2026-09-11 · EN
FLEX — Flex Ltd.
SpeculativeOpen this week — full figures
Deep-value analysis — FLEX (Flex Ltd., NASDAQ)
Analysis date: 09/11/2026 · Reference price: $108.01 · Market cap: $39.90 bil
Fiscal year ending March 31 · Reporting and pricing currency: USD (no conversion risk)
Primary sources: FY2026 10-K (filed 05/20/2026), Q1 FY2027 10-Q (filed 07/31/2026), FY2022-FY2025 10-K,
data pack rapoarte/deep/data-pack-FLEX-20260911.md, simulation rapoarte/deep/mc-FLEX-20260911.json.
Executive summary
Flex is no longer the company it was three years ago, and the market understood that before I did. In two years, the cost-plus assembler with a 3.2% operating margin (FY2024) has become a group with a 4.9% GAAP operating margin and a segment — Cloud and Power Infrastructure — growing 35% per quarter, carrying a 9.7% margin, and set to represent roughly a third of revenue in FY2027. Management announced on 05/05/2026 the separation of CPI into an independently listed company (SpinCo), with a Form 10 expected in September 2026 and completion in calendar Q1 2027. FY2027 guidance was raised after Q1 to $33.7-35.2 bil revenue (+23% at the midpoint, versus $27.9 bil in FY2026) and $4.42-4.74 adjusted EPS (+40%). The business is in its best shape in a decade.
The problem isn’t the business, it’s the price. At $108.01 the stock trades at 41.7x trailing-four-quarter GAAP net profit ($2.59/diluted share) and 23.6x guided FY2027 adjusted EPS. Five fiscal years ago, at 03/31/2022, the same company traded at 7.2x profit. Of the 711% price increase over the last five years ($13.32 → $108.01), earnings per share contributed +40% ($1.85 → $2.59); nearly all the rest is multiple re-rating. The median P/E across the five fiscal years is 12.5x; today we pay 41.7x.
Free-cash-flow yield is the sharp end of the argument. Trailing-four-quarter FCFE (through 06/26/2026) is $833 mil — $1,562 mil CFO minus $736 mil capex plus $7 mil disposals — i.e. 2.09% of market cap. It isn’t an accidentally depressed figure: the company’s own FY2027 guidance is FCF conversion of ~40% of adjusted net profit (revised down from 60%, for separation costs), which on adjusted profit of ~$1,713 mil gives ~$685 mil FCF in FY2027, i.e. a 1.72% yield — below today’s. Capex rises from $633 mil (FY2026) to a guided $1.5-1.6 bil (FY2027), 2.4x, and in Q1 FY2027 free cash flow was $41 mil versus $268 mil a year earlier, an 85% drop in a quarter when revenue grew 21%.
I triangulated five models. The FCFE DCF with a normalized base ($900 mil, i.e. TTM FCFE plus one-off separation costs) and assumptions g1=9%, r=10.5%, gt=2.5% gives $42.87/share, with a margin of safety of −60.3%. The bear model, starting directly from FY2027-guided FCF ($685 mil), gives $25.18 (−76.7%). Earnings Power Value à la Greenwald, at the mid-cycle average operating margin of 4.4% and a 10% WACC, gives $20.35 (−81.2%). Historical multiples (median P/E 12.5x over five fiscal years, applied to an estimated FY2027 GAAP EPS of $3.32) give $41.50 (−61.6%). The only model that reaches the current price is the bull case, and only because I build it generously: base $1,650 mil (estimated FY2028 free cash flow of $2.0 bil, discounted back two years), g1=12% over ten years, r=9.5%, gt=3% — and it still only reaches $112.57, a +4.2% margin. In other words, today’s price is the bull scenario, executed flawlessly.
The Monte Carlo simulation (20,000 scenarios, same central assumptions as Model 2) confirms this: median intrinsic value $43.15, median margin of safety −60.0%, probability the stock is undervalued 0.8%, probability of a margin above 30% 0.1%. The P10-P90 range of the margin is −75.3% … −34.9%: even the optimistic tail of the distribution stays deeply negative. For intrinsic value to reach the current price with a base of $900 mil and r=10.5%, sustained growth of 25.8% per year for five years and 12.9% for the following five would be needed — above the peak-year revenue growth guidance, sustained for a decade.
Verdict: AVOID at $108. This isn’t a short thesis — earnings quality is decent, the balance sheet is solid (net debt $2.4 bil, 1.17x EBITDA), and capital allocation has historically been excellent. It’s a price thesis: there is no margin of safety, and none appears except below ~$55-60 for a buyer requiring a 30% MOS, or below ~$70 for a buyer who accepts the bull thesis with its full assumptions. The level to watch is $70 — the price at 03/31/2026, before the rally that took the stock to $162 in June and from which it has already corrected 33%.
Why the analysis was rerun. The mechanical triage deep_delta.py flagged FULL because a falsifier on
fcf_yield tripped. The check is mandatory and the answer is unflattering for the procedure: the falsifier
in the 08/19/2026 thesis was set at a threshold of 0.0186 — exactly the value of the metric at the time it
was written. The price fell from $120.24 to $108.01 (−10.2%), the yield mechanically rose to 2.09% (TTM) or
2.64% (on the FY2026 basis used by the indicator script), and the threshold gave way. It didn’t break
because the business generated more cash — on the contrary, quarterly FCF fell 85%. A threshold set at the
metric’s current value falsifies nothing, it merely detects price movement. The correct threshold for this
thesis is 0.045, a level that would actually change the verdict if reached.
Business and moat
Flex Ltd. has, since 1994, been one of the three or four major global electronics manufacturing services (EMS) providers, with operations in roughly 30 countries. The basic model is the classic one for the industry: the customer designs or co-designs the product, Flex buys the components, assembles, tests and ships them, and the gross margin is single-digit. In FY2026 revenue was $27,914 mil, gross margin 9.2%, GAAP operating margin 4.90% ($1,368 mil), net profit $880 mil. Return on invested capital (NOPAT $1,053 mil over invested capital of $6,506 mil) was 16.2% in FY2026, versus 14.9% in FY2025 — above the cost of capital, but achieved through turnover, not margin: 1.38 asset turns per year over a 3.15% net margin.
The segment structure was reorganized in Q4 FY2026, precisely to isolate the business that matters. The three current segments, with revenue (mil USD) and margins over three fiscal years (FY2026 10-K, MD&A):
| Segment | FY2024 revenue | FY2025 revenue | FY2026 revenue | FY2024 margin | FY2025 margin | FY2026 margin |
|---|---|---|---|---|---|---|
| Integrated Technology Solutions (ITS) | 12,636 | 11,336 | 11,109 | 3.9% | 4.8% | 5.4% |
| Regulated Manufacturing Solutions (RMS) | 10,535 | 9,678 | 10,191 | 5.1% | 5.2% | 6.0% |
| Cloud and Power Infrastructure (CPI) | 3,244 | 4,799 | 6,614 | 9.4% | 10.2% | 9.2% |
In Q1 FY2027 (quarter ended 06/26/2026) the picture is: ITS $3,056 mil (+20%, margin 5.2%), RMS $2,670 mil (+12%, margin 6.6%), CPI $2,202 mil (+35%, margin 9.7%), total $7,928 mil (+20.6%). All three segments improved margin year over year, which is rare and worth noting: it isn’t just a mix story, it’s execution too. RMS gained 130 basis points from favorable mix in Industrial and Automotive, ITS 10 points despite pressure from memory pricing.
The moat differs radically across the three pieces, and that’s the essence of the investment case.
ITS (38% of revenue) has no moat. Communications, networking, satellites, premium consumer products. It’s pure contract manufacturing, with clients able to move programs to Jabil, Celestica or Foxconn at contract renewal. The proof is in the numbers: ITS revenue fell from $12,636 mil (FY2024) to $11,109 mil (FY2026), −12% in two years, and in FY2025 the Communications business lost 19% in a single year. The 5.4% margin is the result of cost cuts and mix, not pricing power. Here Flex is a price taker.
RMS (34%) has a modest but real moat. Industrial, automotive, healthcare — regulated markets, where moving a program requires revalidation, FDA or IATF recertification, and rewriting the compliance file. Switching cost is measured in years, not quarters. The 6.0-6.6% margin and relative revenue stability (10,535 → 9,678 → 10,191 mil) confirm this. This is the kind of business that justifies a 10-13x earnings multiple, no more.
CPI (28%, heading to ~33%) is something else — and the only part that justifies today’s price. It isn’t contract assembly, it’s a proprietary product portfolio: utility-scale and rack-level power systems, liquid cooling for high-density racks, integrated compute systems. Flex sells equipment, not labor, and that shows in the margin (9.2-9.7% versus 5.2% at ITS). Growth is guided at 65-75% in FY2027 and over 80% in FY2028, with a substantial part already in booked programs. The acquisition of Electrical Power Products Inc. (EPP), closed 05/01/2026 for $1,187 mil cash, adds critical utility-grade electrical equipment. The moat here is of a different nature: scarce installed capacity in a market with explosive demand, long-standing relationships with hyperscalers, and a timing advantage — whoever has the factory ready now wins the contract. It’s a real moat, but cyclical and unproven over a full cycle: nobody knows what the 9.7% margin looks like once the capacity rush ends, and Flex itself lost 100 basis points of margin at CPI in FY2026 due to “ramp costs and unfavorable mix in Cloud and Cooling.”
Customer concentration is the structural risk running through all three. The top ten customers represented ~49% of sales in Q1 FY2027 (48% a year earlier), and a single customer made up 12% of total sales, mostly in the CPI segment. In a company with $32 bil of revenue, a 12% customer means nearly $4 bil: losing or renegotiating it would wipe out more than CPI’s entire organic growth for FY2027. Geographic dispersion ($13.8 bil Americas, $8.4 bil Asia, $5.7 bil Europe) doesn’t offset customer concentration, because data infrastructure demand comes from the same five or six buyers regardless of where the manufacturing happens.
Two fragility elements the filings state plainly. First: tariffs. Flex explicitly discloses that it passes tariff costs to customers, which “would increase net sales, reduce operating margin, and adversely affect the timing of operating cash flow until paid tariffs are recovered.” In Q1 FY2027 the impact was ~1% on revenue and costs, negligible on profitability — but the mechanism is a revenue inflation without profit. On 02/20/2026 the US Supreme Court struck down tariffs imposed under IEEPA, and Flex sought and received a portion of refunds, with no disclosed material effect. Second: memory prices. DRAM and NAND have stayed elevated due to AI demand; Flex says it passes increases on “where contractual mechanisms allow,” but that this “could continue to favorably increase net sales, unfavorably affecting gross margin percentage and increasing inventory balances and working capital requirements.” So: part of the Q1 revenue growth of 20.6% is component-price inflation flowing through the books, not volume — a lower-quality kind of growth.
One detail that appears in no model but says something about operational risk: in FY2026 Flex recorded $51 mil of impairments and other charges following a missile strike on its Mukachevo, Ukraine facility. A manufacturer with a footprint in 30 countries carries geopolitical risk on the balance sheet, not just in the risk factors.
Management and capital allocation
Here Flex has a much better story than the GBL score suggests. Revathi Advaithi has been CEO since February 2019 and delivered, during her tenure, exactly what she said she would: exiting low-margin programs, restructuring the footprint, shifting the mix toward regulated markets and data infrastructure. GAAP operating margin climbed from ~2.5% pre-pandemic to 4.9% in FY2026, and diluted earnings per share from $1.21 (FY2021) to $2.33 (FY2026) — nearly doubled, in a period when revenue grew only 15.7% cumulatively (24,124 → 27,914 mil).
Buybacks are the clearest case of good capital allocation in the entire file. From successive 10-Ks, the average price paid per share, year by year:
| Fiscal year | Amount repurchased | Average price paid | Price at 03/31 of the year |
|---|---|---|---|
| FY2023 | $337 mil | $17.06 | 17.34 |
| FY2024 | $1,298 mil | $25.65 | 28.61 |
| FY2025 | $1,257 mil | $32.66 | 33.08 |
| FY2026 | $944 mil | $49.08 | 65.46 |
| Q1 FY2027 | 0 | — | 108.01 (today) |
Four years, $3.84 bil spent at a weighted average price of ~$31, with the stock now at $108. Weighted average diluted share count fell from 483 mil (FY2022) to 374 mil (Q1 FY2027) — a 22.6% reduction, i.e. nearly a quarter of the company bought back at a fraction of today’s price. And, just as important, in Q1 FY2027 they stopped buybacks completely ($0 versus $247 mil in the corresponding quarter), despite having $1.1 bil authorized and unused as of 03/31/2026. That’s textbook contrarian behavior: buy aggressively at $17-33, stop at $108-160. Compare that with the usual pattern — mechanical buybacks, accelerated exactly at highs.
The cash not used for buybacks went into an acquisition: EPP, $1,187 mil, closed 05/01/2026. Here the judgment is more reserved. Of the purchase price, $951 mil (80%) went to goodwill ($473 mil) and intangible assets ($478 mil: trademarks $132 mil over 15 years, customer relationships $256 mil over 20 years, know-how $46 mil over 10 years, order backlog $44 mil over 2 years). Tangible assets acquired are only $44 mil of PP&E and $99 mil of inventory. The company explicitly states it does not present pro-forma results because the acquisition wasn’t material to consolidated results — which, translated, means it paid $1.19 bil for a business that doesn’t move the needle on a $7.9 bil quarter. At acquired AR of $69 mil and inventory of $99 mil, EPP’s annualized revenue is probably in the $350-500 mil range, so a multiple of 2.4-3.4x sales for electrical equipment. Not absurd in the middle of a data-center power capacity rush, but the opposite of the discipline shown in buybacks: it’s an acquisition paid at the top of the cycle, with 80% of the price in intangible assets. Worth watching at the annual goodwill impairment test.
The second major capital-allocation act was monetizing Nextracker: an IPO in FY2024, with $552 mil raised from Nextracker share issuance and then the distribution/deconsolidation, which produced $373 mil of profit from discontinued operations and eliminated minority interests from the balance sheet ($239 mil of profit attributed to minority interests in FY2024, zero in FY2025 and FY2026). Structurally, it was a well-timed separation — Nextracker was taken public at peak enthusiasm for solar energy.
Debt management is conservative, with one caveat. As of 06/26/2026, total borrowings $5,240 mil, cash $2,840 mil, net debt $2,400 mil, i.e. ~1.17x TTM EBITDA (≈$2,050 mil). Weighted average interest rate is 4.8%. Interest coverage, TTM operating profit $1,449 mil over interest expense of ~$224 mil, is 6.5x. The $2.75 bil revolving facility is fully undrawn, and the company states it’s in compliance with all covenants and expects to remain compliant for the next 12 months. The caveat: the maturity schedule shows $2,349 mil due in fiscal FY2028 (i.e., between April 2027 and March 2028), including the $1.45 bil bridge loan taken in April 2026 to finance EPP (due November 2027) and notes due January 2028. Zero due in FY2027, $32 mil in FY2029, $685 mil in FY2030, $701 mil in FY2031 and $1,473 mil thereafter. So the entire refinancing wall is concentrated in a single fiscal year, exactly the year the CPI separation completes. With $2.84 bil cash and $2.75 bil of undrawn revolver, coverage exists; the risk is refinancing price, not liquidity.
Important O’Glove signal: no debt has been raised to fund dividends or buybacks. Flex pays no dividend at all. The $1.45 bil borrowed in April 2026 went entirely into an industrial acquisition. In FY2026, borrowings ($1,251 mil) and repayments largely offset each other, and the $944 mil of buybacks came from $1,685 mil of operating cash flow.
Key-person risk is now acute and underestimated. Under the leadership plan announced alongside the separation, Revathi Advaithi becomes CEO of SpinCo (CPI), and CFO Kevin Krumm follows her there too. The remaining Flex — the ITS and RMS segments, over $22 bil of revenue — will be led by Michael Hartung, with Advaithi remaining non-executive board chair “for a transition period.” In other words, the team that produced the turnaround leaves for the growth asset, and the large, mature business gets new leadership. Whoever buys FLEX today is buying two companies, one of which is losing both leaders who built its track record.
Skin in the game: weak. yfinance’s heldPercentInsiders is 0.61% of capital, versus 96.8% held by
institutions. For a $40 bil company, 0.61% means ~$245 mil distributed across the full board and executive
team — not insignificant in absolute value, but as a percentage it’s below what the Thorndike grid would
expect from an owner-allocator. Stock-based compensation was $142 mil in FY2026 ($125 mil FY2025, $152 mil
FY2024) and $51 mil in Q1 FY2027 alone, versus $34 mil a year earlier (+50%) — so management’s share
accumulation comes from grants, not market purchases.
What changed in the last 4 quarters
The comparison is between 06/27/2025 (Q1 FY2026) and 06/26/2026 (Q1 FY2027). The balance sheet grew by nearly a third in a year, and every large item has a specific explanation — some good, a few requiring attention.
| Item (mil USD) | 06/27/2025 | 06/26/2026 | Δ | Explanation |
|---|---|---|---|---|
| Total assets | 19,132 | 25,198 | +31.7% | EPP acquisition + working-capital surge |
| Cash | 2,239 | 2,840 | +26.8% | $1.5 bil net borrowings in the quarter |
| Receivables (AR) | 3,920 | 5,036 | +28.5% | revenue +20.6% and reduced factoring (see below) |
| Contract assets | 751 | 1,386 | +84.6% | the largest divergence on the balance sheet |
| Inventories | 5,210 | 6,453 | +23.9% | raw materials (memory) +13.2% in a single quarter |
| Other current assets | 1,347 | 2,522 | +87.2% | $1.5 bil customer-managed inventory (was $1.3 bil in March) |
| Net PP&E | 2,329 | 2,655 | +14.0% | CPI ramp capex |
| Goodwill | 1,369 | 1,831 | +33.8% | +$473 mil from EPP, −$6 mil from a disposal, −$5 mil FX |
| Intangible assets | 330 | 736 | +123% | +$478 mil from EPP purchase price allocation |
| Payables (AP) | 5,799 | 9,195 | +58.6% | extension of payment terms — the key to the cash flow |
| Deferred revenue + customer advances | 1,876 | 2,053 | +9.4% | $1.7 bil of working-capital advances from customers |
| Long-term debt | 3,003 | 5,219 | +73.8% | $1.45 bil bridge loan for EPP |
| Equity | 5,089 | 5,500 | +8.1% | $973 mil TTM profit minus buybacks |
| Shares outstanding | 376.2 mil | 369.4 mil | −1.8% | but +1.0% in the latest quarter alone |
Income statement, quarter by quarter (revenue / COGS from filings, mil USD): Q1 FY2026 6,575/5,987, Q2 6,804/6,181, Q3 7,058/6,373, Q4 7,477/6,747, Q1 FY2027 7,928/7,177. Gross margin rose from 8.7% to 9.4% year over year; net margin from 2.9% to 3.6%; net profit from $192 to $285 mil (+48%); diluted EPS from $0.50 to $0.76 (+52%). Profit is growing two and a half times faster than revenue — a good sign, achieved from mix (CPI from 25% to 28% of revenue) and from the disappearance of restructuring charges ($16 mil in COGS last year versus $4 mil now, plus a $2 mil reversal below the gross line). But watch one one-off component: “other net income” of +$37 mil in Q1 FY2027 (versus −$7 mil expense last year), part of which is the gain on the disposal of a North American site with $90 mil of proceeds. Without this element, pre-tax profit would have been ~$340 mil instead of $377 mil, and EPS growth would have been ~30%, not 52%.
The two variations requiring detailed explanation:
Contract assets, +84.6%. They rose from $616 mil (03/31/2025) to $751 mil (06/27/2025), $1,063 mil (03/31/2026) and $1,386 mil (06/26/2026). As a percentage of quarterly revenue: 11.4% a year ago, 17.5% now. Contract assets are revenue recognized over time (ASC 606) for products with no alternative use and an enforceable right to payment, but not yet invoiced. The fourfold-faster-than-revenue growth is explainable by CPI mix — integrated racks and made-to-order power systems, where over-time recognition is normal, and $62 mil came directly from the acquired EPP balance sheet. It isn’t manipulation, but it’s a structural quality shift: a growing share of revenue is recognized before invoicing, i.e. before there’s an enforceable receivable.
Payables, +58.6%. AP rose from $5,799 to $9,195 mil, while COGS grew 19.9%. Days payable outstanding (quarterly AP/COGS × 91) went from 88.1 to 116.6 days: +28.5 days in a year. For FY2026, the $2,858 mil AP increase was by far the largest item in the year’s operating cash flow. I checked whether this is reverse factoring: it isn’t. The note on supplier finance programs shows confirmed obligations of only $154 mil (03/31/2026) and $134 mil (06/26/2026) — irrelevant against $9.2 bil of AP. So it’s genuine term extension, obtained from the negotiating power of a $25 bil-a-year component buyer, plus the effect of the CPI model where the customer finances components ($1.7 bil of customer working-capital advances and $1.5 bil of customer-managed inventory). It’s a legitimate source of cash, but it’s an unrepeatable one: DPO can’t grow indefinitely, and normalizing back toward 88 days would mean a cash outflow on the order of $2.2-2.5 bil.
Cash conversion deteriorated visibly in the latest quarter, despite record profit. CFO was $276 mil in Q1 FY2027 versus $399 mil in Q1 FY2026 (−31%), even though net profit grew 48%. It’s the first time in the eight-quarter series that quarterly CFO falls below net profit. The difference is entirely in working capital: −$149 mil variation in Q1 FY2027 versus +$65 mil a year earlier, a $214 mil reversal. Capex doubled, from $133 to $236 mil. Result: FCF of $41 mil versus $268 mil, of which $24 mil is CPI separation costs paid in cash. On this quarter’s basis, FCF conversion from adjusted profit was ~11%; annual guidance is 40%. The three remaining quarters must deliver ~$644 mil of FCF for guidance to be met — over $200 mil per quarter, with capex of $420-455 mil per quarter.
Balance sheet analysis — Quality of Earnings (Thornton O’Glove method)
O’Glove’s premise: profit is an opinion, the balance sheet is closer to fact. Here the balance sheet tells a more nuanced story than the headlines — with a final quality verdict of MEDIUM, rising on some axes and falling on others.
(1) Receivables versus revenue — and the mandatory factoring correction
The raw comparison is alarming: AR +28.5% year over year versus revenue +20.6%, a 7.9-percentage-point divergence. But the raw figure is wrong, and O’Glove would be the first to demand the correction, because Flex sells receivables. From the “Trade Receivables Sales Programs” note (mil USD):
| Date | AR on balance sheet | Receivables sold and uncollected | Adjusted AR | Quarterly revenue | Raw DSO | Adjusted DSO |
|---|---|---|---|---|---|---|
| 06/27/2025 | 3,920 | 700 | 4,620 | 6,575 | 54.3 days | 70.2 days |
| 03/31/2026 | 4,679 | 500 | 5,179 | 7,477 | 57.0 days | 63.1 days |
| 06/26/2026 | 5,036 | 600 | 5,636 | 7,928 | 57.8 days | 64.7 days |
The volume of receivables sold per quarter fell from $0.9 bil (Q1 FY2026) to $0.6 bil (Q1 FY2027), and per fiscal year from $4.0 bil (FY2025) to $3.1 bil (FY2026) — a 22% reduction. The company confirms this indirectly in the MD&A: interest expense fell in FY2026 “primarily from lower costs due to reduced receivables factoring.” Corrected, adjusted AR grows 22.0% versus revenue +20.6% — essentially in line — and adjusted DSO improves by 5.5 days. The raw signal was a false positive created by a financing decision (less factoring, because it’s more expensive than corporate debt at current rates).
Over three years, adjusted DSO: ~68-70 days at 03/31/2024 and 03/31/2025 (AR $3,671 mil plus $700 mil factored, over quarterly FY2025 COGS), 63.1 days at 03/31/2026, 64.7 at 06/26/2026. The trend is stable to slightly better. Verdict on the receivables axis: positive.
A nuance not to be lost: if I add contract assets to the “unbilled or uncollected revenue” measure, the picture worsens. Adjusted AR plus contract assets: $5,371 mil at 06/27/2025 (74.4 days of revenue) versus $7,022 mil at 06/26/2026 (80.6 days). A 30.7% increase against revenue +20.6%. So on the most comprehensive measure there is a 10-point divergence, concentrated entirely in contract assets. This is the signal to watch.
(2) Inventory versus COGS, DIO and its structure
| Date | Inventory | Quarterly COGS | DIO | Raw materials | Work in progress | Finished goods |
|---|---|---|---|---|---|---|
| 06/27/2025 | 5,210 | 5,987 | 79.2 days | n/a | n/a | n/a |
| 03/31/2026 | 5,845 | 6,747 | 78.8 days | 4,834 | 474 | 537 |
| 06/26/2026 | 6,453 | 7,177 | 81.8 days | 5,471 | 457 | 525 |
DIO is practically unchanged: 79.2 → 81.8 days in a year when revenue grew 21%. But the structure is the good part, and it’s exactly the reverse of the worst signal in the O’Glove playbook. Finished goods FELL (537 → 525 mil, −2.2%) and work in progress too (474 → 457, −3.6%); the entire $608 mil increase is in raw materials (4,834 → 5,471, +13.2% in a single quarter). The company explains: high DRAM and NAND prices and accumulation of components for booked CPI programs. Rising raw materials with falling finished goods = production that’s selling, not merchandise piling up in a warehouse. It’s the opposite signal from collapsing demand.
The countervailing risk: if memory prices collapse, a $5.47 bil raw-materials stock bought at high prices generates net realizable value write-downs. Flex values inventory at the lower of cost (FIFO) or net realizable value, so the exposure is real. At a 20% decline in component prices on a third of the inventory, the write-down would be on the order of $350 mil — nearly 40% of annual net profit.
Additionally, $1.5 bil of customer-managed inventory sits in “other current assets” (it was $1.3 bil in March). That isn’t Flex’s ownership risk, but it consumes balance sheet space and is partly financed by customer advances ($1.7 bil).
Verdict on the inventory axis: positive, with a memory-price exposure to watch.
(3) Debt — gross, net, maturity, cost, covenants, three-year evolution
| ($ mil) | 03/31/2024 | 03/31/2025 | 03/31/2026 | 06/26/2026 |
|---|---|---|---|---|
| Current debt | ~0 | 1,209 | 0 | 0 |
| Long-term debt | ~3,680 | 2,483 | 3,751 | 5,219 |
| Cash | 2,474 | 2,289 | 2,389 | 2,840 |
| Net debt | ~1,200 | 1,403 | 1,362 | 2,400 |
| Net debt / EBITDA | ~0.8x | ~0.8x | ~0.7x | ~1.17x |
| Equity | ~5,100 | 5,002 | 5,144 | 5,500 |
Maturity schedule at 06/26/2026 (mil USD): FY2027 0, FY2028 2,349, FY2029 32, FY2030 685, FY2031 701, thereafter 1,473, total 5,240. Weighted average rate 4.8% (was 4.9% at 03/31/2026). $2.75 bil revolving facility, fully undrawn. The company states compliance with all covenants and expects to remain compliant for 12 months; numerical covenant thresholds aren’t quantified in the 10-Q — a common transparency gap in interim filings, but a gap nonetheless, and I note it as such.
The concentration in the 12-24 month window is the only real problem: $2,349 mil, i.e. 45% of total debt, comes due in a single fiscal year (FY2028), including the $1.45 bil loan taken in April 2026 for EPP and due November 2027. It’s the same window in which the CPI separation completes, so refinancing will be negotiated by two entities still sharing a balance sheet. Coverage is comfortable ($2.84 bil cash plus $2.75 bil revolver), but cost is uncertain. Worth noting a technical detail from the debt note: the $1.45 bil loan was syndicated in May 2026, with $1.38 bil transferred to new lenders, and the transfer was accounted for as debt extinguishment — which is why Q1’s financing cash flow shows gross borrowings of $2,830 mil and repayments of $1,385 mil, figures that overstate actual financing activity.
O’Glove signal: no debt used for dividends or buybacks. Flex pays no dividend, and the $944 mil bought back in FY2026 came from $1,685 mil of CFO. In Q1 FY2027, the $1.5 bil net borrowings entirely funded the EPP acquisition ($1,134 mil net of cash acquired), and buybacks were stopped. Discipline remains intact.
(4) Discretionary expenses and changes in estimates
| ($ mil) | FY2024 | FY2025 | FY2026 | Q1 FY2027 (annualized) |
|---|---|---|---|---|
| SG&A | 922 (3.49% of revenue) | 904 (3.50%) | 1,052 (3.77%) | 1,336 (4.21%) |
| Restructuring + impairment | 175 | 86 | 138 | ~8 |
| Intangibles amortization | 70 | 70 | 68 | ~92 (with EPP) |
| PP&E depreciation | 431 | 439 | 457 | ~560 |
| Capex | 530 | 438 | 633 | 1,500-1,600 (guided) |
| Capex / depreciation | 1.06x | 0.86x | 1.20x | ~2.5x |
| Stock-based compensation | 152 | 125 | 142 | ~204 |
| Effective tax rate | ~18% | 18.1% | 23.0% | 24% |
Three observations.
SG&A grows faster than revenue, but for identifiable reasons. In FY2026, +$148 mil (+16.4%) against revenue +8%, of which $44 mil is “legal and other costs associated with portfolio optimization projects, including the CPI separation.” In Q1 FY2027, +$101 mil (+43%) against revenue +20.6%, of which $53 mil separation costs and $18 mil increase in stock-based compensation. Stripped of the two, SG&A grows $30 mil, i.e. +12.9% — below revenue growth. It isn’t operational erosion, it’s the cost of a corporate transaction. The signal is that SG&A should return to ~3.6-3.8% of revenue after the separation; if it doesn’t, the margin-improvement thesis wobbles.
R&D isn’t presented separately either in the income statement or the notes — it’s embedded in COGS and SG&A, and the segment definition mentions “corporate general, research and development, and administrative expenses” allocated to segments. For a company now claiming to sell proprietary power and cooling products, not just labor, the absence of an R&D line is a real information gap: I can’t verify whether CPI’s product advantage is supported by internal investment or bought (EPP, Crown, the Poland acquisition). I flag this explicitly as missing data.
The underinvestment year. In FY2025 capex ($438 mil) was below PP&E depreciation ($439 mil) — a 0.86x
ratio. That year produced a reported FCF of $1,082 mil, the best in the series, partly because there was no
investment. FY2026 corrected to 1.20x, and FY2027 goes to 2.5x. Conclusion for valuation: the historical
FCF of ~$1.07 bil per year is not a repeatable earnings power, it’s the average of one underinvestment year
and one normal-investment year, and the correct maintenance base is somewhere around depreciation
($570-600 mil per year).
Accounting estimates: I found no change of policy or estimate that inflates profit. The effective tax rate has risen (18.1% → 23.0% → 24%), so EPS isn’t being helped by taxation — it’s actually held back by it. The two discrete items in FY2026, a $19 mil provision from a settlement with a foreign tax authority for fiscal years 2010-2020 and the disappearance of a $26 mil Brazilian benefit from FY2025, are disclosed and work against profit. ASU 2024-03 (expense disaggregation) doesn’t apply until Q4 FY2028; ASU 2025-12 from Q1 FY2028. No early adoption. The OECD’s Pillar Two (15% global minimum tax) had no material impact in FY2026, but is a latent risk for a group headquartered in Singapore (17% statutory rate) with tax exemptions in China, Costa Rica, Malaysia, the Netherlands and Israel.
(5) Summary: net profit versus CFO divergence (accruals)
| ($ mil) | FY2024 | FY2025 | FY2026 | TTM Jun-2026 |
|---|---|---|---|---|
| Net profit (continuing) | 872 | 838 | 880 | 973 |
| CFO | 1,326 | 1,504 | 1,685 | 1,562 |
| CFO − net profit | +454 | +666 | +805 | +589 |
| Total assets | ~19,000 | 18,381 | 22,060 | 25,198 |
| Sloan accruals (NI−CFO)/assets | −2.4% | −3.6% | −3.6% | −2.3% |
The accruals ratio is negative across all four years — CFO consistently exceeds net profit by $450-800 mil, exactly what you’d expect from a company with ~$570 mil of annual depreciation and no aggressiveness in recognition. On the Sloan measure, earnings quality is good, and the deterioration from −3.6% to −2.3% in TTM is explained by the reversal in Q1 FY2027, when CFO ($276 mil) fell below net profit ($285 mil) for the first time.
(The data pack reports “+nan%” for this metric and for revenue comparisons, because yfinance doesn’t have the quarterly income statement for Q1 FY2027; the figures above are calculated from SEC filings and XBRL companyfacts tags, not from the data pack.)
Quality verdict: MEDIUM
Arguments for “high”: consistently negative accruals across four years, improving adjusted DSO, finished goods falling while sales grow 21%, rising tax rate (i.e., no fiscal support), no significant reverse factoring ($134 mil against $9.2 bil of AP), no debt for buybacks, rising margins across all three segments simultaneously.
Arguments that pull the verdict down to “medium”: (a) contract assets grow 84.6% against revenue +20.6%, and on the comprehensive measure adjusted AR plus contract assets grows 30.7% against 20.6%; (b) DPO +28.5 days in a year is an unrepeatable cash source that, on normalization, would cost 2.2-2.5 bil; (c) CFO fell 31% in a quarter with profit +48%; (d) no published R&D line for a company now claiming a product advantage; (e) Q1’s EPS growth is helped by a one-off gain on a site disposal.
Signal to watch at the 10/28/2026 report (Q2 FY2027): the contract assets / quarterly revenue ratio, threshold 17%. If it stays there or rises, CPI revenue is being recognized ever earlier relative to invoicing and quality falls to WEAK. Second signal, equally important: if the payables balance falls below ~$8.5 bil with COGS still rising, DPO normalization has begun and FY2028 FCF will fall well below every model in this report.
CEO profile — Outsider traits (William Thorndike method)
The subject is Revathi Advaithi, CEO since February 2019 — seven and a half years in the role, so not “too new.” But the profile now reads with a large asterisk: under the leadership plan announced alongside the separation, Advaithi leaves Flex to run SpinCo (CPI), taking CFO Kevin Krumm with her; the remaining Flex moves to Michael Hartung, and Advaithi remains only non-executive board chair, transitionally. Bill Watkins is expected to be non-executive chair at SpinCo.
Capital allocation as priority number one — good score. There is no dividend. All free cash has gone, in order, into buybacks ($3.84 bil over four years), debt reduction and, since 2026, capacity acquisitions. The Nextracker monetization ($552 mil raised, $373 mil discontinued-operations profit, deconsolidation) was timed at peak enthusiasm for solar. The CPI separation is the second structural move of the same type.
Contrarian buybacks — very good score, the clearest element in the file. Average prices paid ($17.06 → $25.65 → $32.66 → $49.08) rise, but stay below the fiscal year’s closing price every year, and the complete stop in Q1 FY2027, with $1.1 bil authorized and unused, at a price of $108-160, is exactly the opposite of mechanical buying. It’s the best discipline signal in the entire file and deserves emphasis: most boards accelerate buybacks at highs, because that’s when “things are going well.”
Decentralization and frugality — medium score. SG&A at 3.5-3.8% of revenue is below industry average for a company with 30 countries and three segments, and successive restructurings ($175 mil FY2024, $86 mil FY2025, $87 mil FY2026, all severance-related) show management cutting excess capacity without drama. Working against this: separation costs ($44 mil FY2026, $53 mil in Q1 FY2027 alone, heading toward ~$250-340 mil per year per the FCF conversion revision from 60% to 40%) are large for a corporate transaction and not frugal.
Independence from Wall Street — weak score. Flex gives quarterly and annual guidance, with a range on revenue, adjusted operating margin and adjusted EPS, plus segment growth. Communication is built around adjusted EPS — a metric from which stock-based compensation ($204 mil annualized), intangibles amortization (~$92 mil) and restructuring are excluded. The gap between guided adjusted EPS (4.42-4.74) and the GAAP EPS I estimate (~3.32) is ~28%, and public communication is done almost exclusively on the adjusted figure. Free cash flow per share isn’t the primary communication metric; it’s mentioned as “conversion,” which is a gentler way of saying it fell 85% in a quarter.
Skin in the game and incentives — weak score. 0.61% of capital held by management and the board, versus 96.8% institutions. Stock-based compensation is growing 50% year over year ($34 → $51 mil per quarter). I didn’t identify significant open-market purchases by management — the proxy isn’t part of the downloaded SEC file, so I mark this point as unverified directly, not disproven.
Proven contrarian behavior — good score. Three pieces of evidence: aggressive buybacks in FY2024-FY2025 when the stock was at $25-33 and consensus treated Flex as an EMS without a future; stopping buybacks at highs; the pivot toward CPI begun in FY2023-FY2024, when the segment had $3.2 bil of revenue and nobody called it “AI infrastructure.”
Outsider score: 3.5 of 5. Capital allocation and buyback discipline are textbook; independence from Wall Street and skin-in-the-game pull it down.
Key-person risk: high, and it’s the main governance risk of this investment. Whoever buys FLEX at $108 is buying, in about 70% of revenue, a company (ITS plus RMS) that after Q1 2027 will be run by a new CEO with no public capital-allocation track record, with the CFO departed and the architect of the turnaround remaining only as transitional chair. The other ~30% piece (CPI/SpinCo) gets the good team, but also the most stretched valuation. There are no related parties or controlling shareholder complicating the picture — capital is widely held institutionally (96.8%), which is neutral-to-positive on the governance axis, but also means there’s nobody positioned to oppose a bad decision.
Accounting red flags
Ordered by how much they’d change the thesis, not by accounting severity.
1. Contract assets, +84.6% year over year against revenue +20.6%. From $751 to $1,386 mil, from 11.4% to 17.5% of quarterly revenue. $62 mil came from the acquired EPP balance sheet, so organic growth is ~76%. It’s revenue recognized before invoicing. The mix explanation (made-to-order racks and power systems, recognized over time under ASC 606) is credible and isn’t a violation, but it shifts recognition earlier relative to collection and is the only balance-sheet item moving decoupled from everything else. This is red flag number one.
2. Days payable outstanding, +28.5 in a year (88.1 → 116.6). The $2,858 mil AP increase in FY2026 was by far the largest item in the year’s operating cash flow, in a year when COGS grew only 7%. I ruled out reverse factoring as the explanation — confirmed obligations across the six supplier finance programs are $134-154 mil, i.e. 1.5% of AP. It isn’t an accounting irregularity; it’s an unrepeatable cash source that funded $944 mil of buybacks in FY2026 and that, on normalization, costs 2.2-2.5 bil. Any model that extrapolates FY2026’s CFO implicitly extrapolates a permanent extension of terms.
3. The GAAP-adjusted chasm, 28%. Guided adjusted EPS $4.42-4.74; GAAP EPS I estimate $3.32 for
FY2027. The difference comes from stock-based compensation ($204 mil annualized, +50% year over year),
intangibles amortization ($92 mil with EPP included), restructuring and separation costs (~$250-340 mil).
Stock-based compensation is a real expense, paid in diluted ownership; excluding it from the communication
metric is standard sector practice, but at $204 mil a year it means 16-20% of GAAP net profit.
4. Dilution has resumed. Shares outstanding fell 1.8% over four quarters (376.2 → 369.4 mil), but rose 1.0% in the latest quarter alone (365.7 → 369.4), because buybacks stopped and 3 million shares were issued from vested restricted units. With buybacks stopped and stock-based compensation growing 50%, the short-term trend is net dilution of ~1-1.5% per year. It’s a rational allocation choice (the money went into EPP at a share price of $108-160), but it flips the sign of a component the last four years had made positive.
5. EPP’s goodwill and intangibles: $951 mil out of a $1,187 mil price (80%). Tangible assets acquired are $44 mil PP&E and $99 mil inventory. The company states the acquisition “wasn’t material to consolidated results,” meaning it paid $1.19 bil for a business with probably $350-500 mil of annual revenue, at a peak moment for data-center electrical equipment demand. The $44 mil order backlog amortizes over two years ($22 mil/year drag on GAAP profit), and the group’s total goodwill rose to $1,831 mil. The first annual impairment test after the power cycle normalizes is the event to watch. The measurement period remains open for 12 months, so the allocation can still change.
6. Segment recast in Q4 FY2026. Segments were redefined (“Flex Agility” and “Flex Reliability” became ITS and RMS, and CPI was created as a new segment) right in the quarter before the separation announcement. Historical data was recalculated, which is procedurally correct, but means the CPI margin series we read (9.4% / 10.2% / 9.2%) is a retroactive reconstruction by management, not a series reported in real time. Not an irregularity; a limit on verifiability.
7. Revenue growth from component inflation and tariffs. The company explicitly states that high DRAM and NAND prices “could continue to favorably increase net sales, unfavorably affecting gross margin percentage,” and that it passes tariff cost to customers, which “would increase net sales, reduce operating margin.” So an undetermined portion of the 20.6% revenue growth is pass-through without profit. Tariffs impacted ~1% of revenue and costs in Q1; for memory there’s no quantification — the second missing data point I flag.
8. The Q1 one-off gain. “Other net income” of +$37 mil (versus −$7 mil expense last year), tied to the disposal of a North American site with $90 mil of proceeds. Contributes ~$0.08 to diluted EPS of $0.76. It isn’t hidden, but the “GAAP EPS +52%” headline includes it.
What is NOT a red flag, though it looks like one: the 28.5% AR increase (explained by reduced factoring, adjusted DSO improves by 5.5 days), the 23.9% inventory increase (entirely in raw materials, with finished goods falling), the tax rate (rising, i.e. working against profit) and Sloan accruals (negative across all four years).
Triangulated valuation
The FCF bridge
Before any model, the definition of the flow. CFO − capex isn’t the default answer, so I specifically
checked the consolidated cash flow statement in the FY2026 10-K and the Q1 FY2027 10-Q.
| Item | FY2026 (mil USD) | TTM Jun-2026 (mil USD) | Source / tag |
|---|---|---|---|
| Reported CFO | 1,685 | 1,562 | NetCashProvidedByUsedInOperatingActivities |
| − PP&E purchases | −633 | −736 | PaymentsToAcquirePropertyPlantAndEquipment |
| + Proceeds from PP&E disposals | +8 | +7 | cash flow statement, investing |
| = FCF (company’s definition) | 1,060 | 833 | explicitly reconciled in MD&A |
| Interest paid | 212 | ~224 | InterestPaidNet — already in CFO (US GAAP) |
| Interest in investing or financing | 0 | 0 | verified: no such line exists |
| Finance lease principal | immaterial | immaterial | explicitly disclosed: “a number of finance leases with an immaterial impact” |
| Operating lease payments | in CFO | in CFO | US GAAP, operating expense |
| Dividends to minority interests | 0 | 0 | no minority interests after Nextracker deconsolidation (FY2024) |
| Dividends to shareholders | 0 | 0 | Flex pays no dividend |
| = FCFE | 1,060 | 833 | identical to simple FCF |
The divergence between simple FCF and FCFE is 0.8% (just disposal proceeds), below the 10% threshold.
The reason is that Flex reports under US GAAP, where interest paid and operating leases sit in operating
activities. It isn’t the case of a 20-F-style IFRS filer, where interest would be in financing and
CFO − capex would be a flow before interest, not one to the shareholder. Here the two definitions
coincide, and the check was done line by line, not assumed.
Consequence for the DCF: net debt is NOT subtracted again. FCFE is already post-interest, so it’s discounted at the cost of equity and gives the share value directly. Subtracting the $2,400 mil of net debt a second time would double-tax the same thing. The Monte Carlo script warns that the balance sheet gives a calculated net debt of $1,927 mil (yfinance total debt of $5,930 mil, which includes leases, minus cash and short-term investments) — the normalization to 0 is intentional and is justified exactly here.
External check, mandatory. My TTM FCFE yield is 833 / 39,899 = 2.09%. Published figures, for
comparison: yfinance reports freeCashflow of $626 mil (1.57%); the company-reported FCF for FY2026 was
$1,060 mil (2.66%); the company’s own FY2027 guidance — ~40% conversion of adjusted net profit of ~$1,713
mil — gives ~$685 mil (1.72%). My figure sits in the middle of the published range of 1.57-2.66%. It isn’t
an outlier and doesn’t contradict the market; it only contradicts the price.
The owner earnings base used in the DCF
I start from TTM FCFE of $833 mil and make a single normalization: plus ~$50 mil of CPI separation costs
paid in cash within the TTM window ($24 mil explicitly disclosed in Q1 FY2027 plus the cash portion of
the $44 mil of “legal and other portfolio optimization costs” from FY2026). This yields $883 mil, rounded
to $900 mil, the figure given as --oe to the simulation.
Why not other bases:
- Not $1,060 mil (FY2026): the year benefited from a $2,858 mil increase in payables, an unrepeatable source, and followed a year (FY2025) with capex below depreciation.
- Not $685 mil (guided FY2027): it’s a peak investment year (capex 2.5x depreciation) plus one-off separation costs; it’s a bear-scenario base, not a base case.
- Not $1,015.7 mil (the tracker’s mechanical base, margin × revenue, three-year average): this includes
exactly the underinvestment year FY2025 and ignores the capex cycle now beginning. Without an explicit
--oe, the simulation would have started here — exactly the error pattern that cost a rerun at AFYA. - Cross-check: owner earnings in the Buffett sense (net profit plus depreciation minus maintenance capex, with maintenance capex approximated by depreciation for a company whose historical capex ≈ depreciation) gives $973 mil on a TTM basis and $880 mil on FY2026. The $900 mil sits between the two.
The five models
Model 1 — DCF on FCFE, bear scenario. Base $685 mil (the company’s FY2027-guided FCF, at 40% conversion), g1 = 6% (baseline EMS growth, no AI premium), r = 11% (beta 1.654, cyclicality, customer concentration, separation uncertainty), gt = 2.0%, net debt 0, 374 mil diluted shares. Intrinsic value $25.18, MOS −76.7%. Terminal weight: 45%.
Model 2 — DCF on FCFE, base scenario (identical to the Monte Carlo assumptions). Base $900 mil, g1 = 9% over five years and 4.5% in years 6-10, r = 10.5%, gt = 2.5%, net debt 0. The 9% is three times the historical five-year revenue growth (2.96% CAGR FY2021-FY2026) and less than half of the guided FY2027 growth — recognizing CPI without extrapolating the peak year. The 10.5% is below the mechanical CAPM (4.2% + 1.654 × 5% = 12.5%), because beta also reflects the speculative rally in June, not just business risk. Intrinsic value $42.87, MOS −60.3%. Terminal weight: 51%, below the 60% threshold above which the valuation stops being about the business and becomes about a single division by (r − gt).
Model 3 — DCF on FCFE, bull scenario, built generously. Here I let the market’s thesis defend itself fully. I assume FY2028 delivers: revenue ~$45 bil (CPI at ~$20 bil after the guided growth above 80%), adjusted operating margin 8.5%, adjusted net profit ~$2.9 bil, FCF conversion 70% after the capex peak → $2.0 bil FCF in FY2028. Discounted back two years at 10% gives an equivalent base today of $1,650 mil. On top of that I apply g1 = 12% over ten years, r = 9.5% (a larger, less cyclical company after the separation), gt = 3%. Intrinsic value $112.57, MOS +4.2%. In other words: the scenario where everything goes right — FY2028 doubles free cash flow and then compounds at 12% for a decade and the market accepts a 9.5% cost of capital — barely reaches today’s price. With the same $1,650 mil base but g1 = 8%, value falls to $89.28 (−17.3%).
Model 4 — Earnings Power Value (Greenwald), no growth. Mid-cycle average GAAP operating margin, calculated from FY2024 (3.23%), FY2025 (4.53%), FY2026 (4.90%) and Q1 FY2027 (4.94%), is 4.40%. Applied to TTM revenue of $29,267 mil, that gives normalized EBIT of $1,288 mil; after tax at 23%, NOPAT $992 mil. WACC calculated from the current structure (equity $39.9 bil at 11% cost, debt $5.24 bil at 4.8% deductible) is 10.15%, rounded to 10%. Firm EPV $9,916 mil, minus net debt of $2,400 mil, gives $7,516 mil of equity value, i.e. $20.35/share, MOS −81.2%. A milder variant, at the current 4.94% margin on guided FY2027 revenue: $28.98 (−73.2%). A variant at the guided adjusted margin of 7.1% (which excludes stock-based compensation and amortization, so isn’t an economic margin): $44.49 (−58.8%). Even the EPV’s upper bound is under half the price. This is the model’s central conclusion: with no growth, Flex isn’t worth anywhere near $108. The entire market value is the value of future growth — which is legitimate for a growth company and fatal for a deep-value thesis.
Model 5 — Historical multiples over five fiscal years. P/E at each fiscal year’s close, on reported diluted EPS:
| Fiscal year | Price at 03/31 | Diluted EPS | P/E |
|---|---|---|---|
| FY2022 | 13.98 | 1.94 | 7.2x |
| FY2023 | 17.34 | 1.72 | 10.1x |
| FY2024 | 28.61 | 2.28 | 12.5x |
| FY2025 | 33.08 | 2.11 | 15.7x |
| FY2026 | 65.46 | 2.33 | 28.1x |
| today | 108.01 | 2.59 (TTM) | 41.7x |
Median 12.5x, mean 14.7x. Applied to my estimated GAAP EPS for FY2027 ($3.32, i.e. the adjusted $4.58 minus stock-based compensation, intangibles amortization, restructuring and separation costs, after tax): $41.50, MOS −61.6%. At the mean of 14.7x: $48.80 (−54.8%). At a multiple of 20x on adjusted EPS: $91.60 (−15.2%). To justify $108 on historical multiples, you need 28.1x — the series’ absolute maximum, hit only once, six months ago — applied to adjusted EPS: $128.70. On EV/EBITDA the picture is identical: 20.2x today (EV $42.99 bil / TTM EBITDA $2.12 bil) versus 13.4x at 03/31/2026 and 8.3x at 03/31/2025.
Monte Carlo (20,000 scenarios)
Run with Model 2’s assumptions: --oe 900 --g1 0.09 --r 0.105 --gt 0.025 --nd 0 --px 108.01. The table is
copied as-is from the script’s output, with no recalculation:
Reference price: 108.01 · median intrinsic value: $43.15
| Percentile | P5 | P10 | P25 | P50 | P75 | P90 | P95 |
|---|---|---|---|---|---|---|---|
| MOS (%) | -78.5 | -75.3 | -69.1 | -60.0 | -48.5 | -34.9 | -25.2 |
- Probability the stock is undervalued (MOS > 0): 0.8%
- Probability of a margin of safety above 30%: 0.1%
- Central assumptions: OE 900 · g1 9.00% · r 10.50% · gt 2.50% · net debt 0 · shares 374 mil · FX 1
- Dispersions: OE ±25% (lognormal), g1 σ=4.5%, r σ=1.0%, gt σ=0.50%
- λ (access to net cash): not applied — net debt is positive, so it’s never clipped
Interpretation. The P10-P90 range of the margin of safety spans 40.4 percentage points (−75.3% … −34.9%), which for a simulation of this type is a tight distribution. What widens the range is, in order, the owner earnings base (lognormal ±25%: at +25% the base would be $1,125 mil, i.e. above the FY2026-reported FCF) and the growth rate (σ = 4.5 points, so plausible scenarios run from 0% to 18%). What’s remarkable isn’t the width, but the position: the entire distribution sits below zero. Not even the 95th percentile — the combination where the base is large, growth exceeds 18% and the discount rate is below 9.5% — reaches the current price, stopping at −25.2%.
An undervaluation probability of 0.8% doesn’t mean “the stock will fall 60%.” It means that, under these assumptions, there’s almost no reasonable combination of base, growth, rate and terminal growth that justifies the price. And that’s exactly the correct diagnosis: the market’s thesis isn’t a more optimistic variant of my model, it’s a different model, the one in Model 3, which starts from a flow base of $1.65 bil instead of $0.9 bil. The difference between the two isn’t a matter of marginal assumptions, but of whether or not you believe FY2028 delivers $2 bil of free cash flow. The simulation can’t arbitrate between the two models, but it clearly shows that within the conservative model there’s no escape hatch.
For position sizing: at a 0.8% probability of undervaluation and a 0.1% probability of MOS above 30%, the correct position size at $108 is zero. Not because the bull thesis is impossible, but because you aren’t paid for the risk of being wrong: if you’re right, you gain about what the market already pays; if you’re wrong, the central model says −60%. The asymmetry is inverted from what a deep-value position requires.
Triangulation summary
| # | Model | Intrinsic value | MOS |
|---|---|---|---|
| 1 | DCF FCFE bear (base $685 mil guided FY2027) | $25.18 | −77% |
| 2 | DCF FCFE base (900 mil, g1 9%, r 10.5%) | $42.87 | −60% |
| 3 | DCF FCFE bull (base $1,650 mil FY2028E, g1 12%, r 9.5%) | $112.57 | +4% |
| 4 | EPV Greenwald (mid-cycle margin 4.40%, WACC 10%) | $20.35 | −81% |
| 5 | Historical multiples (median P/E 12.5x on GAAP FY2027E EPS) | $41.50 | −62% |
| Range | $20.35 — $112.57 | −81% … +4% | |
| Median | $41.50 | −62% | |
| Monte Carlo, median | $43.15 | −60% |
The deterministic median ($41.50) and the simulation’s median ($43.15) converge within ~$2 of each other, which is a useful check: two independent methods see the same intrinsic value of ~$42-43. The range is wide — from $20 to $113 — but skewed in the wrong direction for a buyer at $108: four of five models are under half the price, and the fifth was specifically built to be generous.
Pre-mortem
Three scenarios in which the “AVOID at $108” thesis turns out wrong, with each one’s concrete trigger.
Scenario 1 — The flow base is wrong by a factor of 2, because I confused an investment year with earning power. My central assumption is that $900 mil USD is the company’s cash-generation power. But FY2027 capex ($1.5-1.6 bil) is explicitly stated by management as a peak — “capital spending is expected to peak at $1.4-1.6 billion in fiscal year 2027.” If in FY2029 capex returns to $700-800 mil while revenue stays above $40 bil and adjusted operating margin at 8%, free cash flow would be $1.9-2.2 bil, i.e. 2.3 times my base. At $108 that would mean an FCF yield of 5.3% on a business still growing — a reasonable, even attractive, valuation. Model 3 is exactly this scenario and gives $112.57. What would confirm it: quarterly capex below $300 mil in FY2028 combined with CFO above $600 mil per quarter; or a reported FCF conversion above 60% for a full year without separation costs. Subjective probability: ~25%. It’s the most likely way I’m wrong, and it’s why the verdict is “avoid,” not “short.”
Scenario 2 — Sum of the parts after separation exceeds today’s market cap. CPI will have ~$11 bil of revenue in FY2027 (6.6 bil × 1.70) with a segment margin of 9.7-10.2%, so segment profit of ~$1.1 bil. Comparable data-center power infrastructure companies trade at multiples of 25-35x profit in this cycle. At 28x on an attributable net profit of ~$700 mil, SpinCo would be worth ~$19.6 bil. The remaining Flex (ITS plus RMS, $22+ bil revenue, segment profit ~$1.35 bil, net profit ~$750 mil) at a mature-EMS multiple of 13-15x would be worth $9.8-11.3 bil. Total: $29-31 bil — still below today’s $39.9 bil, but at 32x for SpinCo the sum reaches ~$33-35 bil, and at 40x it exceeds $39 bil. In other words, the separation can justify the price if the market grants SpinCo a pure-growth-company multiple. What would confirm it: the pro-forma figures in the Form 10 (expected September 2026, i.e. this month) showing a standalone CPI margin above 12% and a balance-sheet structure with low net debt. Subjective probability: ~20%.
Scenario 3 — The margin has structurally shifted and the EPV model is a relic. My EPV uses a mid-cycle average operating margin of 4.40%, calculated over four periods, two of which (FY2024, FY2025) belong to the old company. If post-separation mix stabilizes the remaining Flex’s adjusted operating margin at 6.5-7% and SpinCo’s at 11-13%, the correct “cycle” margin isn’t 4.4%, it’s 7%+, and the EPV rises from 20 to over $44 from that single change alone. It’s the Celestica scenario: an EMS that, through data-center mix, re-rated from 8x to 30x profit in three years and stayed there because the margin genuinely shifted. What would confirm it: CPI segment margin above 12% for two consecutive quarters, or consolidated GAAP operating margin above 6.5% for two quarters. Subjective probability: ~20%.
The scenarios in which I’m right, for symmetry. DPO normalization (a reversion from 116 to 90 days would consume 2.2-2.5 bil of cash and wipe out two years of free cash flow); loss or renegotiation of the 12% customer (~$4 bil of revenue, mostly CPI); a correction in hyperscaler capex, which would simultaneously hit CPI growth, margin and multiple; a memory-price collapse that would produce write-downs on a $5.47 bil raw-materials stock; or an expensive refinancing of the $2,349 mil due in FY2028, exactly during the separation. Any one of these alone would bring the stock back below $70, and two of them together below $55, i.e. into the zone where Model 2 would give a real margin of safety.
Verdict compared with the GBL tracker score
The tracker’s prior GBL score, evaluated 09/06/2026: 30.8% — AVOID (Graham 2.5, Buffett 3.5, Lynch 3.0, total 8.25 of 30, Piotroski F-Score 6.0, denominator 60.0).
Convergence on verdict, divergence on reasons. We both reach AVOID, but from partly different angles, and the difference is informative.
Where the tracker is right and my analysis confirms it. The Graham component of 2.5 of 10 is severe and
correct: the Graham number calculated from EPS 2.48 and book value per share 14.89 is
√(22.5 × 2.48 × 14.89) = $28.83, i.e. a −73.3% discount versus the current price (tracker’s disc_gn:
−0.7331 — matches). P/B of 7.25 and trailing P/E of 43.6 are both outside any Graham criterion. Debt to
equity of 107.8% (calculated on yfinance total debt, which includes leases; on pure financial debt it’s
95.3%) crosses Graham’s threshold. On the Graham axis there’s no ambiguity: the stock is expensive by any
classic value criterion.
Where my analysis is more generous than the tracker’s. The Buffett component of 3.5 of 10 underestimates today’s business quality. ROIC of 16.2% (FY2026) and 14.1% (TTM) is above the ~10% cost of capital; margins are rising in all three segments simultaneously; earnings quality as measured by Sloan accruals is negative in all four years; and capital allocation — $3.84 bil bought back at an average price of ~$31, with a complete stop at $108-160 — is better than most companies in the tracker. If I re-scored Buffett purely on operational quality and capital allocation, I’d give 5.5-6 of 10, not 3.5. What pulls it down is something else, and the tracker has no way of seeing it: the absence of a verifiable moat on 38% of revenue (ITS), the 12% concentration on a single customer, and the fact that the CEO who built the track record is leaving for the other company after the separation.
Where my analysis is more severe than the tracker’s. The Lynch component of 3.0 is too lenient if you
look at the PEG the tracker displays: the forward peg of 0.94 looks attractive (under 1 = “cheap for
growth” in the Lynch grid), but peg_trailing is 3.17. The difference comes from the forward PEG
resting on a yfinance forwardEps of $7.11, an unreliable figure — it’s 2.9 times TTM EPS and 55% above the
company’s own adjusted guidance ($4.42-4.74). A PEG built on a mis-estimated EPS is a false cheapness
signal, and it’s the only indicator in the tracker that looks good here. On the company’s actual guidance,
forward PEG is 0.59 if you accept the 40% growth of adjusted EPS as sustained — and 2.6 if you use a
structural growth rate an EMS can sustain long-term. Lynch would have demanded, here, a check on whether the
denominator’s growth rate is repeatable; it isn’t.
The divergence with analyst consensus is the largest in the entire file and deserves stating directly.
The tracker shows a disc_consens of +48.6%, i.e. an average target of $160.50 from 10 analysts — 49%
above the current price. My triangulated median value is $41.50. The ratio between the two is nearly 4x.
I’m not claiming the market is wrong by 75%; I’m claiming our models answer different questions. The $160
consensus is a 12-month price target, built on the multiple the market currently grants to estimated FY2028
flow, and Model 3 shows that reasoning isn’t absurd — it gives $112.57 with assumptions an optimistic
analyst would consider moderate. My value is a ten-year holder’s intrinsic value, with a margin-of-safety
requirement. A deep-value investor doesn’t buy at $108 because they aren’t paid for the risk of being wrong;
a growth investor with a one-year horizon can also be right. What can’t be sustained is that FLEX would be,
today, a value position.
Worth noting the price context too, since it changes the meaning of the numbers: the stock was at $65.46 at 03/31/2026, rose to $162.07 in June 2026, and has corrected to $108.01 now — a 33% drop from the peak in three months. Someone looking only at the correction sees “a good stock on sale”; someone looking at the starting point sees a stock still 65% above the level of six months ago, on a business whose free cash flow fell 85% in the same quarter.
My position versus the 30.8% score: convergent on conclusion, with the recommendation to raise the Buffett component to 5.5-6 and to ignore the forward PEG for this ticker. The verdict remains AVOID at $108.01. Levels at which I’d reconsider: below $70, Model 3 (bull) offers +60% and even Model 2 approaches neutral; below $55, Model 2 gives a margin of safety above 20% and normalized FCFE yield climbs toward 4%; below $43 the stock would be undervalued by the median of both methods. The upward reconsideration trigger, without a price decline: pro-forma figures in the Form 10 (expected this month) showing a SpinCo with margin above 12% and a remaining Flex with margin above 6.5% — in that case Model 4 (EPV) would need to be rewritten from a 4.4% cycle margin to a 7% one, and the entire valuation range would shift up by ~60%.
Next check: 10/28/2026, Q2 FY2027 report. To watch, in order: (1) contract assets / quarterly revenue ratio, threshold 17%; (2) payables balance, threshold 8.5 bil; (3) quarterly free cash flow, threshold 150 mil; (4) whether buybacks resume — if they resume above $100, the file’s best capital-allocation argument disappears.
The figures in this report come from: FY2026 10-K (filed 05/20/2026), Q1 FY2027 10-Q (filed 07/31/2026), FY2022-FY2025 10-K, SEC XBRL companyfacts tags, the 09/11/2026 data pack and the Monte Carlo simulation mc-FLEX-20260911.json. Market figures (price, market cap, beta, insider ownership, consensus target) are from yfinance as of 09/11/2026. FY2027 guidance and the post-separation leadership plan are from the company’s Q1 FY2027 press release and earnings presentation. Gaps explicitly flagged: the research and development line (not presented separately by the company), quantification of memory price impact on revenue (undisclosed), numerical covenant thresholds (unquantified in the 10-Q), and management’s open-market purchases (the proxy isn’t part of the downloaded SEC file).
DEEPFCF;FLEX;FCFE (CFO - capex + cedari de imobilizari, TTM iun-2026);833;US GAAP - dobanda de 212 mil e raportata in operating, leasingurile financiare sunt declarate nemateriale, nu exista minoritari si nu se plateste dividend, deci divergenta fata de FCF simplu e 0,8% DEEPDCF;FLEX;-81;-61.6;4;5 DEEPDCFM;FLEX;1;DCF FCFE urs;-77 DEEPDCFM;FLEX;2;DCF FCFE baza;-60 DEEPDCFM;FLEX;3;DCF FCFE taur;4 DEEPDCFM;FLEX;4;EPV Greenwald;-81 DEEPDCFM;FLEX;5;Multipli istorici 5 ani;-62 DEEPMC;FLEX;-75.3;-69.1;-60.0;-48.5;-34.9;0.8;20000 DEEPIN;FLEX;900;0.09;0.105;0.025;0 EXPO;FLEX;capex-hyperscaleri;0.45;CPI ajunge la ~33% din venituri in FY2027 (crestere ghidata 65-75%) si un singur client face 12% din vanzarile totale, majoritar in CPI; toata prima de evaluare sta pe ciclul de investitii al hyperscalerilor, factor pe care eticheta de sector Electronic Components nu il separa de restul pozitiilor din plan