her work can also be found at Canada's National Observer, where she writes on the business side of climate action. Before joining Motley Fool Canada, Amy wrote for major news organizations including HuffPost, CTVNews.ca, and CBC. When she’s not researching investing strategies, Amy’s time is pretty much monopolized by her two wild daughters, but in what little spare time she has she loves to do yoga, go on walks with her dog, and travel with her husband. Social Media Follow Amy on LinkedIn and Canada Stock Digest. Favorite Investment Quote “October: This is one of the particularly dangerous months to speculate in stocks. The others are July, January, September, April, November, May, March, June, December, August and February.” – Mark Twain

Published September 29, 4:10 pm EDT

Artificial intelligence (AI) is everywhere, and so are investments into chip stocks. However, selling the chip is only the beginning.

Once thousands of AI processors arrive at a data centre, somebody has to supply enough electricity to run them, connect the facility to the grid, build substations and transmission, manage cooling, and keep power flowing around the clock.

That physical buildout could create a second group of AI winners long after the servers have been delivered.

data center server racks glow with light — Source: Getty Images

Follow the electricity

The International Energy Agency (IEA) estimates electricity consumption from data centres reached about 485 terawatt-hours (TWh) in 2025 and could almost double to 950 TWh by 2030. AI-focused facilities are growing faster, but electricity use could triple over the same period.

The problem is speed. A data centre can be constructed faster than a power plant or major transmission project. Transformers, gas turbines, interconnections, and grid approvals have already become bottlenecks.

That creates two obvious places to look for revenue: companies selling electricity and companies helping build the infrastructure needed to deliver it. Capital Power (TSX: CPX) covers the first.

Sell the power

Capital Power owns power-generation facilities across Canada and the United States. In July, it signed a greater-than-10-year agreement to provide 250 megawatts (MW) of capacity and electricity for Meta‘s planned Sturgeon County data centre in Alberta. Supply is expected to begin during the second half of 2028. Capital Power says the agreement should provide stable long-term cash flow while leaving room for additional opportunities around its Genesee site.

Second-quarter adjusted funds from operations (AFFO), which measures cash generated by the business before certain items, reached $328 million. Management also increased the dividend for a 13th consecutive year.

At $62.58, the $2.80 annualized dividend provides a yield around 4.5%. For investors looking at Canadian dividend stocks, AI demand could become another customer for infrastructure Capital Power already owns.

The risk is that data centre development slows or projects arrive later than expected. Power generation also demands enormous amounts of capital, and new construction can run over budget. Then somebody has to design the next round of infrastructure.

Build the grid

WSP Global (TSX: WSP) provides engineering, consulting, design, and project-management services across transportation, buildings, water, environmental projects, and power systems.

Its February acquisition of TRC Companies expanded that last business at a useful moment. TRC specializes in power and energy infrastructure, including transmission, grid modernization, and interconnections.

Second-quarter results showed the effect. WSP’s net revenue increased 22.9% to $4.3 billion. Its backlog reached a record $20.1 billion, up 23.2% year over year. TRC’s Power & Energy business delivered double-digit net revenue growth compared with its results before the acquisition.

AI isn’t responsible for all of that demand. Utilities were already investing to replace aging infrastructure, connect renewable generation, and electrify transportation and buildings. Data centres add another reason to spend. That makes WSP an interesting way to approach Canadian growth stocks without needing to predict which AI model consumers will use. At $184.84, WSP trades around 15 times forward earnings after falling well below its 52-week high of $290.23.

Considerations

WSP’s main risk is execution. Acquisitions helped drive growth, but leverage climbed after buying TRC. Integrating large businesses while maintaining margins requires discipline.

Capital Power brings a different risk. Its Meta contract improves cash-flow visibility, but one 250-MW agreement doesn’t turn the entire company into an AI stock. Neither company needs AI to be its whole business, however, and that’s what I like.

Bottom line

The chipmakers get paid when AI hardware is sold. Capital Power can get paid for supplying electricity for years afterward. WSP can get paid as utilities and developers build the infrastructure needed to deliver it.

AI may live in the cloud. Yet the next round of revenue increasingly depends on what gets built underneath it.