Celestica's $3 Billion Equity Raise: The Signal You Need to Read in the AI Growth Story

Celestica announced a $3 billion equity offering today, describing it as capital to fuel its expansion across global AI infrastructure. The press release reads like a growth story. The math behind the raise reads like something else entirely.

A $3 billion equity raise against a $41.7 billion market cap is not a routine working-capital event. It is a signal. And it's the signal that matters more than the headline about AI infrastructure demand.

Here's what I'm looking at first - because that's always where the real story lives. Not the price, not the P/E ratio, not analyst opinions. The structure of the demand, the shape of the business model, and whether the capital being raised changes the risk profile of the position.

The Growth Is Real - That's Not the Question

Let me get the easy part out of the way. Celestica's AI infrastructure demand is not speculative. It's accelerating, and management has been raising guidance every quarter since the AI cycle turned.

Q2 2026 revenue hit $4.70 billion, up 62% year-over-year and well above the high end of the $4.39 billion consensus estimate. Adjusted EPS came in at $2.54 versus an expectation of $2.31. The Connectivity & Cloud Solutions segment - the one that builds servers, storage, and networking gear for hyperscalers - grew 84% year-over-year. That's the number that tells you which way the wind is blowing.

Management raised its 2026 full-year revenue guidance to $20.5 billion (up from $19.0 billion just eight weeks ago) and adjusted EPS to $11.30, up from $10.15. They went further, saying revenue growth in 2027 will accelerate beyond the 65% growth rate they're targeting in 2026. That kind of sequential guidance increase, repeated across multiple quarters, is not noise. It's a supply chain confirming what the market already suspected: hyperscaler capex is running hot, and Celestica is on the right side of it.

But the debate is not whether the demand is real. The debate is whether a $3 billion equity raise changes the capital allocation calculus - and whether the business model shift this offering exposes will affect the stock's return profile.

Why $3 Billion - And What It Means

Celestica says the net proceeds from the treasury offering will fund working capital, capital expenditures, and general corporate purposes. That's the official line. The operational line is simpler: this business is growing at 65% and it needs a lot more money to fund the inventory, supply commitments, and balance sheet that growth requires.

Let's look at the numbers that tell you how capital-intensive this growth actually is. Inventory sits at $3.4 billion. That's a massive amount of capital sitting in components and finished goods waiting to flow through the supply chain. Operating cash flow for the trailing twelve months is $1.14 billion, with free cash flow (operating cash flow minus capital expenditures) at $519 million. Capital expenditures run $625 million.

Now put the $3 billion raise against that. The equity offering is roughly five and a half times annual free cash flow. It's nearly three times operating cash flow. That's how much capital this growth engine is consuming.

The reason it's so large comes down to what Celestica actually does. Electronics manufacturing services - and more specifically, original design manufacturing for data center hardware - is a working-capital-heavy business. When you're building AI servers and networking racks on a turnkey basis, you buy the silicon, the memory, the printed circuit boards, the power supplies, and the chassis before the customer pays you. The gap between what you spend and what you collect is the working capital hole. At 65% growth, that hole is expanding fast.

Total debt sits at $7.3 billion with total equity of $2.4 billion. The company has $536 million in cash.

Put plainly: the $3 billion raise is both proof of demand strength and proof of capital intensity. It's a dual signal, and I always re-evaluate my allocation when I see that pattern.

The Consignment Shift That Changes Everything

Here's the detail from the Q2 earnings call that the market hasn't fully processed yet. Celestica told investors on July 27 that the OpenAI program is expected to operate under consigned economics - meaning the customer owns the accelerator silicon and major materials, and those costs don't flow through Celestica's revenue.

This matters enormously for how you read Celestica's financials going forward.

Under a turnkey model - where Celestica buys and resells the networking silicon, memory, and components - reported revenue is enormous but gross margin percentage is thin, and working capital requirements are crushing. Under a consigned model, reported revenue is much lower relative to the total system value, but gross margin percentage is better, operating margin quality is higher, and the working capital hole shrinks dramatically.

Management said the vast majority of their networking programs include silicon in reported revenue (turnkey), while the vast majority of compute programs exclude accelerator silicon (consigned). The implication is that as AI infrastructure shifts from networking-heavy deployments to compute-heavy deployments - and as more hyperscalers move toward customer-owned silicon - Celestica's revenue growth rate could slow even while profitability per deployed rack improves.

I believe the consignment shift is the single most important structural change in how you evaluate Celestica over the next 18 to 24 months. It's the difference between a company whose revenue tells you everything and a company whose revenue tells you less than its margin and cash conversion do.

For investors, this means top-line growth comparisons become increasingly misleading. A company recognizing silicon pass-through can appear larger on revenue while creating less gross profit per unit of deployed infrastructure than a consigned-system manufacturer. The appropriate measures of value capture shift from revenue growth to gross profit, operating profit, cash conversion, and return on invested capital.

Celestica already shows that shift is underway. ROIC sits at 39.7%, ROE at 52.7%. Those are exceptional returns for an EMS business. If the consignment mix grows, those returns could improve further - but the revenue growth rate the market has been rewarding might not look as spectacular.

The Supply Chain Confirms the Demand

I want to make sure we're not losing sight of the supply-side signal here. Celestica told investors on the Q2 call that some product lead times are "well above 52 weeks" and that customers have placed non-refundable orders for long-lead silicon. The company has supply agreements for hard-to-find components including memory and printed wiring boards.

Non-refundable long-lead orders from hyperscalers is one of the strongest possible demand signals you'll see in this industry. It's not guidance. It's money already committed. This is the kind of supply chain evidence that separates real AI infrastructure demand from the kind of cyclical optimism that fades when capex budgets get cut.

The supply chain is also telling you that component pricing power is solid. Memory, silicon, and PCBs with lead times exceeding a year and non-refundable commitments - that's an environment where suppliers don't discount and customers don't delay. The multi-year demand visibility Celestica claims is backed by financial commitments, not management optimism.

The Valuation Already Knows

The stock has run hard. Celestica is up 74.7% over the trailing 12 months. It's trading at $362.76, down 2.3% on the day of the equity announcement. The stock hit a 52-week high of $474 and has pulled back from there - but it's still near cycle highs.

Valuation metrics are telling you how much of this growth is already priced in. The stock trades at 2.7 times trailing sales, 37.4 times trailing earnings, and 70.2 times forward earnings. EV/EBITDA is 27.2x. Compare that to Jabil at 1.1 times sales and 17.6x EV/EBITDA, and you see the premium the market has assigned to Celestica's AI exposure.

The forward P/E of 70x is particularly revealing. That multiple prices in near-flawless execution on the $20.5 billion 2026 guidance, the consignment transition going smoothly, and the $3 billion in equity not creating earnings drag through dilution. It leaves almost no room for disappointment.

That's not a judgment on whether Celestica deserves its premium. It's a statement about what that premium assumes. When I see a stock pricing in multiple consecutive years of accelerating growth at a 70x forward multiple, I start asking a different question: is my capital better deployed elsewhere?

The Risk Framework

Here's how I'm thinking about this trade now.

The demand case for Celestica is stronger than almost any other EMS player. The Q2 numbers confirmed acceleration, not deceleration. The 2026 guidance raise and the 2027 acceleration promise suggest the AI infrastructure cycle still has room to run. The consignment shift, if it executes as described, could improve margin quality and reduce the very working-capital intensity that the $3 billion raise is meant to solve.

However, the equity offering itself changes the equation. A $3 billion raise dilutes existing shareholders by roughly 7%, depending on the offering price. That dilution is the cost of maintaining 65% growth in a capital-intensive model. The question is whether the growth and margin expansion that follow justify that dilution.

I believe they will - over a three-year horizon. But I'm also clear that the near-term return curve is back-half weighted. The stock has already absorbed the first wave of AI infrastructure excitement. The $3 billion raise signals that the next phase of growth requires real capital, not just momentum.

This is what separates the early-cycle Celestica trade from the current one. In 2024 and 2025, the stock was rewarded for proving it belonged in the AI infrastructure value chain. Today, the stock is being rewarded for scaling - but scaling requires capital, and capital costs money in the form of dilution.

Where the Capital Goes

I'm not telling anyone to sell Celestica. The thesis is intact. The demand is real. The consignment transition, if it executes, could be a profit-quality upgrade that the market hasn't fully priced.

But I am saying the allocation decision is no longer binary. A position that made sense at $173 (the 52-week low) carries a different risk profile at $363. The 70x forward multiple means execution needs to stay near-perfect. The $3 billion equity raise means dilution is real, even if it's buying future growth.

For existing holders, the question is whether the position size matches your conviction in the 2027 acceleration story. For new entrants, the question is whether this return profile is better than what you'd find in companies that haven't run as far yet.

The debate is not whether Celestica stays important to the AI infrastructure build-out. It's whether the return profile over the next 12 to 18 months justifies the same allocation as it did at half the price and a fraction of the current multiple.

In my opinion, it doesn't. Not without a meaningful pullback. The long-term thesis - that Celestica will continue capturing AI infrastructure revenue as the EMS market shifts toward design-manufacturing hybrids - is still valid. But much of that return is likely back-half weighted, and the $3 billion equity raise is the clearest signal I've seen that the next chapter of growth is going to cost shareholders something.

That doesn't mean exit. It means the position gets managed differently. Smaller allocation. More active protection. And a willingness to move to the sidelines if the consignment transition slows or if Q3 guidance fails to confirm the acceleration management promised.

The AI infrastructure cycle is not over. Celestica is on the right side of it. But right side of the cycle doesn't automatically mean right time to hold the same position.