FLEX – Flex Ltd.
🔴 Live $110 (yf:FLEX) · 2026-09-05 · research written 2026-09-05 Consensus $142–$180 (mean $158, 12mo, Yahoo consensus (9 analysts)) · 2026-09-05 your watch-zone <$85–$95 (own-DD, on-pullback) — +15% away
Thesis
Verdict: WATCH at $109.51 at writing, 4 Sep. Not a full position here. Buy zone $85–95, or on a clean Investor Day (10 Nov 2026) that puts standalone numbers behind the 80% FY28 growth claim. Conviction in the business: medium-high. Conviction in the price: medium-low.
I want to argue that Flex is two stocks wearing one ticker, and that the market has already priced the interesting one at roughly fair value while giving you the boring one for free. That sounds like a bargain. It is not quite, because the interesting one has to deliver two consecutive years of 70% to 80% growth on hyperscaler programs, funded by the most capital-intensive year in the company's history, while a $4.4bn acquisition and a tax-free spin-off both close in the same six-month window. Any one of those is fine. All three at once, with buybacks suspended and free cash flow conversion cut to 40%, is a lot of execution for a 15x forward multiple that only looks cheap if FY28 lands.
The story in three moves. Flex spent seven years under Revathi Advaithi exiting consumer electronics, spinning off Nextracker, and buying its way into data center power: Anord Mardix (2021), Crown Technical (2024), JetCool (2024), Electrical Power Products (May 2026), and now EPC Power at $4.4bn (announced 3 Sep 2026, closing Q4 CY26). The result is a segment, Cloud and Power Infrastructure (CPI), that did $6.6bn in FY26 at a 9.2% margin, grew 38%, and is guided to grow 65% to 75% in FY27 and "80%-plus" in FY28 on a multi-year Google contract and other hyperscaler, neocloud and utility awards. On 5 May 2026 the board announced it would spin CPI out as its own company in Q1 CY27. The stock went from $96 to $135 the next day and peaked at $162 on 30 June, the day it entered the S&P 500. It has since given back a third of that, to $109.51, on a Q1 print that was strong on earnings and ugly on cash.
My read on the two pieces. SpinCo is the closest thing in public markets to a "Vertiv built by an EMS company": embedded power for the rack, switchgear and modular power out to the substation, liquid cooling from JetCool, and rack-to-pod compute integration for hyperscalers. Its margins (9.7% in Q1 FY27) are half of Vertiv's, because roughly half of it is still integration work at EMS economics. RemainCo is a $21bn, 5.7%-margin contract manufacturer for industrial, automotive, healthcare, networking and consumer customers, guided to low-to-mid single-digit growth, run by a 20-year Flex lifer. It will trade like Jabil on a bad day. Sum the two at multiples I can defend (a Celestica-to-Vertiv blend for SpinCo, a Jabil-to-Hon-Hai blend for RemainCo) and I get $115 to $167 per share on FY28 numbers, $128 central when discounted a year, against $109.51 today. That is a 17% expected return with a bear case near $67 if the FY28 acceleration slips to 40% and EMS multiples reassert. Not enough asymmetry to pay up for. Plenty to own a starter in and add on weakness.
What has to be true to upgrade to BUY: the Investor Day shows SpinCo standalone financials with Power (products) above 40% of CPI revenue and a segment margin path to 12%+; Q2 FY27 (28 Oct) shows CPI growth at or above the 45% to 55% guide with free cash flow turning positive again; and the EPC Power equity component is sized below $2bn. Any one of a pushed spin date, a CPI guide cut, or a customer concentration disclosure above 15% for a single customer drops this to PASS.
Three ways to say the one-line version, pick one:
- Good business, fair price, heavy execution calendar. Wait for $85–95 or the 10 November numbers.
- The market has valued SpinCo correctly and is giving you RemainCo for nothing, but nothing is what RemainCo may trade at post-spin.
- A 15x FY28 multiple that only exists if 80% growth on hyperscaler programs happens on time, twice.
Snapshot
Flex Ltd. (NASDAQ: FLEX), incorporated in Singapore, headquartered in Austin, Texas. Founded 1969 as Flextronics; renamed Flex in 2016. Sector page: ai-infrastructure. Fiscal year ends 31 March; "FY27" means the year to March 2027.
Price at writing $109.51 (4 Sep 2026), 369m shares, market cap $40.5bn, net debt about $3.1bn, enterprise value about $43.6bn before the EPC Power deal. 52-week range $53 to $167. Consensus: 9 to 11 analysts, all buy or strong buy, mean target $158. Short interest 2.8% of float. Beta 1.65. S&P 500 constituent since June 2026.
What it does in one sentence: Flex is the second-largest electronics manufacturing services (EMS) company outside Foxconn, building other brands' products in 100-plus factories across roughly 30 countries with 150,000 employees, and over the last five years it bought and built a data center power and cooling products business that is now large enough to spin out.
Investor materials: Q1 FY27 results (29 Jul 2026), Q4 FY26 results and spin-off presentation (5 to 6 May 2026), SpinCo transaction FAQ v1.1 (29 Jul 2026), EPC Power acquisition deck (4 Sep 2026 8-K exhibit 99.2). No investor day deck exists yet; the first one is 10 November 2026.
Business
Three segments as of Q4 FY26, one of which leaves
Flex re-segmented in May 2026 to make the spin legible. The old Agility and Reliability split is gone.
| Segment | What sits in it | FY26 revenue | FY26 growth | FY26 adj. margin | Q1 FY27 revenue | Q1 growth | Q1 margin |
|---|---|---|---|---|---|---|---|
| Cloud and Power Infrastructure (CPI) | Cloud & Cooling (rack/pod integration, JetCool liquid cooling) and Power (embedded power, critical power, switchgear, modular power, EP²) | $6.6bn | +38% | 9.2% | $2.20bn | +35% | 9.7% |
| Regulated Manufacturing Solutions (RMS) | Industrial (incl. energy-infrastructure contract manufacturing, robotics, warehouse automation), Automotive, Healthcare | $10.2bn | +5% | 6.0% | $2.67bn | +12% | 6.6% |
| Integrated Technology Solutions (ITS) | Communications (advanced networking, optical, NICs) and Lifestyle (consumer devices) | $11.1bn | −2% | 5.4% | $3.06bn | +20% | 5.2% |
| Total | $27.9bn | +8% | 6.3% | $7.93bn | +21% | 6.7% |
Source: Q4 FY26 call (6 May 2026), Q1 FY27 10-Q. Segment margins are adjusted operating margins; Q1 FY27 segment income excludes $23m amortization, $51m stock comp and $67m of "legal and other" (of which $53m is separation cost).
CPI is what leaves. Everything else is RemainCo, which keeps the Flex name.
SpinCo: what "grid to chip" actually means
Advaithi's pitch is that AI compute density has made power and thermal a single engineering problem rather than a bolt-on, that customers no longer want to buy subsystems from six vendors, and that Flex is the only company with embedded power, distributed power, thermal and compute integration under one roof. Strip the marketing and the portfolio is real. Working from the chip outward:
Embedded power is the power shelf and rack-level DC conversion that feeds the GPU trays. This is the business Flex already had before AI; it is now the leader in the 400 V DC transition and building 800 V DC racks with NVIDIA (the "modular platform" partnership referenced on the Q1 call). It also houses the Capacitive Energy Storage System built with Musashi Seimitsu hybrid supercapacitor cells, in production since H1 2025, which buffers the 48 V rail inside an 800 V-fed rack against millisecond GPU load swings.
Critical power is switchgear, busway, power distribution units and modular power rooms, mostly from Anord Mardix (Ireland and UK, bought December 2021) and Crown Technical Systems (Dallas, bought November 2024). Electrical Power Products (Des Moines, Iowa, closed May 2026, about $1.1bn cash, roughly $323m of FY26 revenue at mid-to-high-teens EBITDA margins) adds engineered control and relay panels and integrated control buildings for utilities and generators. This is the piece that reaches "all the way to the substation."
EPC Power (announced 3 Sep 2026, $4.4bn, closing Q4 CY26) is the newest and largest piece. Founded 2010, California, owned by Goldman Sachs Alternatives and Cleanhill Partners. It makes software-defined power conversion: MW-scale rectifiers, DC-DC converters and, in the roadmap, solid-state transformers, with 15+ GW deployed across 62 countries and US manufacturing planned above 30 GW a year in 2027. Guided to about $800m of CY26 revenue, 40% organic growth in CY27, and EBITDA margin expanding by double-digit points to about 30% in CY27. On those numbers Flex is paying 5.5x CY26 sales and roughly 13x CY27 EBITDA. That is a full price, defensible only if the 40% growth and the 30% margin both arrive. The strategic point is that EPC's grid-forming rectifiers are the box that lets an 800 V DC data center skip the UPS, transformer and AC distribution stages. If 800 V wins, EPC is core. If 800 V slips, Flex has paid $4.4bn for a good energy-storage inverter company.
Cooling is JetCool (cold plates, bought November 2024) plus coolant distribution units Flex developed itself and is still qualifying with customers. Management's own word for it is "nascent." Do not underwrite cooling revenue yet.
Cloud, the integration business, is rack and pod assembly and test for hyperscalers and AI customers, including Cerebras CS-3 systems built in the US. This is the lower-margin half. Management has said repeatedly that Power margins exceed Cloud margins and that Power is growing faster (70%-plus), which is why CPI mix is expected to drive margin up 100 bp in FY27 and a further 50 to 100 bp in FY28.
RemainCo: a good EMS company that will be valued as one
Post-spin Flex is ITS plus RMS, about $21bn of revenue at 5.7% blended adjusted margin, guided to low-to-mid single-digit growth. Its best businesses are advanced networking (switches, optical, NICs for large OEMs, riding data center pull-through and winning share), industrial energy infrastructure (contract manufacturing for generation, transmission and storage equipment, which stays in RemainCo even though the power products go to SpinCo), healthcare devices and drug delivery, and warehouse robotics. Its worst is Lifestyle, the consumer devices business Flex has spent years shrinking. Hartung's playbook is the one he and Advaithi ran since 2019: prune low-value revenue, drive productivity, buy back stock. It works. It is also worth 12x to 15x earnings, not 25x.
Customers and concentration
Top ten customers were 45% of FY26 sales. No customer exceeded 10% in FY26. In Q1 FY27 one customer reached 12% of sales, "the majority of the revenue with this customer" in CPI. Management named Google as the counterparty on the multi-year CPI contract announced in May. I am treating "the 12% customer is Google" as an inference. It has not been disclosed. Concentration is rising with the CPI ramp and will be far higher inside SpinCo once separated: if CPI is $11bn in FY27 and one customer is roughly $4bn of consolidated sales, that customer could be a third of SpinCo. Watch the Form 10 for this number.
Other named relationships: Cerebras (CS-3 manufacturing, cooling, next-gen power), NVIDIA (modular 800 V platform), Musashi Seimitsu (supercapacitor cells for the CESS). Awards were described as spanning "multiple hyperscalers, neocloud providers, colocation operators and utility customers." Management says 90%-plus of CPI business for the next three quarters is booked.
Footprint
About 100 facilities in roughly 30 countries; 149,686 employees at March 2026 (61,841 Americas, 60,870 Asia, 26,975 Europe). Sales by region FY26: North America 44%, Europe 20%, China 16%, other 20%. Data center capacity is concentrated in the US (Texas, Iowa, plus JetCool in Massachusetts and EPC in California) and Mexico. FY27 capex of $1.5 to $1.6bn is almost entirely CPI facilities, cooling and manufacturing infrastructure; management expects capex to fall back to 2.5% to 3% of CPI revenue and under 2% of ITS/RMS revenue in FY28.
Value chain position
Flex sits between component makers and the hyperscaler. Upstream it buys GPUs and CPUs (customer-consigned, mostly), memory, power semiconductors, magnetics, copper, switchgear components and capacitors. Downstream it delivers tested racks, power rooms and modular buildings to hyperscaler, colocation, neocloud and utility sites. The layer is fragmented at the integration end (Foxconn, Quanta, Wiwynn, Celestica, Jabil, Sanmina) and concentrated at the power products end (Vertiv, Schneider, Eaton, ABB, Delta). Flex's argument is that it is the only one straddling both. Switching costs are high for power products once designed in; they are moderate for integration, where hyperscalers deliberately dual-source.
Upstream, the two nodes where Flex is a price-taker are power semiconductors (SiC and GaN for the 800 V conversion stages) and passives (film capacitors, supercapacitors, magnetics). Neither is a Flex-specific bottleneck; both are covered in ai-infrastructure and the passives work in the vault. The one small-cap upstream name already researched is Musashi Seimitsu, and that page concludes the supercapacitor exposure is pre-scale. No secondary long emerges from Flex's supply chain that isn't already in the vault.
Industry landscape
Sector context lives in ai-infrastructure (the AI compute and power stack) and datacenter-foundations (the power chain, PUE, cost per MW). No standalone EMS primer exists in the vault; the short version needed for this page follows. Flagged in the industry log as a primer backlog item.
EMS is a 5% to 10% gross margin business because the customer owns the design, the IP and often the components, and the manufacturer competes on footprint, working capital and yield. Foxconn does about $300bn of it at a 6% gross margin. Nobody re-rates an EMS company for growing; they re-rate it for changing what it sells. Flex's data center bet is exactly that: use the integration relationship to sell proprietary power and thermal products at 15% to 25% gross margins alongside the 8% integration work. Celestica did the same with its own networking switches and now trades at 16x forward earnings on 60% growth. Vertiv, the pure products comparison, runs a 38% gross margin and a 20% operating margin and trades at 31x forward earnings.
Why now is easy to state. AI rack power went from 10 kW to 120 kW to 600 kW in three product generations. That forced liquid cooling, then 400 V DC, and by 2027 to 2028 800 V DC with solid-state transformers, which changes the electrical bill of materials from the substation inward. Hyperscaler capex is running at roughly $430bn trailing four quarters and is guided higher for 2026. The US data center pipeline is constrained by power rather than chips, and every quarter of Flex's calls now leads with grid capacity rather than GPU supply. The supply side of Flex's business (power conversion, switchgear, transformers) has one to two year lead times and is short.
What consensus is doing: pricing the CPI ramp as delivered (FY28 EPS $7.09, up 50% on FY27) and pricing RemainCo as a rounding error. The narrative-versus-reality gap I would watch is the second-derivative of hyperscaler capex in CY27. Every EMS name, Flex included, de-rated 40% to 60% in the 2022 to 2023 digestion. FY28 "80%-plus" growth is on booked programs, which helps, but booked programs get pushed.
SemiAnalysis cross-check: the mirror's only substantive mention is the 10 April 2025 tariff piece ("Flex and Jabil are also starting to ramp in the rack assembly supply chain" for Amazon Trainium via Mexico). No SA piece on Flex as a power company, and SA writes constantly about 800 V DC and data center power. That silence is worth noting: SA's power coverage centers on NVIDIA's ecosystem partners, Vertiv, Delta and the Asian PSU makers, and Flex has not featured. Either Flex is not yet a technology leader in SA's eyes, or it is too much of an EMS name for their frame. No contradiction with this page, since the vault has no prior Flex view.
Beyond the Hype (Sanmina piece, 3 Nov 2025) states the bear frame directly: contract manufacturers are "low value add and low margin," and BTH stayed out of Celestica for that reason despite the AI boost. My thesis differs only in that Flex is trying to escape those economics by owning power products. BTH's discipline is the right default until SpinCo's product mix is disclosed.
Financials
Core four
Organic growth turned in FY26 and is accelerating. Revenue went $28.5bn (FY23), $26.4bn (FY24), $25.8bn (FY25), $27.9bn (FY26), so three years of consumer exit and de-emphasis before CPI got big enough to pull the total. FY27 guide is $33.7 to $35.2bn, up 23% at the midpoint, of which about half a point is EP². CPI drives all of it.
Margins are expanding on mix. Adjusted operating margin went 4.9% (FY24) to 5.6% (FY25) to 6.3% (FY26) and is guided to 7.0% to 7.2% in FY27. Adjusted gross margin hit a record 9.9% in Q4 FY26 and 9.6% in Q1 FY27. GAAP lags adjusted by 130 to 180 bp because of stock comp (~$140m a year), amortization (~$90m) and now separation costs ($53m in Q1 alone).
Capital intensity is the problem this year. Capex was 2.2% of revenue in FY26 ($625m) and is guided to $1.5 to $1.6bn in FY27, about 4.5% of revenue, all for CPI. Inventory rose 24% year over year to $6.45bn at June 2026 (56 days net of customer advances, one day worse than a year ago; customer-controlled inventory $1.5bn). Q1 FY27 free cash flow was $41m against $266m a year earlier.
Capital deployment: $944m of buybacks in FY26 (19m shares, average about $50), $1.26bn in FY25, $1.30bn in FY24. Share count fell from 450m (FY23) to 366m (March 2026). Buybacks stopped in Q1 FY27 ($0; $1.1bn authorization remains). FY26 M&A was small ($44m); FY27 has $1.1bn of EP² closed and $4.4bn of EPC pending, funded by a $1.45bn term loan (November 2027 maturity) and a debt-plus-equity package for EPC that is not yet sized.
Second derivative: accelerating, and guided to keep accelerating
| Quarter | Revenue | YoY growth | Change in growth rate |
|---|---|---|---|
| Jun-25 (Q1 FY26) | $6.58bn | +4.2% | |
| Sep-25 (Q2 FY26) | $6.80bn | +3.9% | −0.3 pt |
| Dec-25 (Q3 FY26) | $7.06bn | +7.7% | +3.8 pt |
| Mar-26 (Q4 FY26) | $7.48bn | +16.9% | +9.2 pt |
| Jun-26 (Q1 FY27) | $7.93bn | +20.6% | +3.7 pt |
| Sep-26 guide (Q2 FY27) | $7.95–8.25bn | +19% mid | −1.6 pt |
| H2 FY27 implied | ~$18.4bn | +27% | +8 pt |
Prior-year quarterly revenue from FY25 reported figures. The H2 FY27 line is the full-year guide midpoint less Q1 actual and Q2 guide midpoint, against $14.5bn in H2 FY26. CPI alone is guided from +35% (Q1) to +45–55% (Q2) to roughly +100% in the second half. The exit rate into FY28 is what the "80%-plus" number depends on.
Incremental margins say the operating leverage is real
Q1 FY27 added $1.35bn of revenue year over year and $139m of adjusted operating income, a 10.3% incremental margin against a 6.7% reported margin. Q4 FY26 added $1.08bn and $103m, 9.5% incremental. Full-year FY26 added $2.1bn and about $314m, 15% incremental, flattered by consumer exits. Incremental gross margin in Q1 FY27 was about 12% against 9.6% reported. New revenue is higher quality than the base, which is the whole thesis in one number, with the caveat that Q1 CPI margin was 20 bp lower sequentially on program ramp costs.
Balance sheet
Gross debt $5.24bn at June 2026: senior notes of $399m (6.0% 2028), $654m (4.875% 2029), $670m (4.875% 2030), $651m (5.25% 2032), $598m (5.375% 2035), the $1.45bn term loan, a $500m delayed-draw term loan (December 2027), and HUF bonds of $318m. Cash about $2.4bn at March 2026. Net debt to FY27 adjusted EBITDA (roughly $3.0bn) is around 1x today; after EPC it depends on the equity slice. SpinCo is promised "investment-grade credit metrics," which means Flex will take most of the debt or SpinCo will raise equity. The FAQ says Flex may retain up to 19.9% of SpinCo at the spin.
ROIC on the $8.9bn of invested capital was about 13% in FY26 on adjusted NOPAT (roughly $1.4bn), against a WACC I would put near 9% to 10% given the 1.65 beta. Value-creating, and the FY27 capex will dilute it before FY28 restores it.
Valuation
At $109.51: 23x FY27 adjusted EPS guide midpoint ($4.58) and consensus ($4.73); 15.5x FY28 consensus ($7.09, range $5.91 to $7.96); 42x trailing GAAP; EV/trailing EBITDA 20x; 7.4x book; free cash flow yield under 2% on the FY27 guide (about 40% conversion of $1.7bn adjusted net income, or roughly $0.7bn).
The FY28 consensus needs to be understood before the 15x means anything. It requires revenue of about $45bn (+30%) and EPS growth of 50%. On management's framework that is CPI at about $20bn (FY26 $6.6bn, times 1.70, times 1.80), EPC adding about $1bn, and RemainCo at $24bn growing 4%. It is an internally consistent number if the guide holds. It is not a number to put a low-teens multiple on and call cheap, because the guide has to hold twice.
| Peer | Market cap | Revenue growth (TTM) | Operating margin | EV/EBITDA | Forward P/E |
|---|---|---|---|---|---|
| Flex (FLEX) | $40.5bn | +21% | 5.0% (GAAP) | 20.2x | 15.5x |
| Celestica (CLS) | $39.4bn | +62% | 9.8% | 23.1x | 16.0x |
| Jabil (JBL) | $32.5bn | +12% | 5.2% | 14.5x | 18.4x |
| Sanmina (SANM) | $10.6bn | +70% | 7.0% | 13.0x | 14.3x |
| Hon Hai (2317.TW) | ~$115bn | +41% | 3.8% | 8.6x | 11.1x |
| Vertiv (VRT) | $108bn | +24% | 20.4% | 38.7x | 30.8x |
Yahoo Finance, 4 Sep 2026. Hon Hai market cap converted at a rough NT$31/US$. Forward P/E is on the next fiscal year in each case, which for Flex is FY28; on FY27 Flex is 23x, above every EMS peer.
Management
Who runs what after the spin
| Name | Post-spin role | Background |
|---|---|---|
| Revathi Advaithi | CEO, SpinCo; transitional non-executive chair of Flex | Flex CEO since Feb 2019. Prior: Eaton (COO Electrical Sector), Honeywell. Ran the Nextracker spin. Board of Uber. |
| Michael Hartung | CEO, Flex | 20 years at Flex via the 2007 Solectron acquisition; President Agility Solutions 2020–24; President and Chief Commercial Officer since Jul 2024. UCLA economics. |
| Kevin Krumm | CFO, SpinCo (Flex CFO until then) | Joined Jan 2025 from APi Group (EVP/CFO); 15 years at Ecolab (Treasurer). Flex is searching for a permanent CFO. |
| Hooi Tan | COO, SpinCo | Long-tenured Flex operations executive; COO of Flex today. |
| Bill Watkins | Non-executive chair, SpinCo | Flex board since 2009, chair; ex-CEO Seagate and Imergy. |
| Rob Campbell, Mattias Jansson, Todd Hoover, Chris Butler | SpinCo CCO, President Embedded Power, President Critical Power, CTO/Strategy | Internal Flex executives. |
| Dennis Kirkpatrick, Mike Thoeny, Rodrigo DallOglio, Ivan Brockman | Flex Presidents ITS and RMS, COO, Chief Business Transformation Officer | Internal Flex executives. |
Both companies will be headquartered in Austin. My read: the A-team goes with SpinCo. Advaithi, Krumm, Tan and Watkins all move. RemainCo gets Hartung, who has run the Agility half of Flex credibly for four years, and an open CFO seat. That is the right allocation for shareholder value if you assume you will hold SpinCo and sell RemainCo. It is also a signal about which company management believes in.
Skin in the game: modest, and they sold at the top
All executive officers and directors together own 2.29m shares, 0.62% of the company, about $250m at $109.51. Advaithi holds 1.30m shares (roughly $143m) plus unvested PSUs that could pay up to 1.36m more shares. Hartung holds 126k shares ($14m). Krumm holds 34k ($3.7m).
The June 2026 Form 4 cluster deserves plain description. On 15 to 18 June, three to five weeks after the spin announcement and near the all-time high, Advaithi sold 237,067 shares at $141.74 to $149.19 ($34.2m), Tan sold 46,178 ($6.7m), Offer sold 53,442 ($7.8m), Hartung sold 29,896 ($4.3m) and Wendler sold 6,780 ($1.0m). The same week the annual PSU and RSU grants landed (Advaithi 255,062 units). Some of the selling covers tax on vesting; the CEO's six-month total of roughly 474k shares sold for about $67m (per Quiver, from Form 4s) is more than tax-cover. Hartung sold a further 2,755 shares on 18 August at $120 to $123. There were no open-market purchases by any insider in the period. My judgment: management believes in SpinCo strongly enough to move there, and diversified about a third of the CEO's directly held position at $145 on the way. Both are true.
The supplemental award is a SpinCo option, granted before the spin was announced
Advaithi's FY26 total compensation was $44.4m, up 63%, of which $25.1m is a one-time "Supplemental Equity Award" of performance share units granted 19 June 2025. Excluding it, her pay was $19.3m (salary $1.325m, stock $13.3m, cash bonus $3.95m at 171% to 181% of target, other $0.6m including $119k of personal aircraft use). The pay ratio is 3,431:1 with the award and 1,492:1 without, against a median employee at $12,939.
The award vests on "the Company's CPI business measured at the end of fiscal year 2028," capped by relative TSR over three years from 19 June 2025, and pays up to 1,357,810 shares (about $149m at $109.51). In other words, eleven months before the spin was announced, the board wrote the CEO a contract whose payout is the value of the segment she now runs as a separate company. That is alignment for a future SpinCo holder. It is also an incentive to maximize the size and story of CPI by March 2028, which is exactly what a $4.4bn acquisition folded into CPI three months before the spin does. I do not think the EPC deal is bad; I do think the incentive should be named.
Annual bonus metrics are adjusted operating profit (40%), adjusted free cash flow (35%) and revenue (25%) with a ±10 pt individual modifier. Long-term equity is one-third relative-TSR PSUs, one-third adjusted EPS growth PSUs, one-third RSUs. FY26 PSUs paid at the 200% maximum on both metrics (rTSR above the 75th percentile; adjusted EPS growth averaging 18.9%). No employment agreements, no single-trigger change-of-control vesting, no excise gross-ups, no hedging or pledging. Stock comp is about 0.5% of revenue, low for a company of this market cap.
Board and governance
Nine directors, seven independent. Chair Bill Watkins has served since 2009 (17 years); Lay Koon Tan since 2012. Hurlston (Lumentum CEO) is classed non-independent. Committees: Audit (Harris, Stevens, Tan, Ward), Compensation and People (McSweeney, Stevens, Sylvester, Ward), Nominating and Governance (Harris, Sylvester chair, Tan, Ward). Relevant operating experience is decent: Stevens (ex-GM CFO), Ward (ex-Cummins CFO), Sylvester (ex-ABB), Harris (ex-Raytheon). Average tenure eight years. Auditor Deloitte.
Related-party transactions: none above the $120,000 threshold in FY26. Directors sit on boards of Astera Labs, Cisco, Masco and Nextpower, all arm's-length. No shell entities, no cross-holdings, no litigation of note surfaced. The Singapore-law items that matter: quorum 33.3%, annual shareholder authorization to issue up to 20% of shares and to repurchase up to 20%, and the spin needs High Court of Singapore sanction as well as a shareholder vote and an effective Form 10.
Post-spin, Advaithi becomes non-executive chair of a company she no longer runs, for a "transitional period," and the board has said it will add a lead independent director. That is the one soft governance point in an otherwise clean structure: the former CEO chairing RemainCo while running its biggest customer-adjacent sibling. Watch the transition period length in the Form 10.
Capital allocation record: A-minus
Nextracker is the reference case. Flex IPO'd 24% of it in February 2023 at $24, distributed the rest in January 2024, and Nextpower (as it is now called) is a $13bn company. Buybacks were heavy and well-timed, $3.5bn over FY24 to FY26 at average prices in the $25 to $50 range against $109 today. The data center acquisitions have so far bought growth at reasonable prices (Anord Mardix ~$540m for a business now central to a $6.6bn segment). The minus is for FY27: buybacks suspended, capex tripled, two acquisitions totaling $5.5bn in one year, one of them at 5.5x sales with equity in the mix, all while running a separation. Individually defensible. In aggregate, the most leveraged operating year in the company's recent history.
Management grade overall: Green on governance and related parties, Green on capital allocation history, Yellow on skin in the game (low absolute ownership, heavy June selling), Yellow on compensation (the supplemental award's structure). No Reds.
Catalysts & risks
Calendar
- 28 Oct 2026: Q2 FY27 results. CPI guide is +45–55%; the second-half ramp needs to be visible in the Q3 guide.
- Q4 CY26: EPC Power close, with the debt/equity mix disclosed. An equity raise above ~$2bn (18m+ shares) is a negative.
- 10 Nov 2026: Investor Day. First standalone SpinCo and RemainCo financials, long-term targets, capital structure. This is the single most important date.
- Q4 CY26 or early 2027: Form 10 filing (customer concentration, standalone margins, debt allocation, distribution ratio, Flex's retained stake up to 19.9%).
- Q1 CY27: shareholder vote, Singapore High Court sanction, distribution. Index treatment of the two pieces (SpinCo likely S&P 500 eligible; RemainCo may be cut from growth funds) creates forced flows either side.
- FY28 (Apr 2027 to Mar 2028): the "80%-plus" year. Capex normalizes. CPI margin +50–100 bp on mix.
Tailwinds
800 V DC and solid-state transformer adoption (3 to 5 year), which raises electrical content per MW and is where EPC and embedded power sit. Liquid cooling attach (1 to 3 year), from a near-zero base for Flex. US reshoring of electrical manufacturing (Iowa, Texas, Mexico under USMCA). Grid capacity as the gate on data center delivery, which pulls utility customers into CPI. Advanced networking pull-through for RemainCo.
Risks, ranked
| Risk | Likelihood | Mitigants | Can it close? |
|---|---|---|---|
| FY28 CPI growth lands well below 80% as hyperscalers digest | Medium. Every capex cycle has a digestion year; FY28 is the third year of this one. | 90%+ of the next three quarters booked; multi-year Google contract; customer diversification across colos, neoclouds, utilities. | Only by delivery. Q2 and Q3 FY27 prints partially de-risk it. |
| Cash conversion stays weak: capex $1.5–1.6bn, inventory +24%, FCF conversion 40% | High for FY27 (it is guided). Medium for FY28. | Management says capex is "unique to FY27"; customer advances cover part of inventory. | Yes, if FY28 capex falls to ~3% of CPI revenue as guided. |
| EPC Power: price, equity dilution, integration during a spin | Medium. $4.4bn is 11% of market cap; 13x CY27 EBITDA assumes 40% growth and margin expansion of double-digit points in one year. | Committed Citi/BofA financing; EPC goes into SpinCo where it fits. | Partially, once financing mix is disclosed and CY27 numbers print. |
| SpinCo margins stay at EMS levels because Cloud integration is half the mix | Medium. Q1 CPI margin was 9.7%; Vertiv is 20%. | Power growing faster than Cloud; +100 bp guided FY27, +50–100 bp FY28. | Yes, visible in the Form 10 product/integration split. |
| Spin slips or loses tax-free status | Low-medium. Nextracker went to plan; Singapore court and IRS opinion are routine but not automatic. | Separation office in place; leadership named; $53m/quarter already being spent. | Yes, at distribution. |
| Single-customer concentration inside SpinCo (12% of Flex = possibly 30%+ of SpinCo) | High that it exists; medium that it bites. | Multi-year contract; hyperscalers dual-source but rarely walk mid-program. | No. Structural to the business. |
| Insider incentive to inflate CPI by March 2028 (supplemental award) | Low as misconduct, medium as bias toward growth-at-any-price M&A. | Award capped by rTSR; board independence; clawback policy. | No. Lives until FY28 measurement. |
| RemainCo orphaned post-spin: growth holders sell, no natural buyer, open CFO seat | Medium. | Buybacks resume; 12x earnings floor; Jabil precedent. | Yes, over 6 to 12 months of post-spin trading. |
| Tariffs / Mexico / China (16% of sales) | Medium, ongoing. | USMCA-compliant Mexico assembly; US expansion in Iowa, Texas. | No. Managed, not closed. |
Dilution
Share count fell 19% over three years; that reverses in FY27 with the EPC equity component and zero buybacks. The 20% annual issuance authorization under Singapore law is standard but means the board can issue up to 74m shares without another vote. No converts, no warrants.
Bear case
Hyperscaler capex growth slows to 10% to 15% in CY27; the Google program ramps but the second and third customers push out; CPI grows 40% in FY28 rather than 80%; margins stay at 9.5% because Cloud stays half the mix; EPC's 30% margin proves to be a 20% margin. FY28 EPS comes in near $5, the spin still happens, SpinCo trades at 15x and RemainCo at 11x, and the combined value is about $67 a share. Down 39%. This is an ordinary scenario. It is what an EMS multiple does when growth breaks, and it happened to this stock in 2022 to 2023 when it lost 45% peak to trough.
Bull case
FY28 delivers 80%-plus on CPI at 12% margins, EPC hits its numbers, 800 V transitions on schedule, SpinCo lists into an S&P 500 slot and trades at 28x as a "Vertiv-lite" with faster growth, RemainCo re-rates to Jabil's 18x on advanced networking. Combined value about $190 to $210 a share on FY28. Up 75% to 90%.
Valuation / DCF
Sum-of-the-parts on FY28 (year to March 2028), which is the first year both companies are standalone and the year the guide targets.
| Revenue FY28 | Adj. EBIT | NOPAT (21%) | Multiple range | Value range | |
|---|---|---|---|---|---|
| SpinCo (CPI + EP² + EPC) | ~$21bn | ~$2.5bn | ~$2.0bn | 18x–26x | $36–52bn |
| RemainCo (ITS + RMS) | ~$24bn | ~$1.5bn | ~$1.2bn | 12x–15x | $14–17.5bn |
| Enterprise value | $50–70bn | ||||
| Less pro forma net debt (incl. ~half of EPC in debt, less FY27 FCF) | ~$4.6bn | ||||
| Equity value | $45–65bn | ||||
| Per share (~390m post EPC equity) | $115–167 | ||||
| Discounted 12 months at 10% | $105–152, central $128 |
SpinCo assumptions: CPI $6.6bn × 1.70 × 1.80 = $20.2bn plus about $1bn of EPC in the fiscal year; margin 11% (9.2% + 100 bp + 75 bp) plus EPC's higher margin. RemainCo: $21.3bn growing 8% then 4%, margin 6.2%. The SpinCo multiple range brackets Celestica (16x) and a discount to Vertiv (31x). Consensus mean target $158 sits at the top of my undiscounted range; the Seeking Alpha sum-of-the-parts piece that circulated in June argued the opposite (40x for SpinCo, 12x for RemainCo, and still below the then-price of ~$150), which tells you the same math produced a "sell" at $150 and a "hold" at $109.
Scenario weights: bear 25% ($67), base 50% ($128), bull 25% ($190). Probability-weighted 12-month value about $128, or +17%. Positive skew is there but thin for the execution load. At $85 to $95 the same math gives +35% to +50% base and a bear case of −20% to −30%, which is where I would size a real position.
No DCF run; the FY27 capex step and the spin make a single-entity DCF less useful than the sum-of-the-parts above. /dcf can model SpinCo standalone once the Form 10 gives standalone cash flows.
Decision log
2026-09-05 (deep-dive, new coverage). WATCH at $109.51. Not held. Buy zone $85–95 (own-DD, on-pullback); alternatively upgrade to BUY at market if the 10 Nov Investor Day shows Power above 40% of CPI revenue with a 12%+ margin path and EPC equity below $2bn. Trim/target $150–160 pre-spin (own-DD, by Q1 CY27), which is roughly the consensus mean and the top of my undiscounted base range. Stop for a starter position $85 (own-DD) only if accompanied by a CPI guide cut; a price move alone to $85 is the buy zone. Downgrade to PASS on any of: spin date slips past Q2 CY27, CPI FY27 guide cut below 60%, single customer above 15% of consolidated sales, EPC financing above $2.5bn equity. Reasoning: the business transformation is real and management has done this before with Nextracker; the price already reflects the FY28 acceleration at 15.5x, leaving a 17% probability-weighted return against a bear case near −40%. Insider selling at $145 in June and the CPI-linked supplemental award are noted and do not disqualify the name.
Sources
Primary: Flex Q1 FY27 press release and 10-Q (29 and 31 Jul 2026); Q1 FY27 earnings call transcript (FactSet, 29 Jul 2026); Q4 FY26 earnings call transcript (FactSet, 6 May 2026); FY26 10-K (20 May 2026); DEF 14A proxy (24 Jun 2026); SpinCo transaction FAQ v1.1 (29 Jul 2026, flex.com/downloads/flex-spinco-transaction-faqs); leadership teams press release (29 Jul 2026); EPC Power acquisition press release and investor deck (8-K, 4 Sep 2026); EP² acquisition releases (4 May and 22 May 2026); EDGAR Form 4 filings June to August 2026; Yahoo Finance quotes, estimates, holders and price history (4 Sep 2026).
Secondary: Seeking Alpha "Flex: Sum Of The Parts Is Lower Than The Current Share Price" (June 2026, not fetched, summary via search); TheValueist Q1 FY27 read-through thread on X (Jul 2026); Beyond The Hype "Sanmina had a big runup" (3 Nov 2025, vault raw); SemiAnalysis "Tariff Armageddon?" (10 Apr 2025, vault mirror); Investing.com and Quiver Quantitative earnings and insider coverage; Digitimes EMS coverage (Jan 2026).
Gaps: no Fabricated Knowledge, Vik's, Irrational Analysis, Collyer Bridge or Mike10947310 coverage of Flex found. Twitter sweep returned one substantive thread. 10-K segment tables and long-lived assets by country not extracted (segment data taken from the call and 10-Q instead). The EPC Power financing split, the SpinCo distribution ratio and the Form 10 do not exist yet.
Related vault pages: ai-infrastructure, datacenter-foundations, Forgent Power Solutions (datacenter electrical distribution, a smaller US competitor to the Crown/EP² businesses), Musashi Seimitsu (supercapacitor cells inside Flex's CESS).
Appears in / comparisons: Musashi Seimitsu.