Flex Ltd: The CPI Spin-Off Is Real, But the Stock Has Already Front-Run the Rerating

From a fundamentals angle, the company is stronger than the old EMS stereotype. Flex generated 1.685 billion USD of operating cash flow and 1.06 billion USD of free cash flow in fiscal 2026, ended March with 2.4 billion USD of cash against 3.8 billion USD of borrowings, and kept buybacks large enough to retire 19.2 million shares for 944 million USD in the year. No customer contributed more than 10% of sales, though one customer represented 16% of receivables, almost entirely in CPI. The top ten customers accounted for 45% of sales, which is meaningful but not unusual in this industry. Solvency is not the weak point; mix sensitivity is, because the company is now partly exposed to a hot AI-capex wave and partly to slower, more ordinary end markets.

The right qualitative label is company in transition. That is not a dodge; it is the most accurate description of what shareholders actually own between now and the first quarter of calendar 2027. CPI is large enough and fast enough to change the growth profile, so Flex is no longer simply a mature cash machine. Most of the revenue and much of the operating base still sit inside the diversified manufacturing company, so this is not pure high-quality AI infrastructure either. And it is not just a spin-off story: the remaining company still has to prove that losing CPI does not turn it into a multiple orphan. The stock sits between those identities, which is what makes it analytically interesting and valuation-sensitive at the same time.

Company history and business model

Flex began as a far humbler business than the current market cap suggests. The company traces its origin to 1969, when Joe and Barbara Ann McKenzie started hand-soldering electronic parts onto printed circuit boards for Silicon Valley startups that needed overflow capacity. The 2019 annual report retells that origin in unusually plain language: a family business doing “board stuffer” work for fast-growing technology customers. That origin still matters because it explains the company’s instinct. Flex was never built around proprietary branded products. It was built around being the partner that absorbs manufacturing complexity for other people.

The first decisive turn came with geography. Flex moved early into Asia, especially Singapore, and then into larger multi-site industrial park models. The company’s own history points to Singapore as a crucial bridge between American customer demand and Asian manufacturing economics. In Singapore, it could combine lower cost, export orientation, and more responsive supply chains at a time when many OEMs still built too much in-house. That was the original moat: not brand, not patents in the usual sense, but geographically intelligent manufacturing scale.

The second decisive turn was institutional, not operational. Flextronics International Ltd. was incorporated in Singapore in 1990 to acquire Asian and certain U.S. operations of the predecessor business. The company later returned to public markets in 1994 on Nasdaq. The broad listing path is clear from Flex’s own history materials and later filings. The exact IPO pricing and proceeds, however, were not reliably retrievable from primary archived materials in this research session, so I do not use them in the investment case. That limitation does not change the bigger point: the listing gave Flex the balance-sheet flexibility to expand globally and then to make large acquisitions in the next growth phase.

The third stage was scale through consolidation. Flex spent the 2000s becoming one of the global titans of EMS. Its own historical materials note more than 75 acquisitions since 2000, and among the largest and most consequential was the Solectron transaction, which expanded footprint, capabilities, and customer relationships on a scale organic growth alone would not have matched. The same acquisition logic later reappeared in more focused form inside power and cooling.

The fourth stage was portfolio repair and sharpening. Flex changed its name from Flextronics to Flex in 2015, a symbolic shift that matched a broader effort to emphasize design, engineering, supply-chain management, and higher-value manufacturing rather than bare-board assembly. That same decade also showed management’s willingness to incubate and separate businesses. Nextracker was taken public in February 2023 and then fully spun to Flex shareholders in January 2024. The present CPI transaction is not an unprecedented corporate maneuver. It is the second major separation in a short period, and that cuts two ways: it proves Flex can execute separations, but it also reminds investors that portfolio management has become one of the company’s central value-creation tools.

Revathi Advaithi’s arrival as chief executive in 2019 is the next clear break in the story. Flex’s press release said she joined after senior operating roles at Eaton, where she had run a large electrical-sector business. That background showed up in the strategy. Under Advaithi, Flex put more emphasis on margin quality, portfolio discipline, and businesses where operational know-how could support differentiated engineering and faster secular growth. Michael Hartung, a long-time Flex executive, was first elevated to lead Agility Solutions in 2020 and then became president and chief commercial officer in 2024, consolidating go-to-market leadership across the former Agility and Reliability segments. Kevin Krumm joined as chief financial officer effective January 2025. The management team today therefore mixes long internal operating continuity with outside financial and industrial experience.

The most important strategic move of the current era was inward, not outward: building a data-center power and thermal stack inside a company most investors still filed under “contract manufacturer.” The company bought Anord Mardix in 2021 for 540 million USD to strengthen critical power solutions in the data-center market. It bought Crown in late 2024 for about 319 million USD to deepen U.S. critical power capability. It bought JetCool in late 2024 for 53 million USD to add direct-to-chip liquid cooling. Then, in May 2026, it completed the 1.1 billion USD acquisition of Electrical Power Products, funded in part with a new 1.45 billion USD credit facility. Those are not random bolt-ons. They form a puzzle that becomes coherent only if one accepts that Flex was constructing a higher-multiple digital-infrastructure company inside the shell of a global EMS operator.

That construction is now visible in the current business model. Flex’s fiscal fourth-quarter investor deck shows the company’s actual present reporting structure: ITS, RMS, and CPI. ITS generated 11.1 billion USD of fiscal 2026 revenue and 596 million USD of adjusted operating income. RMS generated 10.2 billion USD and 611 million USD. CPI generated 6.6 billion USD and 610 million USD. This is the unusual picture. CPI is only 24% of revenue, but it produced operating income roughly equal to RMS and above ITS. In other words, the quality difference is visible in segment economics, not something the market is hallucinating.

The spin presentation then maps those economics into the future legal structure. SpinCo will include critical power products, embedded power products, thermal management products, and cloud, with a fiscal 2026 revenue mix of 31% power products and 69% thermal management products and cloud. The remaining Flex will include RMS and ITS, described more simply by end market: health, industrial, automotive, communications, and lifestyle. That split makes the businesses easier to understand than the legacy segment language ever did. SpinCo is being marketed as a critical digital-infrastructure supplier. RemainCo is a diversified advanced manufacturer.

Operationally, Flex still runs like an EMS company, sitting between customers and a global manufacturing network of more than 100 facilities across roughly 30 countries. It is working-capital intensive, with receivables, inventories, contract assets, and payables all large relative to earnings, and it wins by being difficult to replace, not impossible to copy. Customers use Flex to compress time to market, reduce factory ownership, shift geography, manage component sourcing, and increasingly to integrate power and cooling design into systems deployment. That last point is the new one. Power and thermal are no longer just line items. In AI racks, they have become gating constraints. Flex is trying to own that bottleneck.

Financially, the vertical record is mixed in the right way. Revenue has not been a straight growth line. Fiscal 2025 revenue declined 2%, then fiscal 2026 rose 8% to 27.9 billion USD. What improved more steadily was margin quality. Fiscal 2026 gross margin rose to 9.2% from 8.4%, helped by mix shifting toward higher-margin cloud and power, and adjusted operating margin reached 6.3%. Cash conversion remained healthy: operating cash flow was 1.685 billion USD and free cash flow 1.06 billion USD. Debt rose again after March 2026 because the company fully drew a 1.45 billion USD facility to finance the EP² acquisition, which is a reminder that the spin is being built on an actively managed balance sheet rather than on a pristine, static one.

What gives Flex a real moat is narrower than management rhetoric but stronger than skeptics sometimes admit. The first moat is manufacturing scale in the right geographies. The company generated 50% of fiscal 2026 sales in the Americas, 30% in Asia, and 20% in Europe. Mexico alone accounted for 25% of revenue, the U.S. 19%, China 16%, and Malaysia 11%. That footprint matters to customers trying to diversify away from single-country risk while preserving cost efficiency. The second moat is customer stickiness in regulated and operationally demanding end markets, especially healthcare, industrial systems, and now data-center power. The third moat is systems integration in power and thermal, where acquisitions have added capabilities that are worth more together than separately. Flex has patents and engineering know-how, but the practical moat is execution across design, supply chain, manufacturing, and lifecycle services.

The moat is not universal. Brand is irrelevant. Network effects are weak. Switching costs are meaningful for regulated products and deeply integrated programs, but much lower in more commoditized manufacturing work. Pricing power is also uneven. The company’s own revenue-recognition discussion makes clear that many customer contracts include price adjustments, cost-sharing provisions, committed price reductions, and rebate mechanisms. That is classic EMS economics. The reason CPI matters so much is precisely that it pushes more of the portfolio into places where Flex has a better claim on engineering value and can defend gross margin more effectively.

Governance is better than the old Singapore-domiciled manufacturing discount might imply, but it is not frictionless. The board has been shareholder-friendly on buybacks: Flex repurchased 19.2 million shares for 944 million USD in fiscal 2026, and available authorization remained at 1.1 billion USD as of year-end. Basic weighted-average shares fell from 435 million in fiscal 2024 to 391 million in fiscal 2025 and 371 million in fiscal 2026. There is no obvious dual-class or family-control issue. The awkward item is the Amazon warrant issued in August 2025 for up to 3.86 million shares at 51.29 USD, which is strategically understandable but mildly dilutive. The bigger governance question is strategic: whether management can execute a second major separation in two years without over-promising the rerating and underestimating standalone costs.

Industry and competition

The easiest way to misunderstand Flex is to force all of it into one peer set. The diversified manufacturing company belongs with Jabil, Celestica, Sanmina, and, in a broader global sense, Hon Hai/Foxconn. SpinCo belongs in a different conversation, one that includes Vertiv, nVent, Eaton, and to a lesser extent the AI-infrastructure portions of Celestica. The reason the spin was announced at all is that the market pays very different multiples for those two categories, even when the factory skills underneath overlap.

The traditional EMS industry is big, fragmented at the long tail, and structurally lower margin than branded electronics or specialized power equipment. Third-party market-research estimates put the global EMS market in the mid-600-billion-USD range in 2026 and growing at a mid-single-digit rate. As a rough scale check, that means Flex’s 27.9 billion USD of fiscal 2026 sales still makes it large, but not dominant enough to dictate industry terms. The profit pool in classic EMS does not sit in final assembly alone. It sits in program complexity, geographic flexibility, quality systems, regulated manufacturing, component procurement discipline, and customer concentration management.

That is why the diversified peer comparison matters. Jabil has increasingly tilted toward “intelligent infrastructure,” healthcare, automation, and other higher-value verticals, and management’s fiscal 2026 outlook called for about 31.3 billion USD of revenue and 5.6% core operating margin. Sanmina’s January 2026 release showed 8.13 billion USD of revenue, 5.7% non-GAAP operating margin, and explicit strength in communications networks and cloud and AI infrastructure. Celestica has become the clearest market favorite in the EMS-adjacent AI category: its fiscal 2025 revenue rose 28% to 12.4 billion USD, and fourth-quarter 2025 adjusted operating margin hit 7.7%. This is what each company became. Jabil is a diversified large-scale operator with more emphasis on automation and industrial infrastructure. Sanmina is a focused high-complexity manufacturer with steadier, narrower scale. Celestica is the market’s preferred AI-platform EMS proxy because its hardware platform solutions business is seen as closer to the data-center capex flow than generic contract manufacturing.

Flex’s remaining company after the spin will sit between Jabil and Sanmina in business quality and below Celestica in market excitement. It is not an insult so much as the likely outcome of the portfolio map. RemainCo will still have meaningful healthcare, industrial, and automotive exposure, plus communications and lifestyle. RMS and ITS together produced 21.3 billion USD of fiscal 2026 revenue and 1.207 billion USD of adjusted operating income before corporate costs. Those economics are respectable, especially with buybacks and global footprint support, but they do not obviously justify the kind of multiple the market gives to a pure AI-infrastructure supplier.

The data-center power and thermal peer set looks different because the profit pool is different. Vertiv is the cleanest comp in market narrative terms: it sells power, thermal, and digital-infrastructure equipment directly into data centers and communication networks. In the first quarter of 2026, Vertiv reported 20.8% adjusted operating margin and raised full-year guidance. nVent reported first-quarter 2026 sales of 1.24 billion USD, up 53%, with about 40% organic order growth and strong data-center demand led by liquid cooling. Eaton is much broader industrially, but its power-management positioning makes it a useful upper-quality reference. These are not contract manufacturers in the classic sense. They capture more engineering, product, and software value per dollar of revenue, and the market rewards them accordingly.

The market’s current pricing shows that difference starkly.

Source basis: company filings, earnings releases, and live market data.

The table looks odd if one expects Flex to screen as a cheap manufacturer. It does not. The market is already paying up for its transition. Flex’s market cap is above Jabil’s despite lower revenue, and above Sanmina’s by a wide margin. It is only moderately above Celestica’s despite much larger revenue because Celestica’s AI-exposed mix has already won a high multiple. That is the right way to read the numbers. The market is not ignoring CPI. It is trying to pre-price it.

The second peer table is the one that matters for SpinCo.

Source basis: company materials and live market data.

The spread is the entire spin thesis. SpinCo’s fiscal 2026 margin of 9.2% is still far below Vertiv or nVent, but it is much closer to them in business direction than to diversified EMS peers. Management’s fiscal 2027 guide for consolidated Flex says margin expansion should be driven by recouped fiscal 2026 investments, with more than 100 basis points of improvement inside CPI alone. If that actually happens and SpinCo can show a credible path toward low-teens operating margin over a few years, the market will not value it on a normal EMS template. If that does not happen, the rerating case becomes more fragile.

Flex’s ecological niche, then, is unusually specific. Today it is a bridging company: large enough and global enough to win diversified manufacturing programs, but now specialized enough in power and cooling to matter in the AI rack buildout. That niche gets stronger if data-center demand stays solid and customers prize integrated power-plus-cooling execution. It gets weaker if AI infrastructure spend normalizes abruptly, if liquid-cooling standardization compresses differentiation, or if hyperscalers decide to dual-source more aggressively across specialist vendors and lower-cost assemblers.

Current fundamentals and valuation

The latest four-quarter picture is clear. Flex entered fiscal 2026 with strong momentum, then accelerated. In Q1 fiscal 2026, revenue was 6.575 billion USD, adjusted operating margin was 6.0%, and adjusted EPS was 0.72; management’s original full-year guide was 25.9 to 27.1 billion USD of revenue and 6.0% to 6.1% adjusted operating margin. In Q2, revenue rose to 6.804 billion USD, margin stayed at 6.0%, and management raised fiscal 2026 revenue guidance to 26.7 to 27.3 billion USD and operating margin to 6.2% to 6.3%. In Q3, revenue reached 7.058 billion USD, adjusted operating margin improved to 6.5%, and the company again raised full-year guidance, this time to 27.2 to 27.5 billion USD and 6.3% margin. Fiscal 2026 finally closed at 27.9 billion USD of revenue and 6.3% adjusted operating margin, above even the raised third-quarter range.

The business drivers behind that sequence were not subtle. Q1 still spoke in the older Reliability and Agility language and already highlighted strong demand in data-center power. Q2 said Flex expected to reach its fiscal 2027 goal of 6%+ adjusted operating margin one year early. Q3 said revenue remained driven by strong AI demand, with margin expansion coming from favorable product mix and disciplined cost management. By year-end the company had formalized the story into a segment map: CPI up 38% year over year, RMS up 5%, ITS down 2%. The growth engine was CPI. Everything else was either stable or respectable.

Fiscal 2027 guidance is what the market is trading now, alongside the spin. Flex guided to revenue of 32.3 to 33.8 billion USD, adjusted operating margin of 7.0% to 7.1%, adjusted EPS of 4.21 to 4.51, capex of 1.4 to 1.6 billion USD, and roughly 60% free-cash-flow conversion. Management explicitly said this guide does not give effect to the planned spin-off of CPI. The margin step-up matters more here than the revenue line: the real question is whether it proves that CPI is scaling into a business with more operating leverage than a normal EMS substrate.

The market narrative today is split between real fundamentals and a hotter overlay. The real fundamentals are growth in cloud infrastructure, power products, thermal management, and the acquired expansion of critical-power capability. The overlay is AI scarcity. Anything with credible exposure to rack density, cooling bottlenecks, or power availability has been rewarded by investors, sometimes aggressively. Flex is not a fake participant in that theme. Its own acquisitions, segment growth, and guidance support it. The question is temperature. A stock can have a valid fundamental story and still price in too much of it too early.

Cash-flow passthrough is stronger than headline GAAP earnings suggest, but it needs interpretation. In fiscal 2026, operating cash flow was 1.685 billion USD against net income of 880 million USD, a ratio of about 1.9 times. In fiscal 2025 it was 1.505 billion USD against 838 million USD, about 1.8 times. In fiscal 2024 it was 1.326 billion USD against 1.245 billion USD, about 1.1 times. The recent uplift mostly came from working-capital mechanics, especially payables growth, not just from pristine earnings conversion. I therefore do not simply capitalize operating cash flow as if all of it were owner earnings. A more sensible owner-earnings estimate uses maintenance capex near depreciation and treats the excess fiscal 2026 capex as growth spend tied to data-center and footprint expansion. With depreciation at 457 million USD and total capex at 633 million USD, a rough maintenance capex assumption around 450 million USD gives owner earnings in the area of 1.2 to 1.25 billion USD for fiscal 2026. That is materially above GAAP net income, but it is also less flattering than using full operating cash flow.

On that owner-earnings basis, the stock is still not obviously cheap. At 47.9 billion USD of market cap, the implied owner-earnings yield is roughly 2.6%, using 1.25 billion USD as a round estimate. Even if one prefers fiscal 2027 adjusted EPS guidance, the stock trades around 29 times the midpoint. Against a U.S. 10-year Treasury in the mid-4% area in late July 2026, that leaves little classical margin of safety for a company whose highest-growth asset has not yet been fully carved out.

The cleanest absolute valuation method here is a sum of the parts, not a single blended multiple. My framework values RemainCo on forward earnings and SpinCo on forward sales, because the remaining diversified manufacturer will likely be judged on cash generation and steadier growth, while SpinCo will initially be judged on growth rate, margin progression, and category placement among AI-infrastructure comps. The assumptions are deliberately simple and explicit.

Scenario analysis within a research framework, not investment advice. Model uses fiscal 2027-style forward assumptions anchored to company guidance and peer multiple ranges.

The numbers say the stock is not plainly mispriced on the upside or downside. It is above my conservative value, below my optimistic value, and near but below my base case. That is usually where a good company with a live corporate-action catalyst trades when the market already believes the strategic story but still lacks the final documents. The expectation gap is narrower than bulls often claim. Investors are not waiting to discover CPI. They are waiting to discover how much of the current AI premium survives legal separation and standalone disclosure.

The historical valuation center has clearly shifted. Old-cycle Flex was often treated as a mid-single-digit-margin contract manufacturer with buyback support. Current-cycle Flex is treated as a transition asset with a hidden infrastructure-growth business. That explains why peer comparison is less comforting than it first appears. Yes, Flex still has lower margin than pure digital-infrastructure names. But the market is no longer pricing it against old EMS history alone. That is why “cheap versus peers” is the wrong test. The right test is whether the current blended valuation leaves enough room for stumble risk. I do not think it leaves much.

The independent margin-of-safety check comes out lukewarm. The current price sits at a premium to my conservative case, so the classical margin of safety is zero at today’s level. The most fragile assumption in the base case is the multiple applied to SpinCo, not the revenue line. If the base-case SpinCo valuation multiple is cut to 70% of that assumption, the base-case equity value falls into the low-120s per share, roughly the area where the stock already traded before the latest rebound. If earnings were flat for the next three years and the multiple compressed toward a more ordinary high-teens forward P/E, long-term returns from today’s price would be unexciting. Margin-of-safety sufficiency verdict: not obvious.

Risk, catalysts, tracking indicators, uncertainties, and sources

The largest business risk is that AI data-center infrastructure demand stays real but becomes less special. Flex’s CPI segment benefited from a rare period in which power density, cooling complexity, and deployment urgency all rose at once. That combination lifted both volume and margin. If hyperscaler spending normalizes, if rack architectures standardize faster than expected, or if customers spread sourcing across more suppliers, CPI can still grow and yet disappoint the multiple. Probability is medium. Impact is high because the stock narrative is now built around this business. The observable indicator is CPI growth relative to the company’s fiscal 2027 margin promises and to order commentary from Vertiv, nVent, and Celestica.

The second business risk is that the remaining Flex becomes less interesting than management suggests. The company says ex-SpinCo Flex should be positioned for low-to-mid-single-digit growth. That is plausible, but RemainCo will lose the fastest-growing business and may also inherit more of the “ordinary EMS” discount. Healthcare and industrial are useful stabilizers, but communications, automotive, and lifestyle are not immune to softer demand. Probability is medium. Impact is high, because a disappointing RemainCo multiple would offset much of any SpinCo rerating. The indicator is how management frames standalone growth and buyback policy once separation documents provide carve-out detail.

The third risk is financial and structural: debt allocation and separation friction. The spin presentation says Flex targets an investment-grade balance sheet and SpinCo targets investment-grade metrics, but it does not yet specify the final capital structure. The company had 3.8 billion USD of debt and 2.4 billion USD of cash at March 31, 2026, then fully drew a 1.45 billion USD credit facility used for the EP² acquisition. The exact debt split, standalone costs, tax leakage safeguards, and retained-stake mechanics will matter. Probability is medium. Impact is high. The key indicator is the Form 10 and any financing disclosure later in 2026.

The fourth risk is working capital. Flex is a fundamentally working-capital-heavy company. In fiscal 2026, accounts receivable, contract assets, inventories, and other assets all grew, offset by a very large increase in accounts payable. That supported cash generation, but it also means reported cash conversion can flatter the business at peak ramp periods. If customer mix changes or payment terms tighten after the spin, free cash flow can look weaker than EPS. Probability is medium. Impact is medium to high. The indicator is quarterly movement in receivables, inventories, payables, and free-cash-flow conversion versus guidance.

The fifth risk is geopolitical and policy noise, which is a permanent feature of Flex’s operating model rather than a one-off headline. The company’s 10-K flags U.S.-Mexico-Canada Agreement review risk, tariff and trade-policy uncertainty, export controls, and also a newer issue: data-center development can be delayed by power availability, water use, and permitting. In other words, the same grid and cooling bottlenecks that help SpinCo’s sales can also slow customer deployment timetables. Probability is medium. Impact is medium. Indicators include data-center permitting headlines, power-availability commentary from hyperscalers, and any tariff pass-through deterioration.

The positive catalysts are straightforward. The biggest is clean progression from announcement to filing: transaction structure confirmed, management assignments fixed, debt allocation credible, and timing still on track for the first quarter of calendar 2027. The next is evidence that CPI’s FY27 margin step-up is actually happening. After that comes proof that RemainCo can still grow and still buy back stock aggressively enough to look like a disciplined cash machine rather than an ex-growth orphan.

The negative catalysts are just as clear. A delayed or restructured spin would hit both the event-driven and fundamental camps. A guide-down on CPI margins, especially after management’s explicit promise of more than 100 basis points of margin recoup in CPI, would damage credibility quickly. So would any sign that one large CPI customer is stretching payments or slowing deployment, because 16% of receivables already sit with a single customer almost entirely inside CPI.

A small dashboard is more useful here than a long watchlist.

Source basis for dashboard: company filings, earnings materials, transaction presentation, and live market data. The exact next earnings timestamp should be rechecked on Flex IR before trading because event-page extraction was limited in this research session.

Research uncertainties are concentrated, not broad. First, the final debt allocation, pension and tax-sharing mechanics, and retained-stake treatment for the spin have not yet been disclosed in the detail needed for a full carve-out model. Second, the exact fiscal 1994 IPO pricing and gross proceeds were not reliably retrievable from primary archived filings in this session, so the listing-path discussion uses only facts I could verify with confidence. Third, standalone public-company cost dis-synergies for both entities are still unknown. Fourth, CPI’s future margin ceiling is inferential today; the market has good reasons to expect expansion, but the company has not yet published long-form standalone historical carve-out statements.

Principal sources used in this report were Flex’s fiscal 2026 Form 10-K, fiscal 2026 quarterly earnings presentations, the May 2026 spin-off announcement and investor presentation, selected Flex acquisition releases, U.S. Treasury rate data, Reuters coverage of the spin and bond-market backdrop, and peer company filings, earnings releases, and live market data for Jabil, Celestica, Sanmina, Vertiv, nVent, and Eaton.

Cross-synthesis summary

Flex’s long history proves one capability above all others: it knows how to turn manufacturing complexity into customer dependence. That sounds abstract until the company shifts from assembling industrial and healthcare products to solving power density and thermal-management problems inside AI data centers. The through-line is not “AI”: it is that Flex wins when customers need a partner that can combine design, supply-chain control, geographic flexibility, and scaled production without forcing them to own all the factories and operational headaches themselves. That capability was built over decades, long before CPI existed, and it is the reason CPI grew fast enough to become separable in the first place.

Past success came from several sources, but not equally. Geography mattered, and so did cycles, acquisitions, and management discipline; luck mattered less than bulls or bears usually claim. Flex’s earlier eras were shaped by outsourcing growth and the globalization of electronics manufacturing. The Advaithi era changed the mix of success factors. Since 2019, the company has been more selective about where it wants to compete, more comfortable pruning or separating businesses, and more focused on raising margin quality instead of simply defending scale. The fiscal 2026 results show that this was not just branding: the company delivered 6.3% adjusted operating margin and segment-level proof that CPI has fundamentally better economics than the rest of the portfolio.

Those success factors are still present, but not in identical form. The geography and supply-chain engine, the acquisition engine, and the buyback discipline all remain. What changes after the spin is the proportions. The future SpinCo will rely much more on product-and-systems positioning in power and thermal, where it has stronger differentiation but also faces richer, better-known peers. The future RemainCo will rely more on program stickiness, regulated manufacturing, and operating discipline. That means today’s shareholders are being asked to underwrite two businesses whose strengths are connected but not the same.

That is why the horizontal conclusion matters. Versus Jabil and Sanmina, Flex’s diversified base is credible, cash-generative, and globally useful, but not obviously superior enough to command a lasting premium without CPI. Versus Celestica, Flex has been later to win the market’s imagination, but CPI gives it a better chance than traditional EMS ever would have. Versus Vertiv, nVent, and Eaton, SpinCo has the right direction of travel but not yet their margin structure or market status. The real advantage, then, is not that Flex is best in class at all of these things. Management built a bridge from one class into another, and that bridge is both the source of upside and the source of valuation risk, because bridges are transitional by nature.

The market is pre-spending future success, but not absurdly. At 127.39 USD, Flex is not priced like a sleepy assembler: it is priced like a company whose most exciting business is likely to get a better multiple after separation and whose remaining business will still be stable enough to avoid a collapse in blended value. That expectation is plausible. It is not free. Investors are already accepting a high trailing earnings multiple and a still-demanding forward multiple because they assume the documents to come will validate the story management has drawn in broad strokes. If those documents are clean, current holders can still do well. If those documents are messy, the downside does not rely on a collapse in AI spending. It only requires a lower-than-hoped multiple bridge between EMS and infrastructure.

What the market is most likely misjudging right now is how easily the sum of the parts will be recognized, not the existence of CPI. In many spin-offs, the hardest part is not getting the deal done so much as establishing two stable valuation centers afterward. SpinCo may well trade at a premium to old Flex, but RemainCo may also trade at a discount to what bullish spreadsheets assume if investors decide it is merely good, not special. This is why I do not think the stock deserves a simple “cheap because hidden asset” label. The hidden asset is no longer hidden. The unresolved variable is the shape of the two eventual valuation boxes.

For the next year, the critical variables are transaction execution, CPI margin progression, and whether management can keep consolidated FY27 guidance credible while preparing the spin. For the next three years, the key question is whether SpinCo can demonstrate enough standalone quality to move from high-single-digit to low-teens operating margin, while RemainCo proves it can still compound earnings through mix, operational discipline, and buybacks. For the next five years, the real test is whether Flex has created two durable franchises or merely extracted a temporary multiple opportunity from a hot infrastructure cycle.

The investment becomes better under three conditions. First, the stock falls back toward a true margin-of-safety zone, which for me means a level in the low-90s rather than the current high-120s. Second, the Form 10 and subsequent transaction materials show a sensible debt split and manageable standalone dis-synergies. Third, quarterly disclosures prove that CPI’s promised margin lift is real even before the market gets the full listing of SpinCo. The investment should be re-examined quickly if any two of the following occur together: transaction timing slips past the first quarter of calendar 2027, CPI margin improvement stalls, or peer AI-infrastructure multiples compress sharply.

【Bull reasons】

• CPI generated 6.6 billion USD of fiscal 2026 revenue, grew 38% year over year, and posted a 9.2% adjusted operating margin, which is already meaningfully better than the rest of Flex.
• The announced separation gives investors a direct path to re-rate the CPI business against power and thermal infrastructure peers rather than against old-line EMS names.
• Management’s fiscal 2027 guide implies another step up in profitability to 7.0%–7.1% adjusted operating margin and 4.21–4.51 USD of adjusted EPS, without yet giving effect to the spin.
• Flex still throws off meaningful cash and buys back stock aggressively, having retired 19.2 million shares for 944 million USD in fiscal 2026.
• The acquisition chain from Anord Mardix through EP² built a more complete power-and-cooling stack than Flex had five years ago, which makes the CPI story harder to dismiss as mere assembly exposure.

【Bear reasons】

• The stock already trades as a transition premium asset at about 55 times trailing GAAP EPS and roughly 29 times the midpoint of fiscal 2027 adjusted EPS guidance.
• The final spin capital structure is still incomplete: retained stake, debt allocation, and standalone policies are only partially disclosed today.
• RemainCo loses the highest-growth business and may be valued closer to ordinary diversified EMS peers than bull models assume.
• One customer represented 16% of receivables at year-end, almost entirely in CPI, which is a manageable but real concentration risk if deployment timing shifts.
• Cash conversion has been helped by working-capital dynamics, especially a large rise in payables, so headline operating cash flow should not be capitalized too generously.

【Pre-mortem】

A plausible 50% drawdown script does not need an AI crash. The more realistic version is this: by late 2026 the Form 10 reveals a heavier-than-expected debt load and standalone costs at both entities, while CPI margin in the second half of fiscal 2027 fails to improve as promised because hyperscaler orders are still growing but pricing and deployment mix are weaker. The market then stops valuing SpinCo on a 3.5x to 4.0x sales logic and falls back toward roughly 2.0x to 2.5x forward sales, while RemainCo is marked closer to a normal EMS earnings multiple. In that script, a stock already carrying a transition premium could plausibly halve.

A second script is more cyclical. AI infrastructure demand remains decent, but power-availability bottlenecks, permitting delays, and customer digestion push large installs to the right. Vertiv and nVent guide down together, the peer basket derates, and Flex loses the narrative support that allowed investors to look through the mixed portfolio. If the market decides CPI is a good business but not a rare one, and simultaneously decides RemainCo is stable but unexciting, the blended equity value can compress much faster than the underlying revenue would suggest.

Flex is worth owning only with discipline about price. The company has done real work: segment economics improved, the CPI platform is large enough to matter, and management made the decisive move to separate the fast-growth infrastructure business from the diversified manufacturing base. Those are not cosmetic changes. They reshape how the company should be analyzed. What they do not do is guarantee a cheap entry today. The present price already assumes a fair amount of successful separation and a reasonably constructive market for AI-infrastructure assets.

My conclusion is therefore straightforward. Flex is a stronger company than the old EMS label implies, but the stock is no longer a bargain built on neglect. It is a live special situation with credible upside and limited margin of safety. I would rather own it after either cleaner transaction disclosure or a more forgiving entry price. What would change my mind positively is not another thematic headline. It would be hard evidence that CPI margins are stepping up as promised and that RemainCo can still be framed as a disciplined compounder rather than as the ex-growth leftovers.

【Company-profile scores】

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

【Investment rating】

• Rating: Hold
• One-line thesis: CPI is real and separable, but today’s price already capitalizes much of the spin-off rerating before final carve-out details are public.
• Three price signals:
  • 【Ideal Buy Price】88–94 USD Basis: at least a 20% discount to my conservative SOTP value, which is about 110 USD per share.
  • Acceptable hold price: 116–156 USD
  • Clearly overvalued price: 179 USD and above
• Current-price classification: acceptable hold
• Whether to wait for a better price: yes. A buy becomes more attractive below the mid-90s, or at a somewhat higher price only if transaction documents show a cleaner-than-feared debt split and lower standalone dis-synergies. The opportunity cost of waiting is missing a successful initial rerating after a clean Form 10 process.
• Target holding horizon: 1–3 years
• Expected annualized return:
  • Conservative: about -14%
  • Base: about +7%
  • Optimistic: about +28%
• Max-loss risk: about 50% in a combined script of spin slippage, lower-than-promised CPI margin lift, and peer-multiple compression.
• Reassessment-trigger signals:
  • If CPI margin fails to improve by at least 50 basis points over the next two to three reported periods.
  • If transaction timing moves materially beyond Q1 calendar 2027.
  • If free-cash-flow conversion falls below 50% without a clear growth-capex explanation.
  • If one CPI customer rises above 20% of receivables.
  • If Vertiv and nVent both cut growth expectations tied to AI data-center power or thermal demand.

【Valuation Range】

• current: 127.39 USD (close as of 2026-07-21 U.S. market session)
• bear (conservative · ideal buy zone): [88, 94]
• base (fair · acceptable hold zone): [116, 156]
• bull (optimistic · above the clearly-overvalued line): [179, 195]

Other tickers mentioned

• JBL.US: closest large diversified EMS peer for the remaining Flex business.
• CLS.US: most relevant AI-exposed EMS reference, useful for understanding how markets reward data-center adjacency.
• SANM.US: high-complexity manufacturing peer and a cleaner read on how cloud and AI infrastructure affects EMS margins.
• VRT.US: best public comp for SpinCo’s power and thermal infrastructure ambition.
• NVT.US: useful comparator for liquid cooling and electrical-infrastructure demand linked to data centers.
• ETN.US: broader power-management benchmark that sets an upper bound for infrastructure-quality multiples.
• NXT.US: precedent spin-out from Flex that shows management has recent separation experience.

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