MCHP

We rate Microchip Avoid: at $75.85 we would not buy. Our central fair value is $47.59, 37.3% below the price. Avoid means fair value is under 0.85 times the price (the line between Avoid and Hold, $64.47); here it is 0.627 times. A buyer would also need the 12% yearly hurdle, which takes a price of $45.58 or lower, and the price is $30.27 above that. Only the bull case ($73.93) is near the price, 2.5% under it, and it earns +10.7% a year from here, under the hurdle. The central case earns +2.5% a year and the case-weighted value ($48.11) is 37% below the price. The view does not depend on our exit-multiple adjustment: on a GAAP-consistent basis the method's default 29.0x gives $59.61-$61.60, and even the unrestated 29.0x on pre-amortization EBIT gives $62.98 (section 3).

Terms. Fair value is our estimate of what a share is worth today. We value each company three ways. The central case is our best estimate, the bear case is a severe but plausible bad outcome, and the bull case is the matching good outcome; only the central case sets the call. A fiscal year ends in March, so "fiscal 2028" is the year to March 2028, and "Q1 FY2027" is the quarter to June 2026. EBIT is operating profit before interest and tax, and EV/EBIT divides enterprise value (the whole business, debt included) by it. Acquired amortization is the non-cash write-down of intangibles bought in past acquisitions. Non-GAAP figures are the company's adjusted ones, which leave out stock pay and special charges. Two estimates set the exit value in 2032. The fade assumes Microchip's growth and its return on new investment drift over 15 years to long-run levels (here 3.7% growth and the peers' 37.8% return on new capital). The outside view applies the EV/EBIT multiple that Microchip's own history and its peers' show. "The method" is Capfolio's standard valuation method, which model.py runs. The Street means the analysts who follow the stock. A flag is a warning from our automated report check that a number looks off or an argument is missing.

How blind the report is. We first value a company without seeing the price, the Street's estimates or its targets (blind), so the market can't steer the model. The model and assumptions were committed blind at 0a4b5f64e, with a central fair value of $49.95. The blind step was only partly blind: the filings, masked to hide market information, still showed the conversion prices of the three convertible notes, that the 2017 notes became convertible because the stock traded above 130% of their conversion price, a fair value of the 2017 notes that implies about $109 a share on 2026-06-30, and a preferred share count that implies about $62. An analyst on the Q1 call said the Street's December estimate was "up a little", and the analyst's recall from training places the stock somewhere around $50-100 in 2024-25 (assumptions.md lists each). None is a model input except the conversion prices, which set when convertible holders take shares. The market pass (the later step in which we read the price, estimates and targets: market.py, market.txt, odds.json) changed no model input.

The revision of 2026-10-21 is unblinded. It answers an adversarial review (an outside reviewer asked to find what is wrong) and was made with the price, the Street's estimates and the reviewer's findings in hand. It changes four central inputs and one bull input; section 4 gives each change, its dated evidence and its dollars, and the central moves from $49.95 to $47.59.

1. A normal year's revenue sets the value, not the speed of the recovery

Microchip sells catalog chips that customers design in once and buy for years: mixed-signal microcontrollers (small programmable chips that run machines; 50.0% of fiscal 2026), analog (28.2% of fiscal 2026) and "other" (21.8%, mostly FPGAs, which customers program after buying, plus licensing, memory and timing systems) (10-K FY2026). Sales go mostly through distributors, which is what made the last cycle so violent: revenue went from $8,439M in fiscal 2023 to $4,402M in fiscal 2025, a 48% fall, as distributors and customers worked off inventory bought in the shortage (market data). Gross margin followed how full the fabs ran, from a 67.5% GAAP peak to 56.1%, because Microchip runs its own fabs and carries their fixed cost.

The value question is therefore where a normal year sits, not how fast the recovery is. The fiscal 2023-26 mean revenue is $6,297M (two peak years and two trough years, centred on March 2024); the fiscal 2019-21 mean is $5,354M (centred September 2019). Between them revenue grew 3.7% a year (4.1% if the gap is taken as four years rather than 4.5; market data). Our central path takes the +13% restocking year of fiscal 2028, a +4% digestion year, and then that measured 4% a year, which gives fiscal 2032 revenue of $8,587M, 1.6% above the 4% trend line of $8,450M because fiscal 2028-29 sit above it. The path lands on management's long-term model (65% gross margin, 25% operating expenses, 40% non-GAAP operating margin; Q4 FY2026 call, Citi conference September 2026): 65.0% gross and 40.2% non-GAAP operating in fiscal 2032, or 34.8% after stock pay and special charges. That gives fiscal 2032 EBIT before acquired amortization of $2,991M and NOPAT of $2,587M at a 13.5% tax rate.

2. Microchip has beaten its own revenue guidance six quarters running

Each quarter Microchip guides to a revenue range, and it has beaten the midpoint in each of the last six quarters: +1.1%, +2.9%, +0.9%, +5.0%, +4.1% and +2.0%, median +2.5% (releases; market data quarterly). Tags in brackets say where a fact comes from: reported (in a filing or release), claim (management on a call) and inference (ours). The June 2026 quarter (Q1 FY2027) brought:

• Revenue $1,484.7M, up 38.0% on the year and 13.2% on the quarter, against a $1,456M midpoint (reported).
• Non-GAAP gross margin 63.8%, above the guided range, still after $38.5M of underutilization charges (reported; charges from the call).
• Non-GAAP EPS (earnings per share) $0.76 against a $0.67-0.71 guide (reported).
• Inventory days down from 185 to 175, against a 130-150 target; distribution inventory 25 days, near the low end of its 17-43 history (reported; history from the call).
• Data centre: about $591M in calendar 2025 and about $1B expected in calendar 2026, with revenue from PCIe Gen 6 (the next data-centre connection standard) switch and retimer chips only from calendar 2027 and Gen 6 design wins up from 6 to 12 programs (claim).

The September quarter guide is $1,589-1,618M with non-GAAP gross margin 66-67%. That margin includes one-off gains from an IP licensing sale and a write-up of channel prices after the August price increase (claim), so we carry 66.0% for the rest of fiscal 2027 and let it fall to 64.5% in fiscal 2029, the digestion year, and 65.0% after. Net debt is about $5.2B and leverage 2.85x; there are no buybacks and the dividend is held at $0.455 a quarter (reported).

3. Forecast and valuation: central fair value $47.59

The model is model.py, run on the method; it writes output.txt, forecast.json and valuation.json. The inputs and their sources are in assumptions.md (the blind record; section 4 and drivers.json carry the revised values); the drivers and their ranges are in drivers.json.

Fiscal 2027 is Q1 as reported plus Q2 at the guide midpoint beaten by the median +2.5% ($1,643M), Q3 at $1,659M (management says December will fall less than the usual 3-5%) and Q4 at $1,709M. Fiscal 2028 at +13% is +7.4% over the March 2027 run-rate. Fiscal 2029 at +4% is a digestion year, because the upturn is already in its seventh above-seasonal quarter.

Debt and other claims counted today. From enterprise value we take $5,398.9M of debt at face (commercial paper, senior notes and the convertible notes), $580.2M of long-term tax payable, $333.9M of customer deposits under supply agreements, 25% of the $480.2M Malaysian tax assessment, $159.4M of preferred dividends to the March 2028 conversion, and the $1,485M mandatory convertible preferred (preferred shares that must convert into common stock in March 2028). Its settlement has three tiers (10-Q note 9, applicable market value): above $62.48 each preferred converts into 16.006 shares (23.8M shares in all); from $51.00 to $62.48 into $1,000 of shares at that value, so a $1,485M claim; below $51.00 into a fixed 19.608 shares each (29.1M shares in all, worth less than the preference). The bull takes the 23.8M shares and the central and bear, both under $51, take the fixed 29.1M shares. Convertible notes add shares only in a case whose exit price is above their conversion price: only the bull's $116.87 exit clears the 2026 notes' $104.17, adding 0.93M shares. Not valued, named: the Hailo purchase (closing was due "later this month" in September; price undisclosed), the capped call on the preferred ($55.1M cost, cap price masked; it only reduces dilution), and pension and lease liabilities inside other long-term liabilities. The share count is today's 543.7M basic plus the 2017 notes; the Q1 FY2027 diluted count is 550.5M because restricted stock units are in it. Using it costs $0.56 a share, which would partly count stock pay twice (we charge it as a cost), so we keep 543.7M. Before the preferred correction of 2026-10-21 the central and bear counted a $1,485M claim; the fixed-share tier is worth +$0.05 to the blind central and +$1.48 to the bear.

Cost of capital. Beta (how strongly a stock moves with the market) of 1.403 is the Blume-adjusted (pulled toward 1) median of eight peers' market data betas (raw 1.601); Microchip's own Blume beta is 1.509, worth -$1.88 a share. LSCC is in the beta set but drops out of the exit multiple's peer set for too few valid years; the beta without it is 1.349 and would add $1.01. The cost of equity is 11.18%. The method gives debt no weight, so the WACC (weighted average cost of capital) is 11.18% too. That discounts cash flows that are already after interest at face (debt is counted today) at a cost of equity; a WACC with 10% or 15% debt weight, discounting all flows at 10.61% or 10.32%, would add $2.66 or $4.14. We did not adopt it because it changes the shared method, not this report (proposal in process.md).

The exit multiple: we restated the method's default 29.0x down to 18.9x. The method's default outside view is EV over GAAP operating income: 29.0x, the mean of Microchip's own ten-year median of 39.3x and the peers' 18.8x. Microchip's GAAP history carries $534M to $1,047M a year of Atmel and Microsemi amortization (10-K XBRL), which falls to about $231M by fiscal 2031, and our year-five EBIT is before it. So we restated every year of Microchip's and each peer's history by adding back its amortization of intangibles: Microchip's median falls to 19.6x, the peers' to 18.2x, and the mean to 18.9x, which we use. That is 0.65x the default, a bigger gap than the 1.5x at which our automated report check requires reviewers to challenge a change from the default; the restatement is the reason, and anchor_amortization.py reruns it. The reviewer agreed the restatement is right in principle.

The unrestated 29.0x applied to EBIT before amortization mixes two bases, so the honest counterfactual is the GAAP one: 29.0x on EBIT after the scheduled acquired amortization. The 10-K FY2026 schedule runs $454.8M in fiscal 2027, $344.0M, $270.2M, $258.1M and $231.3M in fiscal 2031; fiscal 2032 is our assumption of $100-231M. That is 28.1x or 26.8x on EBIT before amortization and gives fair value of $61.60 or $59.61, 0.81 or 0.79 of the price, still Avoid (model.py "R2"; market.py). On the blind inputs the same calculation gave $64.48 and $62.52 (the reviewer: $62.56-$64.42), the top of it a cent above the $64.47 Avoid line, a Hold by a cent on that version. The unrestated 29.0x on pre-amortization EBIT would give $62.98. On the revised model every version is under the line, so the restatement does not decide the view; it moves the central by $12-14.

Known limits of the restatement, from the review. (Finding 3) The XBRL add-back is total intangible amortization, which includes purchased software: the 10-K FY2026 income statement shows acquired amortization of $431.1M, $490.9M and $605.4M in fiscal 2026, 2025 and 2024 against totals of $534.3M, $584.3M and $689.1M, so software was $103.2M, $93.4M and $83.7M. Earlier years' split is not in the filings we saved. Taking the known split for fiscal 2024-26 and $83.7M or $103.2M earlier lifts Microchip's restated median, the multiple to 19.4x-19.5x and fair value by $0.73-$0.88. We did not change the headline because the peers' add-backs may carry the same mix and cannot be split from the XBRL facts. (Finding 4) STM has no 10-K facts and stays unrestated, and MPWR (3 of 12 years) and NXPI (8 of 12) have gaps left unrestated; the peers with 8 or more restated years alone give 18.7x and -$0.29.

The outside view is four-fifths higher than the fade ($98.70 against $54.70). The fade discounts at 11.18% with no debt in the WACC, and at that rate 12.0x year-six NOPAT is what a 3.7% long-run growth rate at a 37.8% return on new capital is worth. The history multiples were set by markets that priced semiconductor companies at lower blended costs of capital and, for Microchip, often at cyclical troughs when EBIT was depressed. We take the mean, as the method says, and show both: on the fade alone fair value is $35.24, on the outside view alone $59.94.

What moves value. Each change below is from the central $47.59.

The cases. The bear ($23.34, exit $45.18) is a second correction (a repeat of the inventory-driven downturn): flat fiscal 2028, -18% in fiscal 2029 with gross margin back to 58%, then a recovery to $7,037M of revenue by fiscal 2032 at 62% gross margin, with the mature return on new capital cut to 0.7x the peers'. It uses the same 18.9x exit multiple as the central; at 14x it would be $19.04 (-$4.30). The bull ($73.93, exit $116.87) is restocking plus the data-centre ramp: +20% in fiscal 2028, 7-8% a year after, and gross margin held at 66-67%, reaching $10,600M of revenue and a 44.9% non-GAAP operating margin, above management's own model.

4. Revision after the adversarial review: the central falls $2.36 to $47.59

The review (the reviewer's 15 numbered findings, answered in the table below) found the central and the Avoid view sound, put a central near $50 in a range of $46-$56, and asked for changes where an input did not match Microchip's own history. The rule for a change made after the market pass is dated evidence that does not depend on the price. The two findings whose changes qualify are 6 (revenue trend) and 8 (margin), both on Microchip's own filings and management's dated model. Findings 10, 11 and 12 are sensitivities, not inputs. The largest change is the revenue trend (-$4.67), which rests on market data statements and no claim or Street figure. The net move is -$2.36, away from the price; two of the four changes raise value. A separate correction, the preferred's fixed-share tier below $51, adds $0.05 to the central and $1.48 to the bear.

Dollar bridge, central, from the blind $49.95 (model.py "Revision after the adversarial review", output.txt; each step added to the one before):

The steps interact by $0.14 in all. The bull moves from $72.33 to $73.93 (+$1.60) because its operating cost growth falls from 3.0% to the central's 2.2% (a bull cannot carry a higher cost base than the central); the bear moves from $21.86 to $23.34 on the preferred alone. The case-weighted value moves from $48.52 to $48.11, the exit price from $85.74 to $76.70, the purchase price from $47.72 to $45.58 and the five-year return from +3.7% to +2.5% a year.

Finding 6, revenue trend, -$4.67. The blind central grew 6.5%, 6% and 6% in fiscal 2030-32 and called it a 5% trend. Microchip's own measured trend is 3.7% (4.1% over a four-year gap): fiscal 2019-21 mean $5,354M to fiscal 2023-26 mean $6,297M (market data statements, read 2026-10-21). The data-centre ramp (about $591M in calendar 2025 to about $1B in calendar 2026, a management claim, Q1 FY2027 call 2026-08-06) is already in the +13% of fiscal 2028; PCIe Gen 6 revenue from calendar 2027 and the August 2026 price increase are claims about a step, not a rate of growth, so they do not support a higher trend for a decade. We stepped down to 4% a year and moved the growth after fiscal 2032 (the start of the 15-year fade) from 5.5% to 4.0% so the fade does not start above the trend (-$1.18). Keeping the 5% would be worth $2.10 and 6% $4.29. The upside stays in the bull (7-8%) and in the top of the range (9%).

Finding 8, margin, +$1.05 and +$2.40. The blind central reached a 39.1% non-GAAP operating margin in fiscal 2032, under management's 40% model and far under fiscal 2022-24 (GAAP operating income plus amortization, before stock pay, was 43.9%, 47.9% and 45.0% of revenue in fiscal 2022-24; the company reported 43.9% non-GAAP for fiscal 2024, release of 2025-05). Two inputs now rest on dated evidence. Gross margin is 65.0% in fiscal 2030-32 (fiscal 2029 stays 64.5%): management's model is 65% (Q4 FY2026 call 2026-05-07, Citi conference September 2026, claims); fiscal 2024 was 65.8% non-GAAP; the June 2026 quarter was 63.8% after $38.5M (2.6 points) of underutilization charges; and non-GAAP gross margin runs 0.4-0.9 point above GAAP (65.8% against 65.4%, 57.0% against 56.1%, 58.5% against 57.7% in fiscal 2024-26), so the blind 64.3% had been placed against a GAAP mean of 62.2%, a mixed basis. Operating cost now grows 2.2% a year plus a quarter of revenue growth, not 3.5%: research plus selling and administrative cost grew 2.2% a year from fiscal 2019 ($1,509M) to fiscal 2026 ($1,760M) while revenue fell from $5,350M to $4,713M, and non-GAAP operating cost was $1,676M in fiscal 2024, $1,432M in fiscal 2025 and about $1,706M on the Q1 FY2027 run-rate (market data; releases). The result is a 40.2% non-GAAP operating margin in fiscal 2032 (opex 24.8% of revenue against management's 25%).

Evidence against, disclosed. After stock pay and special charges the central margin is 34.8%. Microchip's pooled history (GAAP operating income plus 10-K intangible amortization, revenue-weighted) is 35.2% over fiscal 2019-26, 36.3% over fiscal 2020-26 and 38.5% over fiscal 2019-24; the reviewer quoted 35.2% and 36.9%, and we reproduce the first only. That history includes about $84-103M a year of purchased-software amortization in the add-back, which overstates it by roughly 1-2 points in recent years. Moving the fiscal 2032 margin to those pooled values is worth +$0.57, +$2.37 and +$6.17. Stock pay is 5% of revenue (5.4% in fiscal 2026, 5.1% in Q1 FY2027) against 3.2% pooled over fiscal 2019-26 and 2.0-3.8% in fiscal 2022-24; at 4.4%, 4.0% and 3.5% (finding 9) value would be +$1.01, +$1.68 and +$2.52. We kept 5% because it is the current reported rate and the history is at peak revenue.

Answers to each finding (the reviewer's dollars were reproduced where noted; the saved working is in output.txt under "Review checks" and market.txt):

5. Microchip's timing business shows almost nothing about SiTime

Short answer: almost nothing quantitative. SiTime is a timing-chip maker we also cover (report). Microchip is SiTime's best documented US-listed competitor, but it does not report timing as a line, so it cannot confirm or refute SiTime's share, pricing or MEMS (silicon-based oscillator) claims. Evidence is tagged as in section 2: reported, claim and inference.

• Reported. Timing products sit inside the analog line (28.2% of fiscal 2026 revenue, $1,329M) and "timing systems" inside "other" (21.8%, $1,029M) (10-K FY2026). Research and development covers timing systems. The 10-K does not use the words MEMS or oscillator, gives no timing revenue, and gives no product-line margins.
• Claim. Timing and synchronization is one of five areas Microchip sells into data centres outside its Data Center Solutions unit; that catalog data-centre revenue was about $288M in calendar 2025, and timing is named among eight or so product families inside it (Q1 FY2027 call). The CEO called timing hard to break out because business units mix analog and microcontroller parts (Q4 FY2026 call). An analyst on the Q1 call listed timing among Microchip's leading defence positions.
• Inference: size. If timing were material, a company that breaks out data centre, end markets and the Data Center Solutions unit would have shown it. Timing is probably a few percent of Microchip's revenue, and Microchip's valuation turns on microcontrollers and the cycle, as our timing-sector note warned (sector note, section 6; linked in section 10).
• Inference: MEMS against quartz. Microchip sells both: MEMS oscillators (the DSC families) and quartz OCXOs (oven-controlled quartz oscillators) and time systems from Vectron and Symmetricom (sector note, section 4). It is hedged on the technology rather than betting on MEMS, which fits SiTime's claim that no rival sells MEMS oscillators at scale into AI optics, but does not prove it. Our memory, not checked in a source this run, is that the DSC line came from Discera through Micrel.
• Inference: pricing. Microchip's August 2026 price increase and channel price write-up are company-wide; nothing ties them to timing. Microchip's long-term 65% gross margin target for the whole company is about SiTime's own level, so a broad-line seller can price timing parts near SiTime's margins without that showing in its reported numbers.
• Not found. Microchip's timing revenue, its share of any timing segment, its timing prices, and its MEMS revenue. A web search found no disclosed figure.

6. The other side: why the value could be too low, and too high

Why the value could be too low. The restated 18.9x exit is our judgment on top of the method's default, and it cuts $15.39 a share against the unrestated 29.0x (the GAAP-consistent version cuts $12-14). Markets have paid Microchip far more than 19x pre-amortization EBIT in good years, and our own pre-amortization return on capital median is 65%, well above the 37.8% peer return we fade toward. Management has hit or beaten its guide six quarters running, says bookings are the strongest in about four years, and expects data-centre revenue to grow sharply from calendar 2027. The 4% trend is Microchip's measured history; a data-centre and defence mix that lasts would justify 5-6%, worth $2.10-$4.29. A bull who holds 66-67% gross margin and 7% growth gets $74; one who also takes the unrestated 29.0x default gets well above that.

Why the value could be too high. The upturn is seven quarters old, and the last one ended in a 48% revenue fall. Inventory days are still 175 against a 130-150 target, so Microchip is still running fabs below capacity while selling from stock; a demand stall would hit gross margin twice. The balance sheet carries $5.4B of debt, a $1.5B preferred and up to $480M of Malaysian tax, all counted ahead of common holders. The pessimistic case is worth $23.34. The central margin of 40.2% non-GAAP is management's own model, which a stock pay rate of 5% of revenue (against 3.2% pooled history) and the pooled 35.2%-38.5% margin history after stock pay make neither easy nor unreasonable.

7. Next results preview: Q2 FY2027, early November 2026

Q2 FY2027 (September 2026 quarter). Date not confirmed; last year's Q2 call was on 2025-11-06, so early November 2026 is likely.

Each $100M of quarterly revenue above ours that lasts into fiscal 2028 (about 5 points of fiscal 2028 growth) is worth about $4.3 a share (model.fair_value with fiscal 2028 growth at 18.4%).

8. What settles it

• Q2 FY2027 results, early November 2026. Falsifies the central case if revenue lands below the $1.589B guide low or the Q3 guide midpoint is below $1.55B; supports the bull if non-GAAP gross margin holds above 66% without the one-offs.
• Inventory and channel, each quarter. Distribution inventory above 35 days with falling sell-through, or inventory days rising again, is the bear's first sign.
• Data-centre ramp, calendar 2027. PCIe Gen 6 revenue arriving in calendar 2027 is management's claim; a slip of a year moves fiscal 2028 growth toward the bear.
• Malaysian tax ruling, by about late 2027. The 10-K says the matter could be decided within 18 months; losing in full costs $0.63 a share over our 25%.
• Preferred conversion, March 2028. It settles the share count (10-Q note 9): above $62.48 the holders take 23.8M shares, from $51 to $62.48 shares worth the $1,485M preference, below $51 a fixed 29.1M shares.

9. The price needs far more than our model gives, and the Street is far higher

All figures from market.txt and odds.json unless named. Our automated report check raised several flags on this report: warnings that a number looks off or an argument is missing. Each is answered below.

The stock closed at $75.85 on 2026-10-20, up 1.9% on the day, 28% under its 52-week high of $105.91 and 56% over the low of $48.52 (market data quote). The vendor's official close is the same $75.85 (sip-list-exchange-close). The price sits 2.6% above the bull case ($73.93) and 59% above our central. On Microchip's own market data beta (1.759, Blume 1.509, cost of equity 11.63%) fair value is $45.71 and the purchase price $44.82; the view does not change.

What the price needs (market.py, everything else central):

• A higher exit. An outside view of 37.9x EV over EBIT before acquired amortization on central fiscal 2032 EBIT of $2,991M, which is an exit price near $127 as the mean of the fade ($55) and the outside view ($199); put another way an exit price of about $127 on 2032-03-31 against our $76.70. Our 18.9x is the restated peer and own history; the unrestated GAAP default is 29.0x, which gives $62.98, still $13 under the price, and the GAAP-consistent version gives $59.61-$61.60. The price therefore needs more than the method's own default multiple.
• A bigger business at our multiple. Growth every year of fiscal 2028-32 6.8 points above central: fiscal 2032 revenue $11.7B (central $8.6B) and EBIT $4.7B (1.57x), margins held. That is above our bull ($10.6B revenue, 44.9% non-GAAP operating margin), which is why even the bull is below the price.
• No single driver does it (alpha research odds): none reaches the price within its range. The best ends give $72 (fiscal 2028 growth 45%), $74 (fiscal 2030-32 growth 15%), $72 (exit multiple 35x), $59 (gross margin 72%), $56 (growth after 2032 of 12%).
• The odds in our scenarios are under 1% (1 of 3,000 draws; odds.json range 0.0% to 3.9%; a mixture of two simulated worlds, the base world at 75% and the bust world at 25%). Simulated median $42, 10th percentile $29, 90th percentile $55. P is our distributions, not a calibrated probability, and the distributions centre on our view, so P measures how far the price sits outside our ranges, not the chance the market is wrong.

The simulated median is 11% under the central ($42 against $47.59; odds.json flags a gap over 15%, so this one is not flagged). Inference: the bust world, 25% of draws with a median near $30, pulls the mixture down; the base world alone has a median of $46, 3% under the central.

Why we may be wrong. The case for the price is the one in section 6: a higher multiple for a recovering cyclical with a 65% median pre-amortization return on capital, and a data-centre ramp from calendar 2027. We do not widen the ranges toward the price.

Why we are far below the Street. Our report check flags any value under 0.6x the Street's target discounted one year. The average target of $110.50, discounted one year at 11.18%, is $99.39; our $47.59 is 0.48x it, under the 0.6x band. The Street's revenue is close to ours (fiscal 2027 $6,404M against $6,496M, fiscal 2028 $7,460M against $7,340M, fiscal 2029 $8,229M against $7,634M, 9 analysts for 2029), so the gap is not revenue. The vendor's own EBIT estimates (fiscal 2028 $4,334M, a 58.1% margin) have no stated basis and cannot be a company operating margin: management guides Q2 FY2027 non-GAAP operating margin of 38.5-39.5% and its long-term model is 40%. We instead worked the Street's non-GAAP operating margin from its EPS: EPS times 575M non-GAAP diluted shares (the Q2 guide's 574.5-575.5M), divided by one less our tax rate (7.5%, 11%, 13.5%), plus $196M of other expense (four times the $48.5-49.5M Q2 guide), over Street revenue. That gives 38.5% in fiscal 2027 (16 analysts on EPS), 42.3% in fiscal 2028 (16) and 44.6% in fiscal 2029 (5), against our 38.1%, 40.0% and 39.2% on the same before-stock-pay basis (market.py).

Running the Street's path through our model, one sensitivity at a time, each added to the one before (none is an input; market.py):

The target discounted one year is $99.39 at our 11.18% and $100.92 at beta 1.0. Revenue is $6.50 of the gap, the two margins $8.76, the multiple $17.65 and the discount rate $11.84. Even with every step, the model gives $92.33, 0.91x the target discounted at beta 1.0; the remaining $8.59 is what the Street's three years do not show, mostly growth and margin after fiscal 2029 (inference). The margin steps lower opex in the year named and let it grow from there on our cost rule, so a Street margin that lasted to fiscal 2032 would add more. On the blind inputs the same steps give $57.08, $61.69, $79.77, $92.77 and $98.91, where the reviewer's reconstruction of the margin steps reached $62.81, $81.11, $94.31 and $103.63; the first step matches exactly and the margin steps depend on how a year's margin is carried forward, which we could not recover. This replaces the earlier statement that revenue explains $7 of the gap and the rest is the margin and the multiple: the multiple and the discount rate carry most of it.

What Capfolio's ranking model says (read 2026-10-21, alpha rank MCHP). Alpha, our stock-ranking model, ranks MCHP #97 with a core score of 86.5, in the top 20%, with a data warning (net_income_source_review). That is a rank against other stocks, not a fair value, so it does not by itself conflict with an Avoid for new money at this price. We did not run a chart (technical) analysis.

The world in which $75.85 is right. Our report check asks for this whenever the odds are under 1%. We built it as hard as we can from the cases: the upturn does not stop at seven quarters, fiscal 2028 grows 20% and 7-8% after, gross margin holds the guide's 66-67%, and the market pays a multiple above our 18.9x (up to the unrestated 29.0x). That is our bull with a higher multiple; at the bull's own $116.87 exit it earns +10.7% a year from this price, not the 12% hurdle. We keep the central: the bookings and the data-centre ramp are management claims, the last upturn ended in a 48% revenue fall, and the multiple is the method's own history restated for amortization, not our taste.

The price sits outside our 10th-90th percentile range of $29-$55 (the range holding the middle 80% of simulated fair values). The drivers' ranges come from the history and peers each driver's source cites (drivers.json), not from the price; the price is reached by no single driver, and only by several at their ends together (a 35x exit multiple or fiscal 2028 growth of 45% each give $72). We do not widen toward the price.

Street EPS. Street EPS is $3.65 (fiscal 2027), $4.59 (2028) and $5.23 (2029). Our model is run to EBIT and free cash, not to non-GAAP EPS, so we did not compare EPS by year beyond the margin worked above; the Q2 FY2027 preview carries our $0.96 against a Street $0.935 (earnings calendar).

The move from the blind value. $49.95 to $47.59, -$2.36 (-4.7%). The largest change is the revenue trend, -$4.67, on Microchip's own market data revenue history; none of the four changes rests on a management claim or a Street figure alone, and the gross-margin target of 65% is a management claim backed by the fiscal 2024 actual.

The gap between the fade and the outside view, and the 18.9x multiple. The outside view ($98.70) against the fade ($54.70) is one input, the 18.9x exit multiple, and the fade's 3.7% implied growth (section 3). The 0.65x gap to the method's default 29.0x is the amortization restatement in section 3, reproducible with anchor_amortization.py. On a GAAP-consistent 29.0x fair value is $59.61-$61.60, 0.79-0.81 of the price, and on the unrestated 29.0x on pre-amortization EBIT it is $62.98, 0.83 of the price: all under the 0.85 line, so the view is Avoid on each version and the restatement does not decide it.

10. Sources

All read 2026-10-20 or 2026-10-21 UTC, pinned in data-manifest.json.

• Market data statement endpoints only (income, balance sheet, cash flow, annual and quarterly; product and geographic revenue segmentation) for MCHP and the peers TXN, ADI, NXPI, ON, STM, SWKS, MPWR, LSCC; market data profile beta field only for the betas.
• SEC EDGAR: 10-K for fiscal 2026 (year to 2026-03-31); 10-Qs for Q3 FY2026 and Q1 FY2027; 8-Ks of 2026-05-28, 2026-07-23 and 2026-08-18; earnings releases Q3 FY2025 through Q1 FY2027 and 2026-01-05; companyfacts for the amortization add-back. Masked to hide market information with scripts/research-mask.py. The revision uses the 10-K FY2026 acquired-amortization lines and schedule, and the non-GAAP history in the fiscal 2025 and 2026 fourth-quarter releases.
• Earnings call transcripts: Q4 FY2026 (2026-05-07) and Q1 FY2027 (2026-08-06), Quartr through market data.
• Citi conference, about 2026-09-09: read through market data for the margin target and the upturn length; not saved, so its claims rest on this report's quotations of it.
• Market pass and revision: market data quote, price targets and analyst estimates read 2026-10-21T02:18Z (market.py); official close.
• Sector note: docs/research/sectors/timing/note.md; SiTime report: docs/research/2026-10-20-sitm/.

MCHP — Microchip Technology, Inc. — research view: Avoid (we would not buy at $75.85)
As of 2026-10-20 · partly unblinded; revised 2026-10-21, unblinded. What "blind" means
is explained below the box.

Fair value today      $47.59 per diluted share, discounting at an 11.18% cost of equity (the
                      yearly return investors require); blind value $49.95
Price                 $75.85, market data quote, 2026-10-20T20:00:01Z
                      (regular-session close; official close $75.85, previous
                      close $74.43). Fair value is 37.3% below the price.
Five-year exit price  $76.70 on 2032-03-31: the mean of the long fade $54.70 and the outside
                      view $98.70 (18.9x EV over EBIT before acquired amortization). At
                      10/20-year fades: $76.18/$77.11. A value-driver formula gives 12.5x
                      year-six NOPAT (after-tax operating profit) against the fade's 12.0x.
Five-year return      +2.5% a year from today's price on our central case, including the
                      modeled free-cash payouts and today's claims on the business
                      (method.five_year_return, market.py); bear -8.1%, bull +10.7%
Purchase price        $45.58 to earn the 12% hurdle (the yearly return we require)
Bear / bull today     $23.34 / $73.93, each a single consistent business outcome
                      (case-weighted 25/50/25, our weights: $48.11)
Confidence            30% (subjective, not calibrated): the chance the 2032-03-31 value
                      lands within ±20% of the central exit price
Moat                  Narrow: catalog microcontroller, analog and FPGA parts designed into
                      industrial, defence and car programs that run for a decade, mostly
                      made in its own fabs; held back by a channel that overshipped in
                      fiscal 2022-24 and a price-sensitive Chinese market
Market expects        Far more than our central case: an outside view of 37.9x EV/EBIT on
                      fiscal 2032 EBIT before acquired amortization (our 18.9x; the
                      unrestated GAAP default is 29.0x), an exit price near $127 against
                      $76.70; or, at our 18.9x, growth 6.8 points a year above central
                      in fiscal 2028-32, fiscal 2032 EBIT of $4.7B (1.57x central)
                      (market.py)
Price needs           No single driver reaches the price within its range (`alpha
                      research odds`); the best ends are $72 on fiscal 2028 growth of
                      45%, $74 on fiscal 2030-32 growth of 15%, $72 on a 35x exit
                      multiple and $59 on a 72% gross margin
Odds in our scenarios Under 1% that fair value reaches the price (P(price justified),
                      1 of 3,000 simulated scenarios, odds.json; subjective, not
                      calibrated)
Key driver            Fiscal 2028 revenue growth: +3% / +22% instead of +13% is -$7.54 /
                      +$6.94 a share. The exit multiple moves value more in absolute terms:
                      each turn (1x) is about $1.5, and the unrestated 29.0x GAAP default
                      would add $15.39
Street                Consensus target $110.50, median $114.50, range $88-$135 (market data,
                      read 2026-10-21T02:18Z); 3 targets in the last quarter averaging
                      $103.33, 40 in the last year at $93.20, none in the last month.
                      Targets forecast a future price, so they show expectations, not
                      validation or value today. Fiscal 2027 revenue estimate $6,404M
                      against our $6,496M