CALX › analyze
CALX · Analyze
Price at analysis: $37.69 (pre-market, prior close $38.36) · 52-week range $34.26–$71.22 · down ~47% from 52wk high, ~10% above 52wk low · down roughly 30% YTD · Market cap $2.37B · EV $2.19B (net cash) · Sector: Technology / Industry: Software-Infrastructure (filed here under Infrastructure — telecom-access hardware + attached software, per this repo's sector convention)
0. Knowledge check
Knowledge/INDEX.md and python .mcp/kb.py find CALX returned no live note — this is a fresh name for the knowledge base. No prior verdict to reconcile against. Three Playbook pitfalls fired during data-gathering and are applied below:
- [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]] — fired directly (see §3).
- [[pitfall-single-quarter-fcf-read-as-ttm]] — checked against; TTM FCF below is a genuine 4-quarter sum, not a single column (see §1).
- [[pitfall-yahoo-insider-purchases-counts-rsu-grants]] — the
insider_purchasessummary field was not used; the rawinsider_transactionsrows were read directly (see §4).
1. Fundamentals
The cycle, not a smooth trend
Revenue and earnings at CALX have gone boom → bust → recovery over the last four fiscal years, driven by the post-COVID rural-broadband funding cycle (RDOF/ARPA pulling orders forward into 2023, then a channel-inventory digestion crash in 2024):
| FY | Revenue | YoY | Net Income | Diluted EPS |
|---|---|---|---|---|
| 2022 | $867.8M | — | $41.0M | $0.60 |
| 2023 | $1,035.8M | +19.4% | $29.3M | $0.46 |
| 2024 | $831.5M | -19.7% | -$29.7M | -$0.45 |
| 2025 | $1,001.7M | +20.5% | $17.9M | $0.26 |
A naive 3-year revenue CAGR of 4.8% completely hides this — the number that matters is which leg of the cycle the company is on now. The quarterly print says recovery, accelerating:
| Quarter | Revenue | YoY | Net Income | Gross Margin |
|---|---|---|---|---|
| Q2 FY25 | $241.9M | — | -$0.2M (loss) | 56.3% |
| Q3 FY25 | $265.4M | — | $15.7M | 57.3% |
| Q4 FY25 | $272.4M | — | $7.2M | 57.7% |
| Q1 FY26 | $280.0M | — | $11.2M | 56.9% |
| Q2 FY26 | $293.3M | +21.3% | $17.1M (record) | 54.6% |
Revenue growth is genuinely accelerating and Q2 FY26 was a record profit quarter — but gross margin stepped down sequentially (56.9% → 54.6%) on rising memory-component (DRAM/NAND) costs, which management says will bottom in Q3 FY26 (guided non-GAAP gross margin 50.5–53.5%) before recovering via a memory surcharge program — explicitly designed to be gross-profit-neutral over time, i.e. a pass-through, not a margin gain.
Free cash flow — the TTM number is genuinely down, not misread
Applying the discipline from [[pitfall-single-quarter-fcf-read-as-ttm]] (never take TTM from one column — sum four quarters):
| Quarter | FCF |
|---|---|
| Q3 FY25 | $26.7M |
| Q4 FY25 | $40.3M |
| Q1 FY26 | $6.5M |
| Q2 FY26 | $11.9M |
| True TTM FCF | $85.4M |
That is below FY2025's full-year $115.5M, driven by a deliberate inventory build (likely advance-purchasing memory ahead of further cost inflation / tariff exposure — inventory change was -$25.9M in Q2 FY26 alone) and higher working-capital consumption, not a copy-paste error. FCF/share (TTM, ~68.6M avg diluted shares) is ~$1.25; on FY2025's less-representative annual figure it's $1.67. The 3yr FCF "CAGR" of 106% is an artifact of a near-zero FY2022 base and should not be extrapolated.
Quality-of-FCF flag: stock-based comp was $87.9M in FY2025 — roughly 4.9x GAAP net income and ~8.8% of revenue. FCF is legitimately cash, but the SBC add-back is large enough that it materially overstates the economic return to existing shareholders before dilution.
Capital allocation (FY2025)
OCF $135.0M → capex -$19.4M (1.9% of revenue, light) → FCF $115.5M → buybacks -$93.6M (81% of FCF), no dividend, no debt activity. Q1 FY26 alone saw an outsized $170.9M buyback, partly funded by liquidating $66M of the investment portfolio — an aggressive repurchase into the January 2026 trough (~$44–53/share). With hindsight it was well-timed relative to today's even lower price, but it also drew down the cash cushion. Management describes this as a now-completed $431M repurchase program.
Balance sheet — genuinely clean
Total debt $14–16M against $1.06B assets (Debt/Assets ~1.5%), net cash position (~$180M net cash), current ratio 2.83x, quick ratio 1.57x. There is no leverage risk in this name — whatever else is wrong with the thesis, a liquidity crisis is not one of the risks.
Share count
GAAP diluted shares fell from 70.4M (Q4 FY25 avg) to 65.5M (Q2 FY26 avg) in two quarters — an ~7% drop, the first time in years buybacks have outrun SBC-driven issuance rather than merely offsetting it (the 3-year CAGR through FY2025 was essentially flat, +0.2%/yr). This is a real, recent inflection — not yet a proven multi-year trend.
Per-share / returns
Revenue/share (TTM) ~$16.92, ROE (TTM) 6.9%, ROA (TTM) 4.1%. Annual GAAP ROIC (roic.ai) was 1.3% in FY2025 and negative in FY2024 — the absolute level of returns on capital is still low; the story is margin recovery, not an established high-ROIC compounder. No Small/Mid-Cap-overlay red flags beyond what's already covered in Moat/Sentiment below; no dividend/REIT/BDC overlay applies (CALX pays no dividend).
2. Moat & Competitive Advantage
Revenue-stream map: (1) hardware appliances — ONTs, gateways, GigaSpire/GigaPro Wi-Fi systems — increasingly commodity, and the layer absorbing the current memory-cost shock; (2) Calix Cloud / Calix One software platform (Engagement, Operations, Service Cloud) — subscription, recurring, called out this quarter as "record software and services revenue"; (3) SmartHome/SmartBiz/SmartTown managed services — ancillary recurring revenue riding on the same customer relationship.
Quantitative base: Gross margin has genuinely trended up over three years (50.2% FY22 → 49.8% FY23 → 54.5% FY24 → 56.8% FY25) as the software/services mix rises — a real, structural signal, not noise. But ROIC has not followed (1.3% FY25, negative FY24) because operating expense (R&D + SG&A) still runs ~52% of revenue; the margin gain hasn't yet translated into operating leverage. This is an early-stage margin story, not yet a proven compounder.
Adversarial stress-test — "you're a well-funded rival, how do you attack this?" The hardware layer is the easy entry point: ONTs and gateways use increasingly standardized silicon, and Nokia, ADTRAN, and (smaller) DZS all compete for the same regional/rural broadband-provider (BSP) customer base. A rival with a bigger balance sheet (Nokia) could bundle a comparable cloud/ops layer for free to win hardware share — the exact commoditization risk the current memory-cost episode is a preview of. The defensible layer is the switching cost of an embedded cloud operations/marketing platform at small ISPs with thin IT staff — ripping it out is genuinely costly once workflows run on it — but that moat is early and unproven at scale against a determined, well-capitalized attacker willing to lose money on software to win the hardware contract.
Disruption vectors: - Near-term, cyclical: the DRAM/NAND cost supercycle (AI-datacenter demand pulling memory prices up industry-wide) — this hits every networking-hardware vendor, not CALX specifically, and management's surcharge program is a reasonable, if reactive, defense. - Medium-term, structural: the June 2025 BEAD "Benefit of the Bargain" policy shift (NTIA/Trump administration) made the $42B rural-broadband program technology-neutral and lowest-cost-wins, removing the fiber preference that underpinned CALX's original 2025 bull case — fixed-wireless and LEO satellite can now win BEAD dollars against Calix's fiber-centric regional-ISP customers. This is a genuine, permanent cap on the size of the original growth narrative, not sentiment noise (see §4). - Long-term: the software-attach/AI-native-platform strategy is the correct structural answer to hardware commoditization, and the "record software revenue" this quarter is real evidence it's working — but it is still early innings, unproven against a full competitive response.
Evergreen assessment: Not yet a forever business. CALX remains meaningfully levered to hardware cost cycles, a single federal funding program for a chunk of its customer base's capex, and a software transition that is real but incomplete. The moat is "in the making," not established.
3. Valuation
Vendor forward-EPS trap fired — corrected here. Yahoo's forwardPE (16.4x) prices forwardEps ($2.30), which is FY2027 consensus (Calix's fiscal year = calendar year, so this is two years out), not the current fiscal year. The correct current-year read is epsCurrentYear $1.72 / priceEpsCurrentYear 21.9x — a ~25% understatement in the raw vendor field, exactly the failure mode documented in [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]]. The true current-year (FY2026) multiple is 21.9x, not 16.4x.
| Model | Basis | Result | Read |
|---|---|---|---|
| Graham IV (trailing) | EPS $0.75 (TTM, GAAP), BVPS $11.28 | $13.79 | Price 2.7x Graham — but TTM EPS is trough-depressed (includes a near-loss quarter); low weight |
| Graham IV (current-year) | EPS $1.72 (consensus), BVPS $11.28 | $20.90 | Still ~45% below spot even on the optimistic number — Graham keeps saying "not a classic asset-backed value name," useful as a floor check, not a verdict |
| P/E (current-year, corrected) | $1.72 EPS | 21.9x | Roughly a market multiple for a cyclical-recovery networking name — not statistically cheap |
| P/FCF (TTM, real 4-qtr sum) | FCF/share ~$1.25 | ~30x | Not cheap on the depressed trailing base |
| P/FCF (FY2025 annual) | FCF/share $1.67 | ~22.6x | More representative of the "good" run-rate, still not a screaming discount |
| Bogle expected return | No dividend; return = EPS/FCF growth ± multiple change | Wide range | If the guided ramp (rev +15-20%, margin recovery, BEAD Q4 flow) lands and the multiple holds ~22x: attractive (20%+). If the ramp disappoints and the multiple compresses to 15-18x (a second guidance miss after the April event): flat-to-down from here. No fake precision — this is genuinely a coin-flip-shaped bet on H2 execution. |
| DYT / DDM | N/A | — | No dividend; not applicable per framework |
Fair value range: $32–46. This is set by the FCF-multiple bracket (23–30x on a genuinely uncertain FCF base) reconciled against the corrected current-year P/E (21.9x, in-line-to-slightly-rich for the risk profile) and Graham's floor-check skepticism. Current price ($37.69) sits in the lower-middle of that range — reasonable after a real drawdown, not a statistically screaming bargain. Per [[principle-down-a-lot-is-not-cheap]], the 47% fall from the 52wk high does not by itself make this cheap; it is cheap-ish relative to a fair range that itself carries real uncertainty.
Entry: $32–37 — the current price is already inside this zone; actionable now for anyone sizing into the name.
Trim: 60x ttm (~$45 today). Set on trailing (ttm) EPS, not the vendor's fwd field, per [[pitfall-multiple-trim-inherits-the-broken-vendor-field]] — CALX's forwardPE is corrupted (FY2027, see above), and a Trim NNx fwd line would inherit that error directly. Caveat that must travel with this number: TTM EPS ($0.75) is currently trough-depressed because a near-loss quarter (Q2 FY25) is still inside the trailing window. As that quarter rolls off and Q3/Q4 FY26 (guided stronger) roll in, TTM EPS will rise mechanically over the next two prints, lifting this dollar level with it even absent further re-rating — which is the intended behavior for a recovering cyclical, but this multiple should be explicitly re-anchored at the Q3 FY26 print (~2026-10-26) once the base normalizes, not left on autopilot for a year.
4. Sentiment — why is it near its 52-week low, right now?
Four distinct forces, not one clean narrative:
1. A genuine credibility hit (April 21, 2026). CALX disclosed that memory-component cost pressure was compressing appliance gross margins faster than previously flagged; the stock fell ~14% in a day. A securities class action followed (class period Jan 28–Apr 21, 2026), alleging management touted "record" Q4 FY25/Q1 FY26 gross margins without disclosing that the advance-purchased memory buffer supporting them was depleting — i.e., the strong margins investors were buying weren't a sustainable run-rate. Lead-plaintiff deadline was July 27, 2026; litigation is ongoing. This is a real overhang on how much the market will trust the next guide, independent of the underlying business.
2. An industry-wide, plausibly cyclical cost shock. The DRAM/NAND cost supercycle (AI-datacenter demand pulling memory prices up) is compressing hardware margins across the networking-equipment sector, not CALX alone. Management's memory-surcharge program is a pass-through defense, guided to bottom gross margin in Q3 FY26. If this plays out as guided, it reverses — this is the "opportunity" half of the setup.
3. A genuinely structural policy shift. The June 2025 NTIA "Benefit of the Bargain" BEAD restructuring made the $42B rural-broadband program technology-neutral and lowest-cost-wins, removing the fiber preference that drove CALX to its 2025 highs (52wk high $71.22, all-time high $80.95). This permanently lowers the ceiling on the original growth narrative — fixed-wireless and satellite can now win dollars that would previously have gone to Calix's fiber-centric regional-ISP customers. Management still expects a BEAD tailwind starting Q4 FY26, but a smaller, later, less certain one than the original 2023-2025 thesis assumed. This is the "trap" half of the setup and it does not go away when memory costs normalize.
4. Persistent, one-directional insider selling — no golden flag. Reading the raw insider_transactions rows (per [[pitfall-yahoo-insider-purchases-counts-rsu-grants]], the summary field is unreliable) shows CEO Weening, CFO Sindelar, and 10%+ holder/director Carl Russo selling in nearly every month since mid-2025, prices ranging $35–70, via the standard option-exercise-then-sell pattern consistent with scheduled 10b5-1 plans. Zero open-market insider purchases appear anywhere in the transaction history — not a smoking gun on its own, but the analytical framework's golden flag (insiders buying into weakness) is simply absent here.
Sell-side stance stayed constructive but has been trimming confidence all year: JPMorgan Overweight (PT cut $65→$58, Jul 22), Needham Buy (PT cut $70→$62, Apr 23), Rosenblatt Buy (held at $55), Northland upgraded Market Perform→Outperform (Jul 21, $52 PT). Mean target $62.33 — +65% above spot — but the pattern of cutting price targets while holding ratings says analysts are recalibrating the magnitude of the ramp downward even as they keep the thesis alive.
Net read: mixed, not clean. Part of this drawdown is cyclical overreaction to a transient input-cost shock (opportunity); part is a legitimate structural repricing of the BEAD-driven growth story (not fully a trap, but a real haircut to the ceiling); part is a credibility tax from the April disclosure event that the market will likely hold against the stock until at least one more clean quarter is delivered.
5. Synthesis — weighted verdict
CALX is a smaller/mid-cap cyclical-recovery tech name, not a mature dividend payer — per the framework's conflict-resolution weighting, Moat + Sentiment carry more weight here than a static Graham read, but Valuation's FCF/current-year-multiple discipline is still the anchor that keeps optimism honest.
Named tensions: - Fundamentals vs. Valuation: the balance sheet is genuinely bulletproof (net cash, ~1.5% debt/assets) and FCF is recovering — but the stock is not statistically cheap even on the corrected (21.9x current-year) multiple. The "quality on sale" framing is more "reasonable entry after an overreaction" than "screaming statistical bargain." - Moat vs. Sentiment: the software-transition thesis (rising gross margin, "record" software/services revenue) is real evidence of moat-building — but it's early, and the BEAD policy shift structurally caps the size of the original bull case regardless of how well the software pivot executes. - Sell-side vs. price action: analysts still see ~65% upside to mean target, but have cut targets across three of four ratings actions this year — the Street is recalibrating down even while staying long, which should temper how much weight the consensus number itself carries.
Verdict: ACCUMULATE, conviction 6.0/10. This is a "close call because" case: real balance-sheet safety and a genuine (if early) margin-mix improvement justify starting or adding to a position at current levels, inside the stated entry zone — but the unresolved securities litigation, the unproven H2 FY26 margin-recovery-plus-BEAD-tailwind ramp embedded in consensus, and the structural (not just cyclical) nature of the BEAD policy shift keep this from being a high-conviction BUY. The next real test is the Q3 FY26 print (~Oct 26, 2026): does gross margin actually bottom as guided, and does management's BEAD-Q4 commentary firm up or soften further.
Key risks that break this thesis: 1. Gross margin does not bottom in Q3 as guided (memory costs keep rising faster than the surcharge program can pass through) — this would be the second guidance disappointment inside twelve months and would likely trigger a much harsher de-rating given the fresh litigation overhang. 2. BEAD Q4 tailwind fails to materialize or is smaller than expected under the technology-neutral rules — the structural cap on the original growth narrative becomes the dominant story instead of a manageable haircut.