Jabil

Where we land

Jabil is a contract manufacturer close to the heart of the AI data-centre build-out, assembling the racks, servers and supporting gear that cloud companies are buying in bulk. Management's fiscal 2027 plan sets total revenue at $44.5B, roughly half of it from AI-related work. Since that plan appeared, analysts have raised their earnings forecasts 16 times in six weeks without a single cut. We lean positive, and we award the company gold.

The lean comes with reservations. The shares fell when the plan was published, margins are thin, borrowing is heavy, and the price already asks for years of steady earnings growth. What tips the balance is that the demand is showing up in sales today rather than only in promises, and that the people who model this company for a living keep moving their numbers up.

What the business is

Jabil makes things for other companies. A brand owner designs a product, and Jabil sources the parts, builds it, tests it and ships it, a line of work known as electronics manufacturing services. Our figures place all of its revenue in that single line, so the AI hardware, medical devices and consumer gadgets it builds sit under one heading.

That work pays slimly. Gross margin, what is left of each sale after the direct cost of making the product, is 9%. Operating margin, what remains once the cost of running the company is also paid, is 5%. Neither is a weakness in itself: customers keep the brand and most of the pricing power, and pay Jabil for scale, reliability and execution rather than for ideas.

What stands out is how steady that thin margin has been. Over five years gross margin has held at a middle value of 9% with almost no movement either side. That is the mark of a manufacturer that prices its contracts with discipline and keeps a firm grip on costs, and a large part of why we rate the quality of this business so highly.

Why now

Cloud and AI companies are spending heavily on data-centre hardware, and Jabil assembles and integrates the servers, racks, networking and power equipment that go inside those buildings. Its full-year forecast, driven by that demand, came in above what analysts had expected.

On quality, Jabil ranks among the strongest companies we rate, but that is not the same as being the best-placed company in its field. Its nearest rival in AI hardware, Celestica, earns fatter margins and is growing faster, yet the market prices Jabil at a higher multiple of its earnings. The case for Jabil has to explain why that premium is deserved; the case against begins with the suspicion that it is not.

We have held it in the Conviction 10 since September 2026, and it still qualifies for a place there.

The case for

The plainest argument is that the business has woken up. Sales rose 20.6% over the past year, after three years of compound growth of just 1.2% a year. A company that stands still for three years and then accelerates this sharply has usually found a new source of demand, and here that source is named and visible.

Profit tells the same story. Operating profit held at around 1.5bn from 2022 to 2025, a long plateau, then reached 1.8bn in the latest year: the first real step up in some time, and driven by the same AI work that lifted sales.

The outlook points further in that direction. Management has targeted AI-related revenue of $22.1B within its fiscal 2027 plan, and analysts' consensus implies earnings per share rising about 25% next year. The forecast changes have been strikingly one-sided, with 16 increases and no cuts over 45 days. When estimates move that consistently, it usually means the people closest to the company keep finding their earlier numbers too cautious.

There is also a new door opening. In October, Jabil and India's Adani Group signed an agreement to form an equal joint venture making AI and data-centre equipment in India, carrying the AI hardware push into a fresh market, though our figures carry no financial terms for it.

One qualification is needed. Measured against the wider market, Jabil's growth looks far better than it did six months ago, but only a little better than a year ago. This is a recovery more than an established breakout, and the stronger argument for the shares is that the recovery has further to run, not that it is complete.

Where the growth could come from

The AI data-centre build-out explains Jabil best right now. On management's plan, AI-related revenue would be about half the company by fiscal 2027; our revenue figures do not split AI out, so that target is the clearest measure of its size. This single line of work accounts for the gap between last year's growth and the near-standstill before it, and underpins the expected jump in earnings per share. The debate is no longer whether the AI earnings exist but how large and lasting they will be, and steady delivery, quarter after quarter, is what could change how the market values them. The sign it is working would be AI-related revenue tracking at or above the $22.1B target.

The Adani venture is an option on the future rather than something in the numbers today. If it works, it could give Jabil local demand from Indian data-centre builders and a factory base outside China for its global customers, lengthening the AI growth runway beyond the current cycle of American cloud spending. For now it is a memorandum.

The swing factor is the other half of the company. Outside AI, Jabil builds products for healthcare, cars and electric vehicles, renewable energy, industrial customers, phones and consumer devices, and several of those markets have been soft. These businesses most likely explain why growth stayed near flat for years, and why the AI surge has had to carry the results on its own. If this half stops shrinking or starts to grow, earnings would rest less on one AI cycle and investors would have less reason to mark the shares down the way they did after the outlook.

The most interesting lever sits inside the AI business itself. Faster AI networks and hotter chips need optical connections, power systems and liquid cooling, and Jabil makes parts in all three areas, not only the final assembly. Assembly earns thin margins; owning higher-value components could let profit grow faster than sales, supporting the view that the market is underestimating the AI earnings. This is our reading of where the margin opportunity most likely lies, and it would be confirmed by management citing rising sales of optical, power or cooling products while operating margin widens alongside AI revenue.

The case against

Start with the comparison that matters most. Jabil's gross and operating margins trail Celestica, Flex and Plexus. Celestica earns an operating margin of 9%, is growing considerably faster, and trades at a lower 25 times earnings. Reliability is not the issue: Jabil's business quality is close to the best in the market, level with Plexus and Kitron. The gap is in margin and growth, exactly what an investor paying a premium would want more of, not less.

Then there is the market's own verdict. The shares dropped about 10% on the day the fiscal 2027 outlook was published, even though the results beat expectations, most likely because investors had already priced in more than management delivered. The shares have also lost much of their earlier upward drive, weakening over three, six and twelve months.

The headline return on shareholders' money flatters. Return on equity, which measures profit against the money shareholders have in the business, is 64.6%. That looks exceptional until set beside the balance sheet: borrowings less cash amount to 1.6 times shareholders' equity, and the shares trade at 17.7 times the book value of that equity. Much of the high return comes from a small equity cushion rather than rich margins, and a thin cushion magnifies losses as readily as gains.

Bottom-line profit has been erratic too. Net profit fell from 1.4bn in 2024 to 657m in 2025 while operating profit held flat. Whatever drove that swing sat below the operating line, and it means the earnings reaching shareholders have been less dependable than the underlying business.

What the price assumes

To justify today's share price, Jabil's earnings would need to compound at about 9% a year for five years, taking earnings per share to 14.78. That assumes the share price stays flat and that, in five years, the market values the company at 18.5 times earnings, the middle value across every profitable listed company we cover.

Jabil's own record is no help in judging whether that is likely, because its earnings have moved too unevenly to show a trend worth projecting. Carrying the recent pattern forward, which amounts to earnings shrinking about 18% a year, would leave earnings per share at 3.63, only a quarter of the required level. But that line is drawn through a choppy past, and the analysts' view points firmly the other way.

So the price is a bet that the analysts are right and the past is not a guide. The shares have reflected both sides of that argument: up 24% over the past year, but down 20% over the past six months, as enthusiasm for the AI story met doubts about how long it lasts.

The other side of the medal

The risk to keep watching is a combination rather than a single flaw. Jabil's growth rests heavily on a handful of large AI buyers, its borrowings less cash stand at 1.6 times shareholders' equity, and the businesses outside AI are still near flat. Any one of these would be manageable. Together they mean the company has staked much of its future on a narrow group of customers, has little balance-sheet slack if those customers slow, and has no second engine ready to take over.

That, more than the margin gap or the share-price wobble, is the risk that matters, because the price already asks for earnings to grow about 9% a year for five years. If a few big buyers pause their data-centre spending, the AI half of the business would stall, the flat half could not fill the gap, and the debt would turn a disappointment into real damage to shareholders' returns.

The first signs would be hard to miss: AI-related orders falling short of the target, analysts beginning to trim their forecasts after weeks of only raising them, and operating profit sliding back toward its old plateau.

What to watch

The earliest signal sits with Jabil's customers rather than with Jabil. The large cloud and AI companies' spending plans on data-centre hardware move before Jabil's own sales do, and any sign that they are trimming or delaying that spending would reach Jabil's order book soon after. Alongside that, watch whether the AI customer list widens beyond a few big buyers, because that, more than the size of any single quarter, decides how lasting these earnings prove.

The direction of analysts' estimates is the next tell, after an unusually clean run of increases. If forecasters begin cutting the expected jump in earnings per share, or operating profit slips back toward its earlier plateau of about 1.5bn as AI orders fall short, the growth reading here would be wrong.

Beyond AI, two things would strengthen the case. One is healthcare, automotive and consumer markets showing growth on the year rather than declines masked by AI. The other is the Adani venture becoming a signed, funded company with a named factory site and first customer orders, which would turn an option into a second leg for the business.

The numbers

The same industry, for comparison

Research for information, not personal investment advice or a recommendation to buy, sell or hold any security: see Important information. Data as at 31 October 2026.