CLS Stock Just Dropped 8% in a Single Day — Here’s What the Sell-Off Is Actually Telling You

When a company raises its full-year revenue outlook to $17 billion, beats earnings estimates by more than 7%, reports 43% year-over-year revenue growth, and then sees its stock drop 8% in a single session, investors become particularly irritated. On March 31st, Celestica shareholders were in a similar situation, with CLS trading at $257 after opening at $281, falling to an intraday low of $253, and closing almost $23 below the previous session. It wasn’t even a particularly active day on volume. There was no panic in the selling. It was intentional, which almost makes it more eerie.

One of the reasons Celestica’s story is worthwhile is that it is not as well-known as Nvidia or AMD. The company was established in Toronto in 1994 as an IBM spinout, essentially the manufacturing infrastructure that IBM decided it no longer wanted to own. For the first twenty years of its existence, it operated as a contract electronics manufacturer, a type of business that makes modest profits by manufacturing complex components for companies that design but don’t build. That model was successful. It wasn’t thrilling. Then came the hyperscaler buildout, spending on AI infrastructure began to compound at rates no one had quite predicted, and Celestica’s Connectivity & Cloud Solutions division began to expand in ways that fundamentally altered the company’s nature.

The Q4 2025 figures are truly remarkable. $3.65 billion in revenue, a 43.57% increase over the same period last year. In contrast to the historical baseline of contract manufacturing, where extracting a point or two of margin from high-volume, low-complexity assembly is deemed a win, gross margins are growing into the low-teens range. Servers, storage systems, and high-performance networking equipment for hyperscale data center clients are all part of Celestica’s Hardware Platform Solutions business within the CCS segment. Because these products are more engineering-intensive, more deeply ingrained in customers’ infrastructure roadmaps, and more difficult to replicate at scale, they fetch higher prices than traditional EMS work. A pure contract manufacturer is positioned differently in the value chain than a company that can design, integrate, and manufacture at this level for the biggest cloud operators worldwide. The financials are beginning to reflect Celestica’s intentional transition.

The most obvious indication of the company’s positioning is the AMD partnership that was announced in March. Helios, a rack-scale AI platform developed by Celestica and AMD, is essentially a whole hardware system tailored for AI workloads as opposed to a collection of separate parts put together according to specifications. The true engineering complexity of modern data center deployment lies in rack-scale integration, and a vendor that can provide hyperscalers with that capability differs from one that merely assembles circuit boards. The market may not have fully considered the implications of this partnership for Celestica’s long-term pricing power and margin trajectory, or it may have done so and is merely discounting against execution risk.

It is more difficult to categorically reject insider selling. President Todd C. Cooper reduced his direct ownership by 45% in early February when he sold 89,484 shares at an average price of $287.45, a transaction valued at over $25.7 million. Senior executive Yann L. Etienvre sold 86,229 shares for about $24.8 million that same day. During the most recent quarter, executives sold almost 298,000 shares valued at about $88 million when combined with other insider transactions. The current level of insider ownership is 0.52%. For a company whose institutional narrative primarily relies on management conviction in the growth story, that is a thin number. Investors are right to take notice of these sales, even though it’s still unclear if they indicate regular portfolio diversification or something more targeted.

It’s both comforting and strange to see how the analyst community has reacted to the recent weakness. With an average price target of $354.24 and a consensus rating of Moderate Buy, the stock is expected to rise by about 37% from its current position. Canaccord Genuity’s goal is $430. In January, Aletheia Capital increased its goal to $410. Citigroup maintained a buy rating but lowered its target from $375 to $338. In February, Wolfe Research upgraded to Strong Buy. In light of this, the stock is currently trading at a P/E of about 36 times trailing earnings, having dropped nearly 30% from its 52-week high of $363.40. That multiple doesn’t clearly indicate overvaluation for a company with 43% revenue growth, unless you take into account the energy cost issue that is starting to loom over every AI infrastructure play in the industry.

The energy issue is real and becoming more difficult to overlook. The combined spending on AI infrastructure by big tech is predicted to reach $635 billion in 2026, a significant increase from the previous year. The data centers these companies are constructing have enormous power requirements. Celestica is not directly impacted by rising energy costs in the same way that utilities or fuel-intensive manufacturers are, despite the Iran War’s disruption of the world’s oil markets. However, they hit the hyperscalers, and the businesses involved in that supply chain are affected downstream if hyperscalers begin to reevaluate the speed of their capital expenditure programs. Melissa Otto, an analyst at S&P Global, has specifically pointed out that persistent pressure on energy costs may force a reconsideration of the timelines for investments in tech infrastructure, which could have repercussions for the entire vendor ecosystem.

With CLS trading at $257 against an analyst consensus of $354 and a 52-week range of $58 to $363, it is difficult to ignore the fact that this stock has an exceptionally broad range of possible outcomes. The 52-week journey itself reveals how unsure the market has been about how to value this specific company at this specific point in the AI infrastructure cycle, from a low that now seems almost impossible in retrospect to a high that now seems like a distant memory. The pre-market rebound to $261.50 the morning following the 8% decline indicates that the selling was not solely motivated by conviction. However, there are still unanswered questions about energy costs, hyperscaler CapEx durability, and AMD partnership execution that will determine whether the floor holds and where the next significant move goes.