One of the products that has put Celestica Inc. CLS-T at the centre of the artificial intelligence frenzy is something called a networking switch. This hardware is crucial for data centres to function, particularly for AI. A switch acts as a traffic cop of sorts, directing huge amounts of information packets between servers quickly and efficiently so that nothing goes awry.
Financial markets, on the other hand, don’t work that way. Investors pile into the shiny new thing with nothing to stop them. These days, that’s AI. And of all the publicly traded companies in Canada, Celestica is most directly tied to the boom.
The Toronto-headquartered global manufacturer has seen its share price soar close to 340 per cent since last year – and more than 2,500 per cent since OpenAI released ChatGPT in November, 2022, akin to a junior miner that has struck gold.
For Celestica, the question is how much longer the ascent can continue.
According to analysts who follow the company, there are reasons to be bullish. The company itself will provide an update at its investor day Tuesday, when it also releases third-quarter earnings. (Celestica declined an interview, citing a quiet period before its financial results.)
If AI is a bubble, the economy will pop with it
The explosion of generative AI is prompting some of the world’s biggest tech companies to spend hundreds of billions of dollars for data centres to train AI models and support growing usage of the technology. For that, they will need Celestica’s switches, storage and other equipment. Alphabet Inc.’s Google GOOGL-Q and Meta Platforms Inc. META-Q are among its customers.
The top four U.S. cloud companies alone are planning to spend some US$300-billion this year, a nearly 60-per-cent jump over 2024, according to analysts at JPMorgan Chase & Co. The growth rate is expected to slow next year, but these companies are still projected to spend an additional US$80-billion, representing a 30-per-cent jump compared with 2025.
Beyond the eye-popping numbers, however, there are worrying signs. A chasm yawns between the amount companies are spending to build infrastructure and the financial returns from AI today. David Cahn, a partner at Sequoia Capital, estimated that gap at US$600-billion last year. He wrote more recently that the figure now seems quaint.
Moreover, he argued that the only possible way to make sense of the data-centre bonanza is that some companies believe that artificial general intelligence, or AGI, is around the corner. That is the nebulous and hotly debated concept that AI becomes as smart as humans. “Nothing short of AGI will be enough to justify the investments now being proposed for the coming decade,” he wrote.
Some market watchers are concerned about a spate of so-called circular-financing deals in recent months, too. To take one example, Nvidia Corp. said it will invest up to US$100-billion into OpenAI in instalments based on the data-centre capacity the ChatGPT-maker rolls out. OpenAI, of course, is a buyer of Nvidia chips, so it appears that a supplier is subsidizing its customer.
Opinion: Immutable market truths for judging the effects of tariffs, AI and more
Celestica’s share price, meanwhile, is tethered to expectations for data-centre spending. “Any change in that spending trajectory is likely to be swiftly reflected in the share price,” said Jesse Pytlak, an equity analyst at Cormark Securities.
The tizzy over DeepSeek earlier this year showed that. The Chinese company released an AI model that appeared to require a fraction of the chips that competing models need, sparking a mini panic as investors fretted that the tech giants were overspending on infrastructure. Celestica’s stock dropped nearly 30 per cent, though it quickly recovered.
Celestica is far from the only company whose fortunes are currently riding on AI, of course; the market as a whole is increasingly levered toward the technology. According to a JPMorgan report this month, 30 AI-related stocks account for 44 per cent of the market cap of the S&P 500. That compares with 26 per cent in 2022.
“There are certainly reasons to be concerned about a bubble,” Mr. Pytlak said. “While this is not technology that is going away, I believe that spending patterns will inevitably have to evolve.”
Nobody knows when or to what degree. “At some point, there’s a risk that they overbuild,” said Thanos Moschopoulos, an analyst with BMO Capital Markets. “I don’t think we’ll be there for the next two or three years.” The near-term demand for AI infrastructure is too strong, for one thing, and even with the surge in its share price, Celestica remains reasonably valued based on 2027 estimates, he said. Revenue and profit are growing, too.
Celestica becomes Canada’s third most valuable tech company as stock surges
So long as the spending continues, Celestica is uniquely positioned to benefit. The products the company manufactures are not mere commodities, but are sophisticated and customized for clients, and the company has beefed up its research and development capabilities. “All of a sudden the hardware needed to do AI is just a lot more complex,” Mr. Moschopoulos said. There is less geopolitical risk with Celestica as well, which has operations in places such as the United States, Malaysia and Thailand, whereas some competitors are based in Taiwan.
Given the concentration in the industry, a few customers account for a meaningful chunk of Celestica’s revenue. During the last quarter, one customer accounted for 31 per cent of revenue, while another made up 13 per cent.
Celestica is trying to diversify, however, and management has disclosed a “digital native” customer, without identifying it by name. Some analysts believe it’s OpenAI.
Earlier this month, OpenAI announced a partnership with Broadcom to develop and deploy custom chips. The market expects Celestica to benefit through the sale of associated server, storage and networking hardware.
How that deal pans out ultimately is unclear. OpenAI does not have a track record as a chip designer. It has also committed to building tens of billions of dollars worth of data-centre infrastructure. But its ability to pay is partly contingent on massive revenue growth and stemming its losses, though it may always find new sources of capital.
At least one stock could be depending on it.
