The math is unusual for a company that makes boxes: 23 analysts cover Celestica, the Toronto-based maker of computing hardware for data centers, and 20 of them say buy. One says strong buy. Two say hold. None says sell. The average price target among the group stands at $438.86, with the most optimistic forecast at $510 and the most cautious at $300, according to data compiled on Aug. 25.
The consensus rating, a plain “buy,” reflects a broader reassessment of the company’s role in the AI economy. Celestica builds the servers, storage systems, and networking gear that hyperscale operators install by the tens of thousands, and its fortunes track the industry’s capital spending more directly than almost any software company’s. As the biggest cloud providers accelerate upgrades to higher-throughput architectures, Wall Street has concluded that the hardware layer will capture more of the value than it did in earlier technology cycles.
The company’s position has been described by analysts as a “picks and shovels” play on AI: it does not design the chips or the models, but it assembles and ships the physical infrastructure both depend on. That role has historically carried thin margins and cyclical demand, which is part of why Celestica traded for years as a modestly valued contract manufacturer. The current upgrade cycle, analysts said, is different in at least one respect: the customers cannot wait.
Hyperscalers, the operators of the largest cloud networks, are in the middle of the most rapid data-center expansion in the industry’s history. New generations of chips require new servers, new power delivery systems, and new cooling equipment, and the transition is happening on a compressed schedule. Celestica’s manufacturing capacity, its supply chain relationships, and its ability to ramp production quickly have become scarce commodities in their own right.
The earnings picture has followed. Celestica has posted results that beat analyst expectations in recent quarters, and management has guided for growth that assumes the current spending wave continues. The upgrade to higher-throughput architectures, which began with the largest customers, is expected to spread through the rest of the market, giving the company a runway that extends beyond the current order book.
Risks remain. The most obvious is concentration: a handful of customers account for a large share of Celestica’s revenue, and those customers have shown a willingness to shift orders between suppliers as technology changes. The company also faces competition from contract manufacturers with lower cost bases and from its own customers, some of which have begun building hardware in-house. A slowdown in AI capital spending, whatever its cause, would hit Celestica’s order book quickly.
The stock’s run has already been long. Celestica shares have climbed sharply since the AI build-out began, and the current price targets imply further gains even after that appreciation. Valuation is the other side of the coin: at the average target, the company would trade at a multiple that assumes the growth persists, leaving little room for disappointment.
Analysts who remain cautious point to the same fundamentals from the other direction. The two hold ratings are based not on doubts about demand but on questions about margins and timing: whether Celestica can convert record revenue into record profit, and whether the current architecture transition lasts long enough to justify the current prices. The $300 target, roughly a third below the average, prices in a sharper slowdown than any of the company’s recent results suggest.
For the industry, the analyst consensus is a signal as much as a forecast. When the covering analysts of a contract manufacturer all lean in the same direction, it tells the broader market where capital believes the durable value in AI resides. Celestica’s customers, the hyperscalers, have made their own bets in the hundreds of billions; the analysts are betting that the companies that build their machines will do well too.
The comparison to earlier cycles is instructive. In the last major data-center build-out, driven by cloud computing in the late 2010s, hardware suppliers captured steady but unspectacular returns, and their stocks traded at modest multiples even at the peak of spending. The current cycle has differed so far: capacity is tighter, the technology transition is faster, and the customers are more willing to pay for speed. Whether Celestica’s shares hold their gains will depend on whether those differences persist or whether the industry reverts to its old habits of oversupply and margin pressure.
The next test comes with the company’s quarterly results, when investors will see whether the order growth embedded in the price targets is actually arriving. For now, the sell-side message is consistent: in a market where everyone wants to own the picks and shovels, the highest-conviction call is on the shovel maker itself.
