A 90-year connector compounder that the market has re-rated into an AI-infrastructure growth stock — the business is exceptional and the AI cycle is real, but at ~34x forward EPS and 50% above its own decade-average multiple, the price already pays Amphenol to keep printing 30%+ organic growth, leaving no cushion if hyperscaler capex normalizes.
| Date |
|---|
| Type |
|---|
| What happened |
|---|
| Source |
|---|
| 2026-08-10 | editorial note | Valuation figure revised: $161 → $168.97Valuation moved from $161 (deep-dive-2026-06-20.md) to $168.97 (deep-dive-2026-08-10.md). | dossier |
| 2026-08-10 | editorial note | Capex figure revised: $291.6M → $355.5MCapex moved from $291.6M (deep-dive-2026-06-20.md) to $355.5M (deep-dive-2026-08-10.md). | dossier |
| 2026-08-10 | editorial note | Margin figure revised: 25.4% → 40.5%Margin moved from 25.4% (deep-dive-2026-06-20.md) to 40.5% (deep-dive-2026-08-10.md). | dossier |
| 2026-08-10 | editorial note | Revenue figure revised: 91% → 12.7%Revenue moved from 91% (deep-dive-2026-06-20.md) to 12.7% (deep-dive-2026-08-10.md). | dossier |
| 2026-08-10 | editorial note | Verdict changed: A 90-year connector compounder that the market has re-rated into an AI-infrastructure growth stock — the business is exceptional and the AI cycle is real, but at ~34x forward EPS and…Before (deep-dive-2026-06-20.md): A 90-year connector compounder that the market has re-rated into an AI-infrastructure growth stock — the business is exceptional and the AI cycle is real, but at ~34x forward EPS and 50% above its own decade-average multiple, the price already pays Amphenol to keep printing 30%+ organic growth, leaving no cushion if hyperscaler capex normalizes. After (deep-dive-2026-08-10.md): The Q2 print did the one thing the bear case could not survive — it grew earnings faster than the stock, compressing the forward multiple from ~34x to ~32x while the price rose 5%; the valuation objection is now materially weaker, but a 1.23 book-to-bill built on a single demand wave, an adjusted tax rate stepped up to 27%, cash conversion down to 68%, and an $80.8M CEO sale four days before the split announcement are the four things that decide whether this is a compounder or a peak. | dossier |
The verdict
The Q2 print did the one thing the bear case could not survive — it grew earnings faster than the stock, compressing the forward multiple from ~34x to ~32x while the price rose 5%; the valuation objection is now materially weaker, but a 1.23 book-to-bill built on a single demand wave, an adjusted tax rate stepped up to 27%, cash conversion down to 68%, and an $80.8M CEO sale four days before the split announcement are the four things that decide whether this is a compounder or a peak.
Primary sources
SEC filings
Earnings calls
Source documents — open to read in full
What changed — six material moves
What held — does the structural thesis stand?
Yes, and it strengthened on the operating side. Every load-bearing claim in the June dossier survived contact with the Q2 print: the design-in interconnect model, the decentralised GM operating system, the acquisition flywheel, and the AI-datacom engine. Two of the four bear legs are now weaker:
Two bear legs are unchanged or worse: AI-cycle concentration (Communications Solutions is now 61.5% of sales, IT datacom 43%) and the goodwill tower ($22.8B of goodwill+intangibles, 51% of assets). And one new one appeared: the adjusted effective tax rate was stepped up from 24.5% to 27.0% on a "continued shift in income to higher-tax jurisdictions" — a permanent ~3% haircut to the EPS path.
Direction stays NEUTRAL, but the reason changed and the buyable zone must move up. The June seed named a $120–135 accumulation zone off a 25x-on-FY2027 de-rate; on the new earnings base that same arithmetic gives ~$141–160 (22–25x on an FY2027 base of $6.40) — the prior zone is stale and would now require a thesis break, not a de-rate, to reach. Conviction on the valuation cap weakens; conviction on the cycle risk is unchanged. No ACTIVE MarketCall tracks this name, so nothing routes to our position log as an update — this is a fresh-seed refresh only. own_view: deferred — this run is unattended and does not carry Connor's prior.
Lens 1 · Company Overview — unchanged, see the previous dossier. (Segment definitions in the Q2 Form 10-QThe quarterly version of the annual report. Lighter, and not audited. Note 13 are verbatim identical to the FY2025 Form 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes.; nothing in the new filings restates the business.)
Lens 2 · Supply Chain — unchanged, see the previous dossier. One transcript data point worth carrying forward without re-running the lens: asked directly about fiber supply bottlenecks, Norwitt said the company does "not see any significant bottleneck" and deflected to decentralised GM management. That is an assertion, not a disclosure, and it is the single most checkable claim on the call.
Lens 3 · Competitive Advantages — unchanged, see the previous dossier. No new competitor named in the Q2 filing; Risk Factors are explicitly unchanged from the FY2025 10-K.
Q2-2026 external net sales / segment operating income / margin, with prior-year comparison:
| Segment | Q2-26 rev ($M) | % of total | YoY USD | YoY organic | Seg OI ($M) | Seg margin | Q2-25 margin |
|---|---|---|---|---|---|---|---|
| Communications Solutions | 5,383.6 | 61.5% | +85% | +42% | 1,808.3 | 33.6% | 30.6% |
| Harsh Environment Solutions | 1,856.8 | 21.2% | +28% | +22% | 559.3 | 30.1% | 25.2% |
| Interconnect & Sensor Systems | 1,517.7 | 17.3% | +17% | +13% | 318.9 | 21.0% | 19.5% |
| Total | 8,758.1 | 100% | +55% | +30% | 2,686.5 | 30.7% | 26.7% |
Three things this table says that the June table did not.
First, margin expanded in every segment simultaneously — +300bps at Communications, +490bps at Harsh Environment, +150bps at Interconnect & Sensor. MD&A attributes it to "strong operating leverage on the significant growth" plus the $80.0M IEEPA recovery, partially offset by acquisitions running below company average. The offset is real: CommScope's H1 net margin was 9.1% ($190.2M on $2,100.9M) against a company 16.5% — so the underlying organic business expanded margin by more than the reported number, which is the genuinely impressive part.
Second, Harsh Environment's +490bps is the quiet story. The June dossier framed the company as an AI-datacom story wearing a diversified suit. A defence/aero/industrial segment putting up +22% organic and 30.1% margins is not a passenger — it is running at a margin the whole company did not reach a year ago.
Third, the concentration got worse anyway. Communications Solutions is 61.5% of sales (from 52% in FY2025), and IT datacom alone is 43% of Q2 sales vs 36% in FY2025. Management now expects CommScope's IT datacom exposure to reach ~50% of its revenue for the full year, from ~33% in 2025 — the acquisition is increasing, not diluting, the AI beta.
End-market organic growth, Q2-2026 YoY: IT datacom +63% (total +89%) · Defense +24% · Commercial aerospace +21% · Industrial +18% · Mobile devices +14% · Automotive +6% · Communications networks −6%.
The last line is the only negative print in the company and deserves its weight: comms networks organic revenue declined, on "demand moderations from both communications network operators and wireless equipment manufacturers," and management guided it to fall a further mid-teens sequentially in Q3. That is a real end market rolling over inside a company at record everything.
Geography, Q2-2026 (ship-to): US $3,413.8M (39.0%), China $1,113.0M (12.7%), other foreign $4,231.3M (48.3%). US grew +76% USD but only +23% organic — 52pp of US growth was acquisition. Foreign grew +33% organic. The June dossier's flagged conflict ("China ~30% of sales" vs 16% ship-to ) resolves further in the research layer's favour: ship-to China is now 12.7% and falling. The manufacturing-side China exposure is not disclosed quarterly and remains the unmeasured risk.
All figures `` unless noted.
The print.
Consensus. Adjusted EPS $1.35 vs consensus $1.17–1.19 — a +13% to +15% surprise; revenue $8,758.1M vs ~$8,322M consensus, a +$437M beat. This is the second consecutive quarter of beating the high end of the company's own guide — the guide is now systematically conservative, which matters for how you read the Q3 number.
Balance-sheet flags (Q2-2026 vs Dec-2025).
Market reaction. The stock rose +4.15% on 2026-07-29. Since then Citi went to $210 (from $195), UBS to $197, Barclays to $200; Consensus price targetThe average of what published analysts think the share price should be. An opinion poll, not a forecast. $198.06 across 17 analysts, range $170–$230, rating Strong Buy. Price on 2026-08-10: $168.97, Market capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. $208.34B, 52-week range $104.71–$178.52.
Unusual vs its own history. A 90-year connector company printing 40.5% gross margin and a 1.23 book-to-bill is out of band on both dimensions at once. The June dossier said 58% growth was only defensible because a third of it was organic; the Q2 answer is the same test passed again — 30 of the 55 points are organic, and the margin came from operating leverage rather than price (management explicitly attributed CommScope's improvement to "operational execution in the factory" and SG&A discipline, "not pricing").
The June dossier ran this lens on `` paraphrase because no transcripts were on disk. The Q2-2026 transcript is now ingested, so this is the first primary read.
Sentiment shift vs the prior three-to-four calls: more confident, and now with a named number attached. The recurring 2025–early-2026 phrase was "virtually all sequential growth driven by AI." On the Q2 call that became a quantity: AI-related revenue at an annualised run rate of $10.5–11B — management has moved from describing AI as a driver to sizing it as a business. Norwitt framed durability structurally rather than cyclically: customers "need more of everything… more high-speed copper, more fiber optics, and more power," across training, inference and geographies simultaneously.
What is new in the language.
What they still do not say. There is no hedging on AI-capex durability, no discussion of hyperscaler customer concentration inside the IT datacom number, and no framing of any part of the 1.23 book-to-bill as pull-forward. Analyst pushback was thin: William Stein probed whether a backplane-to-PCB architecture shift could cut Amphenol content (Norwitt reframed to portfolio-agnostic "more overall content"); Andrew Buscaglia probed fiber supply bottlenecks (deflected). The June dossier's watch-item — the absence of hedging in a clearly cyclical demand wave — is unchanged and now louder, because the numbers being defended are twice as large.
| Metric | June 2026 dossier | 2026-08-10 | Direction |
|---|---|---|---|
| Price | ~$161 | $168.97 | +5.0% |
| Market cap | ~$157B | $208.34B | +33% (price + reporting basis) |
| Trailing P/E | 44.0 | 42.32 | down |
| Forward P/E | 34.1 | 31.76 on FY2026 cons. $5.32 | down ~7% |
| Consensus PT | $182 (18 analysts) | $198.06 (17 analysts, range $170–$230) | +9% |
Conflict flagged, not resolved: stockanalysis's overview page shows a forward P/E of 28.76 while its own forecast page shows 31.76 against FY2026 EPS of $5.32 at the same price. The gap is almost certainly NTM-blended (rolling four quarters, which reaches into a higher 2027) versus fiscal-year. Both are cited as-is; the FY2026 basis (31.76x) is the one used in Lens 11 because it is the one whose denominator is disclosed.
The read has changed. In June the argument was "you are paying a tech multiple for a connector company with no cushion." Today the price is 5% higher and the forward multiple is ~7% lower, because consensus FY2026 EPS moved from roughly $4.75 (this dossier's June base) to $5.32. Earnings outran price. The premium to TE Connectivity is unchanged in kind but the absolute stretch is smaller. Peer multiples: n/a — not re-sourced this run (TEL, APTV, ST, GLW were n/a in June and no new figures were pulled; do not read the June TEL ~19.9 trailing figure as current).
Post-split note. After 2026-09-02, every price above halves and every EPS figure halves. A $168.97 pre-split price is $84.49 post-split; FY2026 consensus $5.32 becomes $2.66. Multiples are unaffected. This dossier is entirely pre-split.
Moves and drivers since the 2026-06-20 boundary:
The tell is unchanged and sharper. This name trades on AI-capex sentiment and on the organic IT datacom number, not on the diversified-industrial fundamentals. The new wrinkle: a 15% beat now produces a 4% move. The bar has risen.
Tenure, track record and archetype are unchanged from the June dossier. Two things moved.
1. The insider selling got worse, and the timing got worse. Norwitt sold 500,000 shares for ~$80,785,000 on 2026-07-31 — two days after the record print, five days before the board approved the split — plus a Form 144 notice for a further ~$30.2M, and 186,104 shares filed through Morgan Stanley Executive Financial Services on or after 2026-07-31 in connection with option exercises. Reporting services put insider sales at ~$419M over trailing twelve months with the CEO and CFO leading. This sits on top of the June dossier's record (608,333 shares / ~$89.6M on 2026-02-13; Lampo 100,000 / ~$15.0M on 2026-02-18).
Read it honestly in both directions. Most of this is option-exercise-and-sell mechanics on a plan — the equity statement shows 7.7M shares issued on option exercises in H1-2026 for $222.7M of proceeds against 2.8M shares repurchased for $386.0M, i.e. the buyback is not shrinking the count, it is mopping up DilutionIssuing new shares, so each existing share owns a smaller slice of the same company., and the sales are the other side of that machine. But the scale is now the largest in the company's history, it is concentrated in the two people who saw the Q2 numbers first, and the July sale landed inside the window between the print and the split vote. It is a yellow flag, not a red one, and it is the bears' best single fact.
2. A quantified governance signal. At the 2026-05-21 AGM Norwitt received 58.06M votes against — 5.4% of votes cast, roughly four times the median dissent across the other seven nominees (Falck was higher at 4.8%; Lane was lowest at 0.5%). Deloitte's ratification drew 72.15M against (6.4%) — unusually high for an auditor vote. Neither is remotely close to failing. Both are the first measurable expression of the CEO+Chair concentration the June dossier flagged.
Capital allocation, updated. Dividend held at $0.25/quarter ($615.3M declared in H1-2026); buyback authorisation down to $411.1M remaining as of 2026-07-29 under the $2.0B 2024 program expiring 2027-04-28. Three acquisitions closed in H1-2026 for $10,684.0M net of cash, overwhelmingly CommScope. Deleveraging is now visibly prioritised over new deals — $1,134.1M of term-loan repayment between May and July.
Still clean.
regulatory/regulatory-findings.md re-fetched 2026-08-10 across 2021-08-10 → 2026-08-10: zero SEC Litigation Releases, zero AAERs. Q2 Note 15 repeats the ordinary-course language and the explicit "does not believe that the resolution of any existing legal or regulatory action is expected to have a material adverse effect". Risk Factors explicitly unchanged from the FY2025 10-K [Item 1A].Four things that got worse or are new.
One audit-scope gap worth naming. Management's assessment of internal control over financial reporting does not yet include CommScope, which was ~5% of total assets (ex-goodwill/intangibles) and 14% of Q2 consolidated net sales. This is a standard, permitted exclusion for a business acquired in-year, not a deficiency. It does mean the segment carrying the year's margin surprise is the segment whose controls have not yet been tested — worth remembering when reading the 40.5% gross margin.
Goodwill, restated. $17,554.7M goodwill + $5,288.9M intangibles = $22,843.6M, of which CommScope contributed $6,958.0M and $3,289.0M respectively; the entire CommScope goodwill was assigned to Communications Solutions and none is tax-deductible. Purchase accounting is still preliminary as of 2026-06-30 — the allocation can move. Acquired intangibles carry a 16-year weighted average life. If AI datacom turns, the impairment risk is concentrated in exactly one segment, and it is the segment that is 61.5% of sales.
Model integrity — read this before any number below. model.xlsx was rebuilt this run from the newly populated CSVs: 8 quarters of history (2024-Q3 → 2026-Q2), opening balance sheet "sourced from our figures" (not hand-typed), 3 years × 3 cases. But our model reports "No computed values — the workbook has 51 formulas and no cached results," and the share price and CAPM inputs on the Assumptions sheet are unset. The workbook therefore produces no valuation output, no balance check, and no per-share value in this unattended run — it must be opened in Excel and saved before any of it can be cited. Nothing below is taken from the model. Every figure is `` with the arithmetic shown, cross-checked against consensus.
Anchors. FY2025 adjusted diluted EPS $3.34. H1-2026 adjusted diluted EPS $2.42, of which Q2 $1.35 and Q1 therefore $1.07. Q3-2026 guide $1.40–1.42 (mid $1.41). Adjusted effective tax rate now 27.0%.
Valuation cross-check at $168.97:
| Basis | EPS | Implied multiple |
|---|---|---|
| FY2026 base | $5.33 | 31.7x |
| FY2027 base | $6.40 | 26.4x |
| FY2028 base | $7.23 | 23.4x |
| LTM adjusted | ~$4.32 | 39.1x |
The de-rate arithmetic, updated. FY2027 bear $5.45 at a de-rated 22x → ~$120 (−29%). FY2027 base $6.40 at 22–25x → $141–160. FY2027 bull $7.20 at 30x → ~$216. Consensus PT $198.06 implies roughly 31x the FY2027 base — i.e. the Street is underwriting no multiple compression at all.
What changed versus the June projection. The June base was FY2026 $4.75 / FY2027 $5.60 / FY2028 $6.35, and the conclusion was "~34x FY2026, ~29x FY2027 — the price already discounts the base case." Two quarters later the base is FY2026 $5.33 / FY2027 $6.40 / FY2028 $7.23, and the same price is 31.7x / 26.4x / 23.4x. The earnings caught up to roughly the same degree the multiple came down. The June sentence "you need the bull path to make money from here" is no longer supported by the arithmetic — the base path at a stable multiple is now roughly flat-to-modestly-positive over a year, and the bear path is a −29% de-rate rather than a −35% one.
(Per unattended-run rules, no our model create was logged. Suggested base forecast to log later: "APH FY2026 adjusted diluted EPS ≥ $5.25", p≈0.75, resolves 2027-02-15. Suggested falsifier: "APH IT datacom organic growth < +20% YoY in any quarter through Q2-2027.")
Bull case (strengthened). The Q2 print is the cleanest possible answer to the June bear case. Organic growth of 30% with margin expansion in all three segments, a 1.23 book-to-bill on record $10.7B of orders with every end market positive, and the largest acquisition in company history re-guided up 12% on sales and doubled on accretion within two quarters of closing. AI revenue is now a sized $10.5–11B run-rate business, not a narrative. Net leverage fell to ~1.46x from ~2x while the company was still buying back stock and raising the dividend. Crucially for the valuation argument: the forward multiple compressed from ~34x to ~31.7x while the stock rose 5%, because earnings outran price. Amphenol sells copper, power and fiber into the same rack, and Norwitt's "customers need more of everything" is corroborated by the fact that CommScope's AI business roughly doubled year over year under Amphenol's ownership.
Bear case (narrowed, but the core is intact).
Pre-mortem (18 months out, thesis broke). Q3 and Q4 2026 print in line — the guide was conservative and everyone knew it. Then Q1-2027 orders come in below the shipped revenue for the first time in six quarters: book-to-bill 0.9. Management explains it as "digestion of a very strong order intake" and points to the still-large backlog. Hyperscaler capex guidance for 2027 is flat rather than up. IT datacom organic growth falls from +63% to +12% in two quarters as the H2-2026 comps become impossible. The comms-networks weakness that was visible in Q2-2026 spreads to industrial. The distributor channel — 28.7% of sales — de-stocks, taking a quarter's worth of revenue out at once. The 27% tax rate and the ~$855M annual interest bill make the EPS decline steeper than the revenue decline. The Street re-rates from 31.7x to 22x on a bear-case FY2027 EPS of $5.45. Post-split the stock is ~$60 (pre-split ~$120), a 29% drawdown, and the business never broke — margins are still 26%, the balance sheet is still fine, and the compounder is still compounding. It was simply owned at a cyclical peak on a multiple that assumed no cycle existed.
Are multiples too high? Less clearly than in June. At 26.4x the FY2027 base for a business compounding EPS at 20%+ with 30% operating margins and a 37% ROE, the multiple is defensible on quality — the problem is not the multiple, it is that the denominator's growth rate depends on one demand wave. The honest framing: you are not overpaying for the company, you are underwriting the durability of AI datacom capex, and the price offers no compensation for being wrong about it.
Contrarian view (what the market refuses to see). Everyone is arguing about whether AI capex is durable. The more interesting number is that Harsh Environment did +22% organic at a 30.1% margin — a defence/aero/industrial business quietly running at margins the entire company could not reach a year ago. If AI datacom digests and the stock de-rates, the thing you are left holding is materially better than the thing the June dossier described, because the non-AI half of Amphenol got structurally more profitable during the boom. The market is pricing the AI beta and ignoring that the base business re-rated its own economics.
Dismantling the bull case with the new numbers.
the previous dossierThe fifteen questions carry forward intact; the Q2 call answered roughly two of them (Q3 on CommScope trajectory — answered with the raise to $4.6B/$0.30; Q4 on the China tax matter — partially answered, the $230M was paid but new foreign-tax reserves opened). Three additions the new material demands:
Every dossier we have written on Amphenol, newest first.
The Q2 print did the one thing the bear case could not survive
A 90-year connector compounder that the market has re-rated into an AI-infrastructure growth stock
Covered in the Knowledge Base
Hardware — Memory & Compute Architecture
| Industry | Semiconductors |
| Size | Public Company |
Where Amphenol sits against the other names we cover on this beat. Each line is that company’s verdict, not a summary of it.
The Q2 beat was not a peak — July revenue accelerated to +44.7% YoY, the board appropriated a single-tranche US$29.4B of capacity capital, and H1 cape…
The thesis got WIDER and the price got WORSE.
Cash $19.6B
The de-rate the June dossier warned about arrived early and for the exact reason it named
Cash $1.4B
LONG — MEDIUM (conviction DOWN one notch from the prior MEDIUM-HIGH, on wider risk, not a worse business).
Cash $25.0B
The circularity flag stopped being a footnote and became the tape
Cash $13.2B