Sanmina Corporation

My qualitative portrait label is company in transition. That is not a stylistic flourish. It is the central investment fact. Sanmina has not suddenly become Super Micro, and it is not merely old Sanmina with a little extra cloud demand. It is a legacy EMS operator with better vertical integration, a growing regulated-industrial base, and a transformed AI systems manufacturing arm whose durability still has to be proved through at least several more quarters of customer mix, shipment phasing, cash conversion, and margin retention. The business looks stronger than it did a year ago. The stock also knows it.

Company vertical history and financial review

Origins and listing path

Sanmina was founded in Silicon Valley in 1980, and the old core of the business was printed circuit boards, backplanes, and related manufacturing services for complex electronics customers. The company’s modern description still emphasizes deep PCB, backplane, machining, enclosure, optical, RF, microelectronics, and systems-integration capabilities, which is useful because it shows continuity between the early company and today’s broader manufacturing platform. Sanmina did not start as a software-like category creator. It started by solving a harder, less glamorous problem: how to build demanding hardware reliably at scale for OEM customers that cared about quality, engineering support, and production transfer.

It listed on Nasdaq in 1993 under its predecessor form, Sanmina Holdings. The SEC filing trail confirms the public-company lineage, but in this research pass I did not find a primary filing that cleanly stated the original IPO price and gross proceeds without relying on older archival materials of uncertain completeness. I am therefore not repeating a number I cannot verify. What is verifiable is the broader capital-markets story: Sanmina entered the public markets during the long outsourcing wave that helped create the modern EMS industry, when OEMs increasingly shed internal manufacturing and looked for partners that could handle both scale and complexity.

The decisive early turning point was the 2001 merger of Sanmina and SCI Systems. That combination created the Sanmina-SCI identity investors remembered for years, and it transformed the company from a specialized manufacturer into a global EMS platform. It also set up both the promise and the pain of the next decade. Scale brought customer reach, factories, and broader capability. It also brought the usual post-merger burdens in this industry: lower-return revenue, integration friction, cyclical exposure, and a share price that could move much more dramatically than the underlying quality of operations. The current corporate name, Sanmina Corporation, reflects that post-merger evolution.

Stage division and key nodes

The first stage ran from founding through the early public years. Sanmina was building an identity as a high-complexity manufacturing specialist in an industry where outsourcing was still moving from exception to norm. The growth driver was not consumer branding or proprietary demand pull. It was OEM willingness to transfer increasingly sophisticated manufacturing work to outside partners. That early shape still matters because Sanmina’s best businesses today remain the ones where the customer values process discipline, quality systems, and manufacturing know-how more than the lowest bid.

The second stage was the scale-up through the 1990s and into the SCI merger. Here the company leaned into breadth. The market’s belief in that era was simple: larger EMS platforms could absorb more outsourced programs and earn better economics from footprint, procurement, and customer intimacy. The flaw in that belief was that scale in manufacturing often improves survivability more than pricing power. Sanmina emerged from this period much larger, but also more exposed to telecom, computing, and broad hardware cycles. That trade-off explains why the stock could explode in the dot-com era and then spend years paying for excess enthusiasm. The all-time high reflected in public stock-history datasets came in 2000, not in the present AI cycle.

The third stage was the long digestion from the early-2000s bust through the 2010s. This was when Sanmina had to prove it could be more than a roll-up of factories. A large part of the work was portfolio shaping: more regulated and mission-critical end markets, more vertical integration, more repair and services, more emphasis on industrial, medical, and defense, and more discipline in capital allocation. Over time the company improved gross margin from 7.2% in fiscal 2019 to 8.2% in 2021, and operating margin from 3.5% in 2019 to 4.2% in 2021, even though revenue in those years was not growing spectacularly. That is what a maturing EMS operator looks like when it is becoming more selective rather than simply larger.

The fourth stage began around the post-pandemic recovery and ran through fiscal 2025. Revenue rose from 6.74 billion USD in 2021 to 8.94 billion USD in 2023 before slipping to 7.57 billion USD in 2024 and recovering to 8.13 billion USD in 2025. The important thing was not smooth top-line growth. It was that cash generation stayed dependable and margins held up better than in prior cycles. Operating cash flow was 338 million USD in 2021, 331 million USD in 2022, 235 million USD in 2023, 340 million USD in 2024, and 621 million USD in 2025. In that same period Sanmina repurchased stock and kept building credibility as a disciplined operator rather than a chronic value trap. By fiscal 2025 it was not yet an AI darling, but it had become a much sturdier business than the market had historically assumed.

The fifth stage is the one investors are paying for now: the ZT reset. The acquisition was announced in May 2025, financed through a bridge structure and then permanent facilities, and closed in October 2025. Sanmina’s 2025 10-K shows the acquisition originally carried 1.46 billion USD of cash consideration and about 150 million USD in stock at closing, with the seller eligible for up to 450 million USD of contingent cash payments tied to performance over three years. By the March 2026 quarter, purchase consideration in the accounting note had been updated to 1.62 billion USD net of cash acquired and equity issued, with the acquired goodwill allocated entirely to IMS. This is the stage that genuinely changes the company’s fate, because it moves Sanmina closer to the point in the value chain where hyperscaler program timing, thermal design, liquid cooling, and U.S.-based NPI capacity start to matter as much as traditional EMS volume discipline.

Financial vertical review

The long-run financial record says Sanmina is a better cash business than a casual glance at the income statement might suggest, but also a working-capital-intensive one. From fiscal 2021 through fiscal 2025, operating cash flow ran 338 million, 331 million, 235 million, 340 million, and 621 million USD. Total capex or net purchases of property and equipment ran about 73 million, 130 million, 191 million, 111 million, and 147 million USD. That means free cash flow has generally been positive, sometimes strongly so, but never in a straight line. In 2023, free cash flow was squeezed by heavier inventory and working-capital needs. In 2025, the huge increase in deferred revenue and customer advances helped cash generation materially. This is a manufacturer whose cash profile can improve or deteriorate quickly with customer terms, program ramps, and inventory timing.

The margin record is modest by software or semiconductor standards, but relatively respectable for EMS. Gross margin improved from 7.2% in 2019 to 8.2% in 2021, reached 8.3% in 2023, 8.5% in 2024, and then held around 8.5% in 2025. Operating margin moved from 3.5% in 2019 to 4.2% in 2021, 5.1% in 2023, 4.4% in 2024, and 4.4% in 2025. The pre-ZT business was therefore not a low-quality margin collapse story. It was a gradual quality upgrade story. What ZT did was not invent good execution; it changed the revenue mix and temporarily pulled non-GAAP operating margin into the mid-6% range in Q1 and Q2 FY2026, though GAAP operating margin remained much lower because of amortization, inventory fair-value charges, and acquisition and integration costs.

The balance sheet is the clearest place where the transition shows up. As of 2025-09-27, Sanmina had 926 million USD of cash and 300 million USD of long-term debt. By 2026-03-28, cash had increased to 1.58 billion USD, but long-term debt had risen to 2.17 billion USD, with total assets jumping from 5.86 billion USD to 9.67 billion USD and goodwill rising from 30 million USD to 359 million USD. That is a manageable but meaningful change in risk profile. This is no longer a lightly levered mid-multiple manufacturer buying back stock out of routine cash generation. It is a manufacturer with a large acquisition-related debt stack whose leverage case depends on the durability of post-deal earnings and working-capital discipline.

Price and valuation history

Sanmina’s share-price history has moved through three broad chapters. The first was the late-1990s and 2000 hardware euphoria, when outsourced manufacturing names could trade on sheer demand momentum and scale dreams. The second was the long hangover, when the market treated these businesses as cyclical, margin-thin, and structurally underpowered. The third is the current rerating, driven first by better underlying execution and then by AI-infrastructure exposure via ZT. Public stock-history records still show the all-time high from 2000 and indicate that the current run, while powerful, remains below that dot-com peak.

Still, the current move is not trivial. By April 2026, Investor’s Business Daily described the stock as near its highest point since 2001, and by July 2026 Sanmina’s IR page showed a 219.41 USD prior close, while broader market coverage highlighted heavy investor anticipation ahead of the July 27 earnings call. The market is no longer pricing Sanmina as a plain EMS laggard. It is assigning some version of a scarcity premium for U.S.-based AI system manufacturing with AMD ecosystem leverage. The question is whether the premium should be persistent or event-driven.

Business model, moat, and industry cycle

Revenue structure and operating model

Sanmina reports two businesses: IMS and CPS. In fiscal 2025, IMS accounted for 6.56 billion USD of segment sales and 507 million USD of gross profit, while CPS accounted for 1.70 billion USD of sales and 236 million USD of gross profit. CPS is the smaller unit, but it is a larger profit contributor than its revenue share alone would suggest. That fits the strategic logic of vertical integration. If Sanmina can supply not just systems assembly but also higher-value components, memory, interconnect, optical, machining, and certain proprietary products, it gets a better claim on the profit pool than a pure box-builder would.

The cost structure is classic EMS with a twist. A great deal of cost is still variable: materials, subcontracting, labor at scale, logistics, and customer-driven inventory. That is why margins on ordinary production programs remain thin and why working capital matters so much. But the fixed-cost element rises sharply in the businesses that matter most to the current thesis: large AI-ready sites with power availability, liquid-cooling capability, systems integration, NPI capacity, test, and complex supply orchestration. That is one reason the ZT assets matter. They are not just extra square footage. They are a different kind of asset base. Sanmina’s deal-close release stressed large facilities in New Jersey and Texas with significant power and advanced liquid-cooling capability.

Sanmina’s operating leverage therefore cuts in two directions. In legacy EMS, scale helps but does not turn the business into a license model. In AI systems manufacturing, scale plus urgency plus capability can widen margins for a period, especially in NPI and fast deployment windows. The risk is that those windows can narrow once programs mature and customers re-open bidding or shift volume. Q2 FY2026 looks like a good example of this duality: huge revenue, very solid non-GAAP margin, but a management comment that some expected second-half shipments were pulled forward. The operating model is getting better, but it is not becoming smooth.

Moat

Sanmina’s moat is real, but it is not magical. The strongest element is manufacturing depth in complex, regulated, and mission-critical hardware. A company that can move from design support and prototyping into full-volume production, support PCBs, backplanes, enclosures, optics, precision machining, logistics, and after-market repair, and do so across a global footprint in 20 countries is meaningfully harder to replace than a narrow assembler. That breadth is one reason Sanmina has stayed relevant for decades across industrial, medical, defense, communications, and now cloud infrastructure.

The second moat is customer stickiness in qualification-heavy programs. In medical, defense, aerospace, and certain communications applications, once a supplier has been qualified, audited, and embedded in a product lifecycle, switching can be expensive in time, documentation, risk tolerance, and supply continuity. This is not quite a software switching-cost moat, but it is much better than a commodity manufacturing relationship. Sanmina’s continued reliance on existing major customers, and the fact that roughly half its revenue historically comes from its ten largest customers, is partly a risk and partly evidence of durable embedment.

The third moat is vertical integration, which matters more than branding in EMS. CPS generated 236 million USD of gross profit on 1.70 billion USD of sales in fiscal 2025, versus IMS gross profit of 507 million USD on 6.56 billion USD of sales. That tells you Sanmina is not merely a pass-through manufacturer. Some of the economic advantage comes from supplying more of the component stack and capturing higher-value engineering and product content. In an industry where many players can stamp metal and populate boards, profit often pools around the pieces few can do at quality and yield.

The weakest moat element is pure scale. Scale helps procurement, capacity utilization, and global account coverage, but by itself it does not guarantee pricing power. Flex and Jabil are larger. Celestica has a purer AI-infrastructure narrative. Super Micro owns more of the branded-system economics in AI racks. Sanmina’s edge is that it combines decent scale with real manufacturing breadth and, after ZT, a better seat at the AI table. That is a medium moat, not a fortress.

Management and governance

Jure Sola remains the central figure. He is chairman and chief executive officer, and the 2026 proxy shows he beneficially owned about 1.53 million shares, or 2.82% of the company, as of 2025-12-31. That is meaningful ownership, but not controlling ownership. The same proxy also shows BlackRock, Vanguard, and FMR as major outside holders, and total director-plus-executive ownership of about 3.10%. In governance terms, this is founder-led influence inside a still institutionally owned public company, not a controlled company.

On the positive side, the governance architecture has standard alignment tools: board stock-ownership guidelines, anti-hedging language, and a clawback policy updated after the SEC’s compensation-recovery rules. The proxy also states all named executive officers satisfied the ownership guidelines. Those are not deep moats, but they do matter in a business where margin swings and acquisition accounting can easily cloud economic reality.

The more awkward part is that related-party employment exists in management’s family circle. The proxy discloses that Jure Sola’s brother, daughter, and son worked at Sanmina and earned approximately 742 thousand USD, 391 thousand USD, and 209 thousand USD, respectively, in fiscal 2025. The company says such matters are reviewed under its related-person transaction policies. This does not by itself break the thesis, but it does justify a small governance discount relative to a cleaner large-cap peer set.

Execution credibility is mixed but broadly favorable. The company remediated previously identified material weaknesses by the end of fiscal 2025, according to its 10-K, and management’s internal-control assessment was effective at that date. The company also continues to buy back stock, including a new 600 million USD authorization with no expiration date announced in April 2026. At the same time, the real test of capital allocation is no longer buyback cadence. It is whether the ZT integration produces attractive through-cycle returns after debt service and contingent consideration.

Industry structure, cycle, and policy

EMS is a giant industry with hundreds of billions of dollars of revenue globally and generally mid-single-digit structural growth by most market-research estimates, but those averages hide very different profit pools. The low-value end of the market is crowded and price-sensitive. The better economics live where manufacturing complexity, supply-chain orchestration, regulatory qualification, custom power and cooling, and speed-to-ramp matter. Sanmina’s old strategy was to lean toward those better pockets. The ZT deal pushed it harder in that same direction.

This is a multi-cycle business. It is cyclical to macro demand, inventory corrections, and customer capex budgets. It is also exposed to a technology-iteration cycle because hardware platforms do not stand still, and now to an AI data-center capex cycle that is much more explosive than ordinary EMS demand. Flex says directly in its 2026 filing that demand for cloud and power infrastructure serving data-center applications is tied to AI, machine learning, and high-performance computing investment by hyperscalers, and that these capex cycles can be volatile. That same sentence could sit inside Sanmina’s risk section unchanged.

Policy and geopolitics matter through tariffs, export controls, defense rules, environmental liabilities, and regional manufacturing strategy. Sanmina’s own disclosures flag tariffs and trade policy as meaningful risks, while its global footprint across 20 countries gives it some ability to re-balance production. The current environment also favors suppliers that can offer U.S.-based manufacturing for strategic systems. The AMD partnership language is explicit on that point. Sanmina becomes more attractive when customers want geographic optionality, shorter deployment cycles, and less dependence on a single overseas chain. That is a real tailwind, though it comes with higher labor and capital intensity.

Horizontal competitor analysis

The peer set and what each company became

Sanmina has enough real comparables that this is clearly a multi-peer case rather than a one-off niche monopoly. The most relevant group is Flex, Jabil, and Celestica, with Super Micro as a useful boundary marker rather than a true apples-to-apples comp. Flex, Jabil, and Celestica all manufacture and integrate complex hardware at scale. Super Micro overlaps in rack-scale AI systems, but it captures more branded-platform economics and carries more direct product-cycle exposure than Sanmina does.

Flex became the broadest and most diversified of the group. Its fiscal 2026 filing says no customer accounts for more than 10% of annual revenue, its ten largest customers account for 45% of sales, and it generates demand from cloud and power infrastructure tied to hyperscaler AI capex. It also produced 1.06 billion USD of free cash flow in fiscal 2026 on a 4.9% operating-income margin. Customers pick Flex because it can absorb scale, run globally, and offer a more diversified mix than a narrower AI build-out story. Investors pay for that resilience, but they do not usually give Flex the same concentrated AI torque they award Celestica or, lately, Sanmina.

Jabil became a portfolio-shaping story. Its fiscal 2025 filing shows 41% of revenue from Intelligent Infrastructure, 40% from Regulated Industries, and 19% from Connected Living and Digital Commerce. That is a much cleaner mix than the old mental model of Jabil as an Apple-adjacent consumer-electronics contractor. Gross margin in 2025 was 8.9%, down from 9.3% in 2024, and the filing still warns that a reduction in business from one or more of its largest customers could materially hurt results. Customers pick Jabil because it combines enormous scale with a more curated sector mix than classic commodity EMS. The market values it as a high-quality operator, but not as a pure AI systems scarcity play.

Celestica became the highest-purity public “manufacturing plus platforms for AI data-center infrastructure” story in the comp set. Its 2025 filing shows total revenue of 12.39 billion USD, with the CCS segment at 9.19 billion USD, or 74% of sales. Three CCS customers each represented at least 10% of total revenue in 2025, at 32%, 14%, and 12%. Communications revenue rose 81% on data-center networking demand and switch ramps. That profile explains both why the market gives Celestica a premium and why investors should be careful when they use it as a comp for Sanmina. Celestica is what the market looks like when it is willing to ignore concentration because AI growth is strong enough.

Super Micro is the useful edge case. In fiscal 2025, it said the year-over-year increase in server and storage sales was primarily due to GPU and Super Racks, up 5.804 billion USD, including liquid-cooled and air-cooled servers tied to H200, H100, and B200 systems, with average selling price up 34%. That is closer to end-system economics than Sanmina’s model. Customers choose Super Micro for speed, product architecture, and platform delivery. Customers choose Sanmina when they need a manufacturing and integration partner with less direct branded exposure and more contract-manufacturing DNA. The overlap is real in rack-scale AI, but the business models are different enough that the valuation gap should not be mechanically arbitraged.

Peer data snapshot

† Sanmina equity value is estimated from the 2026-07-21 close and the 53.6 million shares outstanding disclosed on 2026-04-21. ‡ Flex’s own fiscal 2026 filings report revenue of 27.9 billion USD, up about 8% from fiscal 2025.

The business reason behind the table is more important than the table itself. Sanmina is cheaper in absolute size because it is smaller, more recently rerated, and still proving that its AI exposure is durable. Flex deserves diversification credit because no customer exceeds 10% of sales. Jabil deserves quality credit because its mix has shifted toward higher-value infrastructure and regulated end markets. Celestica gets the biggest AI scarcity premium because more of its revenue is already concentrated in cloud/networking programs, though that also means far heavier customer concentration. Super Micro trades less like an EMS company and more like a fast-moving system vendor whose economics are tied to product cadence and rack demand. Sanmina sits between the old EMS world and the newer AI-infrastructure world. That in-between identity is why the stock can look both attractive and risky, depending on which peer an investor anchors to.

Ecological niche

Sanmina’s niche is not leadership by sheer size. It is a challenger position built on manufacturing breadth, vertical integration, and increasing relevance to high-value hardware categories. Before ZT, it occupied the upper-middle ground of EMS: credible in industrial, communications, medical, and defense, but not the first name investors cited for cloud AI hardware. After ZT, it moved into a more valuable gap: a U.S.-based, large-scale manufacturing partner for rack- and cluster-level AI infrastructure with real NPI credibility and a broader in-house technology stack than many contract assemblers.

That niche gets stronger if AI deployment remains capacity-constrained, if customers keep valuing U.S. manufacturing optionality, and if Sanmina can convert AMD proximity into a broader hyperscaler supplier role. It gets weaker if AI system assembly commoditizes faster than expected, if customers re-balance volume toward larger incumbents, or if program timing proves too lumpy for the market’s current expectations. The most direct profit-pool competitor is not one single company. It is the combined set of established EMS giants and AI platform builders that want the same rack-and-cluster budget.

Current fundamentals and bull-bear divergence

What is happening in the last four reported quarters

Sanmina’s last four reported quarters are a story of two businesses living inside one ticker. Q3 FY2025 delivered 2.04 billion USD of revenue, 5.7% non-GAAP operating margin, and 1.53 USD of non-GAAP EPS. Q4 FY2025 produced about 2.1 billion USD of revenue and 1.67 USD of non-GAAP EPS, capping a year with 8.1 billion USD of revenue and 6.04 USD of non-GAAP EPS. Then the ZT close hit the numbers. Q1 FY2026 jumped to 3.19 billion USD of revenue with 6.0% non-GAAP operating margin and 2.38 USD of non-GAAP EPS. Q2 FY2026 rose again to 4.01 billion USD with 6.4% non-GAAP operating margin and 3.16 USD of non-GAAP EPS.

Cash conversion also strengthened materially. Q1 FY2026 operating cash flow was 179 million USD, and Q2 FY2026 operating cash flow was 399 million USD. For the first half of FY2026, operating cash flow reached 577 million USD and free cash flow 433 million USD after 144 million USD of capex. Working-capital metrics also improved through March 2026, with days sales outstanding falling to 50 from 60 and cash-cycle days improving to 51 from 57. This is important because one of the easiest ways for a hot manufacturing story to disappoint is for revenue to outrun cash. So far, that is not what Sanmina has shown.

Guidance is where the market’s caution should begin. After the blowout Q2, management guided Q3 FY2026 revenue to 3.2 billion to 3.5 billion USD and non-GAAP EPS to 2.55 to 2.85. For the full fiscal year it guided to 13.7 billion to 14.3 billion USD of revenue and 10.75 to 11.35 USD of non-GAAP EPS. The numbers are strong. The shape is lumpy. Q2 benefited from accelerated compute shipments moving forward from the second half. Bulls read that as faster adoption. Bears read it as evidence that the demand curve is still too timing-sensitive to capitalize at a premium multiple without a discount for volatility.

The company also signaled confidence in capital returns and industrial adjacencies. The board authorized a new 600 million USD repurchase program with no expiration date, and in December 2025 Sanmina announced a new Houston facility focused on medium-voltage distribution transformers, instrument transformers, and switchgear, with production expected to start in 2027 and initial customer commitments already in place. That energy expansion is not the main equity story today, but it matters because it shows management is still using its manufacturing DNA to pursue adjacent categories beyond AI.

What the market is trading right now

The current share price is trading three things at once. First, it is trading real earnings acceleration. Q1 and Q2 FY2026 changed the income statement in a way no narrative alone could. Second, it is trading an AI-infrastructure rerating, not merely an EMS upgrade. Market coverage ahead of the July 27 results explicitly frames Sanmina as an AI-driven hardware beneficiary. Third, it is trading event risk: the next print matters because investors want to know whether Q2 was the first step in a durable staircase or the highest step in a shipment pull-forward.

The danger is that these three forces do not last equally long. Earnings acceleration can persist. Narrative premiums often do not. Event-driven enthusiasm disappears fastest of all if guidance normalizes. That is why the July 27 call matters more than the backward-looking quarter itself. The stock already moved through a major rerating before the call. What investors need next is evidence about the run-rate after the spike.

The main bull and bear disagreements

The bull case rests on durability. Sanmina now has real manufacturing capacity and customer relevance in cloud and AI infrastructure, reinforced by AMD’s choice to retain design/customer-enablement capabilities while making Sanmina a preferred U.S.-based NPI manufacturing partner. The facilities acquired from ZT are purpose-fit for this market, with large power capacity and liquid-cooling capabilities. If that relationship extends beyond one initial program cycle and attracts broader hyperscaler work, the business mix has genuinely moved upmarket.

The second bull point is that this is not a weak legacy base being dressed up by one acquisition. Management said “core Sanmina” grew 7.3% year over year in Q2 FY2026. Communications networks and cloud and AI infrastructure were already strong in Q1 FY2026. That matters, because it means the old company was not falling apart while the new one arrived. It was already improving.

The third bull point is cash. First-half FY2026 operating cash flow of 577 million USD and free cash flow of 433 million USD suggest the earnings jump is not purely accounting. A manufacturer funding buybacks, carrying 1.58 billion USD of cash, and still generating strong cash from operations after a transformative deal deserves more respect than an ordinary low-margin assembler.

The main bear point is concentration. In the March 2026 quarter, two customers each represented at least 10% of sales, the top ten were about 70% of first-half sales, and the acquired ZT business is economically intertwined with AMD and a small hyperscaler universe. That is not fatal, but it makes the stock much more vulnerable to program rescheduling, qualification changes, or a shift in platform competitiveness at AMD or end customers.

The second bear point is timing. Management itself said Q2 included accelerated-compute shipments that had previously been expected in the second half. When a stock rerates on AI scarcity, that kind of pull-forward matters because it can make the new earnings base look more stable than it is. Q3 guidance below Q2 revenue is the first reminder that this market will not move in straight lines.

The third bear point is leverage plus contingent consideration. Long-term debt rose to 2.17 billion USD by March 2026, and the original acquisition agreement includes up to 450 million USD of contingent cash consideration over three years. If the acquired business performs very well, Sanmina owes more cash. If it performs poorly, the strategic rationale weakens. That is an unusual symmetry: both success and disappointment create capital-allocation pressure, just through different channels.

Valuation, risks, catalysts, and cross-synthesis summary

Historical and peer valuation context

At 219.41 USD on the latest confirmed pre-report close, Sanmina is no longer trading like a forgotten EMS stock. Using the finance tool’s trailing data, the stock was around 46x trailing earnings on 2026-07-23 intraday data, though that trailing figure still reflects a pre-ZT denominator and is therefore not a useful way to judge current economics. Against FY2026 guidance of 10.75 to 11.35 USD of non-GAAP EPS, the stock trades at about 19x to 20x that year’s guided earnings. That is a large premium to what old Sanmina historically deserved, but not obviously crazy if the post-ZT earnings base proves durable.

Peer valuation says the market is willing to pay up for AI-linked manufacturing stories, but the reasons differ. Flex, Jabil, and Celestica all trade on much richer multiples than the old EMS template would imply, with current finance-tool P/Es around 55x, 40x, and 41x respectively, though those also reflect each company’s accounting and cycle timing. Celestica in particular is the dangerous comp because it shows what a concentrated AI/networking story can trade at when investors believe growth is still early. Sanmina merits some premium to its own history because its business changed. I do not think it merits a blind copy-paste of Celestica’s narrative premium because its concentration, integration, and deal-financing questions are still fresher.

Cash-flow passthrough and absolute valuation

On a rough five-year view, Sanmina’s operating cash flow has exceeded accounting earnings in the aggregate, but free cash flow after capex has been closer to parity and highly sensitive to working capital. Operating cash flow was 338 million USD in 2021, 331 million USD in 2022, 235 million USD in 2023, 340 million USD in 2024, and 621 million USD in 2025; capex or net purchases of property and equipment ran roughly 73 million, 130 million, 191 million, 111 million, and 147 million USD across those years. From that pattern, I infer that operating cash flow has been around one and a half times net earnings over the cycle, while owner earnings after maintenance needs are closer to GAAP earnings in normal years. The headline lesson is that Sanmina converts profit to cash reasonably well, but the cash arrives unevenly.

For maintenance versus growth capex, the best estimate from the public record is that maintenance spend for the pre-ZT core looked closer to the 70 million to 100 million USD range, judging from 2020–2021 levels and the lower end of later annual spending, while the step-up above that appears tied to growth and portfolio repositioning, including ZT-related integration, capacity, and broader strategic investments. I therefore use an owner-earnings lens that does not simply capitalize GAAP net income and does not fully penalize growth capex as if none of it created future capacity. This is an estimate, not a reported company metric.

The valuation below is a research framework, not investment advice.

The business logic behind these ranges is simple. I do not think Sanmina deserves to be valued as if FY2026’s surge were obviously ephemeral. I also do not think it deserves to be valued as if the current AI mix were already as proven and as broad as Celestica’s cloud/networking exposure or as structurally differentiated as a branded systems vendor’s economics. The fair-value center has moved up. The right question is how far up.

Expectation gap and margin-of-safety recheck

The market is clearly pricing more than a one-quarter beat. It is pricing continued evidence that Sanmina’s AI and cloud-infrastructure exposure will survive the first burst of ZT integration and remain margin-accretive. The most likely expectation gap at the next earnings print is not Q3 revenue by itself. It is the Q4 guide, customer concentration color, and management’s language about whether the Q2 pull-forward was exceptional or increasingly normal. Those are the metrics most likely to change both valuation and narrative at once.

Against my conservative fair value of 172–184 USD, the current price offers no margin of safety. If I cut the base case’s owner-earnings assumption to 70% of plan, the fair-value center stops looking like 205–222 USD and starts looking much closer to the conservative band. Put differently, the most fragile assumption in the present narrative is not that Sanmina has AI exposure; it is that the new earnings plateau is both broad and repeatable enough to deserve a sustained premium multiple.

If earnings were simply flat around the FY2026 guided level for three years and the exit multiple settled near the base case, the annualized return from 219.41 USD would be modest at best. If the exit multiple settled nearer the conservative case, returns would be negative. That means the margin-of-safety verdict at the current price is not obvious. This is the classic “gooder company, harder price” problem. The company improved. The stock front-ran a lot of that improvement.

Risks, catalysts, and tracking dashboard

The biggest permanent-capital risks are specific. The first is customer concentration inside the AI business. Probability medium; impact high. The observable indicator is whether future filings continue to show two or more customers above the 10% threshold and whether management commentary increasingly centers on one ecosystem. The transmission path is straightforward: lower program volume or a platform shift hits revenue first, then factory utilization, then margins, then the narrative multiple.

The second is shipment timing risk masquerading as trend. Probability high; impact medium to high. The indicator is quarter-to-quarter revenue volatility relative to non-GAAP margin stability and management’s explanation of shipment movement. The transmission path is through expectation resets. A lumpy manufacturer can still be a good business, but a stock priced for smooth AI acceleration often cannot absorb lumpy evidence without multiple compression.

The third is leverage and contingent consideration. Probability medium; impact medium. The indicator is net debt movement, amortization schedule, covenant headroom, and any updates on contingent payments under the acquisition agreement. The transmission path runs through lower financial flexibility, slower buybacks, and higher sensitivity to a weaker down-quarter.

The fourth is legal and regulatory overhang from government-contract allegations. Probability low to medium; impact medium. The company disclosed an FCA-related DOJ investigation and an unsealed qui tam case alleging about 100 million USD of claimed fraud damages, with treble damages and other remedies sought, though Sanmina says loss is not currently probable or estimable and the DOJ declined to intervene at that time. The transmission path would be less through immediate operations and more through a valuation overhang, settlement risk, and reputational damage in defense-related business.

The fifth is governance quality around a still founder-centric leadership model. Probability low; impact medium. The indicators are related-party disclosures, succession planning, and whether governance issues remain limited to disclosed family employment rather than escalating into capital-allocation or control conflicts. Alone, this is not a thesis-breaker. Combined with concentration and leverage, it matters more.

The positive catalysts are clear as well: a Q3 report that shows margins staying in the guided 6.4%–6.9% range despite lower revenue than Q2, a Q4 guide that confirms the FY2026 step-up is not simply pull-forward, explicit evidence that AMD-related work is broadening into hyperscaler or OEM relationships beyond one launch wave, continued cash conversion, and faster leverage reduction than investors expect. The negative catalysts are equally clear: a guide cut, margin slippage below the mid-6% non-GAAP range, a more concentrated customer mix than investors assumed, or growing evidence that ZT’s economics are closer to volume manufacturing than high-value NPI after the first ramps.

These indicators matter because they separate business quality from stock excitement. Revenue tells you whether AI demand is real. Margin tells you whether the work is economically better, not just bigger. Cash flow tells you whether earnings are bankable. Customer concentration tells you whether the story is broadening or narrowing. The earnings date matters especially here because Sanmina’s current multiple is highly sensitive to new guidance. Sanmina’s investor-relations calendar and multiple recent market previews point to a Q3 FY2026 earnings call on 2026-07-27.

Cross-synthesis

Looking vertically, the capability Sanmina has genuinely proved is not invention in the consumer-tech sense. It is endurance plus adaptation. It survived the post-dot-com unwind, the limits of scale-for-scale’s-sake EMS economics, multiple hardware cycles, working-capital shocks, and the long dull period in which investors discounted the whole industry. In the process it became a tighter, more selective manufacturer with better regulated-market exposure, better vertical integration, and a habit of generating cash even when the revenue line was noisy. That is real competence, and it predates AI.

Its past success came partly from era tailwinds, partly from cyclical improvement, and partly from management discipline. The era tailwind today is AI data-center build-out. The cyclical piece is communications and cloud-hardware strength. The management piece is the willingness to use that healthier base to do a genuinely transformative deal. What has not yet been proved is whether the ZT acquisition gives Sanmina a durable new perch or merely a more exciting seat in an inherently volatile capex chain. That is the decisive distinction. The market is rewarding future success already, not just past success.

Horizontally, Sanmina’s edge versus most EMS peers is the combination of vertical integration, mission-critical manufacturing depth, and now a meaningful AI systems manufacturing position. Its weakness versus Flex is diversification, versus Jabil is scale and portfolio breadth, versus Celestica is already-proven AI concentration, and versus Super Micro is direct platform ownership. Those are not temporary disadvantages. They are structural features of where Sanmina sits in the chain. That is why I think the market’s biggest likely misjudgment today is not directionally wrong enthusiasm. It is the speed with which it may be assuming durability. A business can improve structurally and still have a stock that gets ahead of the proof.

For the next year, the core variables are Q4 guidance, customer concentration, cash conversion, and leverage reduction. For the next three years, the biggest question is whether Sanmina broadens beyond being mainly an AMD-enabled AI manufacturing story into a broader hyperscaler and high-power infrastructure platform. For the next five years, the real test is whether management can turn this step-change in relevance into a more stable return profile rather than a peak-cycle earnings spike. Sanmina becomes a better investment if the stock comes in, if FY2027 demand holds up after the first ZT excitement fades, and if leverage falls without sacrificing strategic flexibility. The original judgment should be re-examined if margins fall below 6% non-GAAP for two straight quarters, if customer concentration increases rather than broadens, if cash flow starts lagging earnings materially, or if legal/government-contract risk becomes more financially concrete.

Bull and bear reasons

Bull reasons:

• Sanmina now has genuine strategic relevance in cloud and AI infrastructure through the closed ZT acquisition and AMD preferred-NPI partnership, not just thematic exposure.
• The pre-deal business was already improving, with core Sanmina still growing 7.3% year over year in Q2 FY2026.
• First-half FY2026 cash generation was strong enough to support both integration and shareholder returns.
• CPS remains an important internal profit engine, which gives Sanmina better economics than a pure low-value assembler.
• U.S.-based AI-capable manufacturing capacity with power and liquid-cooling capability is scarce and strategically attractive.

Bear reasons:

• Q2 FY2026 included shipment pull-forward from the second half, which means the new revenue run-rate is not fully proven.
• Customer concentration remains high, and the acquired AI business is tied to a small customer universe.
• The deal materially increased leverage, raising the cost of being wrong on durability.
• The stock has already had a large rerating and is now priced much closer to premium AI-linked peers than to its own history.
• Legal overhang from the FCA/DOJ matter is not yet probable or estimable, but it remains unresolved.

Pre-mortem

One plausible 50% drawdown script is this: by early 2027, the market learns that Q2 FY2026 was the high-water mark of an urgent first-wave AI deployment cycle. Revenue settles back near 12.5 billion USD, non-GAAP operating margin falls toward the mid-5% range as NPI mix fades into volume production, and investors stop capitalizing Sanmina at about 20x forward earnings and revert toward 14x to 15x normalized earnings. A stock bought for an AI rerating then de-rates back toward a higher-quality but still cyclical EMS multiple. That combination can cut the equity in half without any insolvency narrative at all.

A second script is more concentrated. AMD’s AI platform gains do not translate into the broad, recurring factory load Sanmina needs, one or two hyperscaler programs pause or shift, and Sanmina’s disclosures begin to show heavier dependence on fewer customers at the same time debt remains elevated and contingent payments still hang over the deal. Revenue misses are then interpreted not as timing issues but as evidence that Sanmina is merely a second-order beneficiary inside someone else’s product cycle. The stock could compress sharply because both the earnings base and the narrative quality would be questioned at once.

Final research conclusion

Sanmina is a better company than the market thought it was two years ago. The legacy business had already become a steadier cash generator with more defensible vertical integration and a healthier mix than the stereotype of a bare-bones EMS contractor implied. The ZT acquisition then gave it something rarer: a direct route into rack-scale AI infrastructure manufacturing, backed by an AMD partnership and by facilities that actually fit the physical requirements of this market. The core fundamental debate is not whether Sanmina improved. It did. The debate is whether the current stock price already assumes too much confidence about how repeatable the new earnings base will be once the first wave of accelerated-compute shipments is behind it.

At the current price, I think Sanmina is ownable for existing holders but not yet attractive enough for aggressive new buying. The company has real upside if the AI mix broadens, margins stay in the mid-6% non-GAAP range, and cash flow keeps deleveraging the balance sheet. The stock, though, is no longer giving investors a comfortable buffer against the probability that FY2026 contains some pull-forward. What worries me most is not collapse. It is paying a near-premium multiple for a business whose new shape still needs two or three more clean quarters of proof. What would change my mind? A better entry price, or evidence after July 27 that the post-ZT demand base is both broader and smoother than Q2 alone could show.

【Company-profile scores】

• Fundamental quality: medium
• Growth: medium
• Moat: medium
• Financial soundness: medium
• Management credibility: medium
• Valuation attractiveness: low
• Risk level: medium
• Suitable investor type: event-driven / cyclical

【Investment rating】

• Rating: Hold
• One-line thesis: The business reset is real, but the stock already discounts a large part of FY2026’s AI-driven ZT uplift before durability is fully proved.
• 【Ideal Buy Price】138–148 USD Basis: this is roughly 20% below my conservative fair-value range of 172–184 USD, which assumes Sanmina keeps a higher post-ZT earnings base but loses some early AI urgency.
• Acceptable hold price: 174–247 USD
• Clearly overvalued price: 281–294 USD
• Current-price classification: acceptable hold
• Whether to wait for a better price: yes. I would prefer new entry below about 148 USD, or after evidence that FY2027 earnings power is broadening beyond the Q2 pull-forward. The opportunity cost of waiting is missing further upside if the July 27 call proves the step-change is fully durable.
• Target holding horizon: 1–3 years
• Expected annualized return: conservative about negative 7%; base about flat to negative 2%; optimistic about positive 6% to 8%
• Max-loss risk: about 45% to 55% in the pre-mortem case where AI program timing normalizes, margins drift back toward legacy EMS levels, and the multiple compresses toward a mid-teens earnings multiple
• Reassessment-trigger signals: if non-GAAP operating margin falls below 6.0% for two consecutive quarters; if a single customer becomes clearly dominant in revenue; if operating cash flow trails net income for two consecutive quarters without an obvious temporary working-capital reason; if net debt is not declining by FY2027 despite strong demand; if the FCA/DOJ matter becomes accrued or otherwise materially more concrete

【Valuation Range】

• current: 219.41 (close as of 2026-07-21)
• bear (conservative · ideal buy zone): [138, 148]
• base (fair · acceptable hold zone): [174, 247]
• bull (optimistic · above the clearly-overvalued line): [281, 294]

Research uncertainties and source note

The biggest blind spots are ordinary but important. Sanmina does not publicly break out the acquired ZT business as a clean standalone segment inside quarterly reporting, so the exact margin profile and customer mix of the acquired operations remain partly inferred from management commentary and consolidated results. I also did not find a primary archival filing in this research pass that let me state the 1993 IPO economics with the precision I would want. Finally, peer comparisons are inevitably imperfect because Celestica and Super Micro sit at different points in the value chain, and current sell-side estimate revisions were not treated as authoritative here unless echoed by company disclosures.

Primary sources used most heavily were Sanmina’s 2025 10-K, the March 2026 10-Q, the 2026 proxy, Sanmina and AMD transaction press releases, Sanmina earnings releases, and peer SEC filings for Flex, Jabil, Celestica, and Super Micro. Secondary sources were used mainly for price history and recent market narrative, not for core financial facts.

Other tickers mentioned

• FLEX.US: closest large diversified EMS peer and a useful benchmark for diversification, margins, and cloud-power exposure
• JBL.US: major EMS peer whose mix has shifted toward intelligent infrastructure and regulated industries
• CLS.US: the most relevant public premium-AI manufacturing/networking comp, with stronger AI concentration and heavier customer concentration
• SMCI.US: boundary comp for rack-scale AI systems economics and liquid-cooled server demand
• AMD.US: counterpart in the ZT transaction and strategic partner that retained design/customer-enablement teams while making Sanmina a preferred NPI manufacturing partner

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.