SMTC · Semtech Corporation · Stock research

Current Thesis

The August recovery has round-tripped. Price closed 2026-08-14 at $140.35; the reference close for 2026-08-21 is $124.09, which sits beneath the 2026-07-17 flush close of $124.96 that the August bounce was supposed to have reclaimed. That is a full give-back of the post-divestiture leg inside four sessions, and it happened in a week when the AI-semis tape was strong (2026-08-17: SanDisk +10%, Micron +6%). RSI(14) is 52.3 — mid-range, so the slide was orderly rather than a capitulation. The stock is 29.0% below the $174.73 52-week high with a three-month price change of -20.9%.

The binary also moved forward. Semtech's conference-call announcement puts Q2 FY27 results on 2026-08-25 after the close, with the call at 1:30 p.m. PT / 4:30 p.m. ET — not the 2026-09-02 date that sat on third-party earnings calendars. Two days out, consensus sits directly on top of the guide: Zacks marks $328.4M revenue (+27.5% YoY) and $0.62 adjusted EPS against management's $328M ±$5M and $0.61 ±$0.02 set on 2026-05-25. An in-line July quarter therefore resolves very little. What the tape will price is the October-quarter outlook, which Benchmark models at $360.1M revenue and $0.73 EPS.

The narrative leg on offer is unchanged from the July and August notes — mix purification (the 2026-08-13 agreement to sell the cellular module unit to Compal for $62M cash, leaving data center interconnect plus LoRa) layered on a data center line compounding off a small base. What has changed is that the market is no longer paying for it above $140, and a new sell-side house has anchored the valuation far lower than the incumbents.

Bullish and bearish views on Semtech Corporation

The model's bull view on Semtech Corporation (SMTC), in brief: Q1 FY27 on 2026-05-25 was a guide-raise, not a beat-and-hold: record revenue $291.0M (+16% YoY), adjusted EPS $0.51 against $0.45 consensus, and a Q2 guide of $323–333M versus roughly $301M consensus at the time. The bear view: Two failed attempts at the $140 shelf since mid-July. Both cases follow in full.

Bull Case

• Q1 FY27 on 2026-05-25 was a guide-raise, not a beat-and-hold: record revenue $291.0M (+16% YoY), adjusted EPS $0.51 against $0.45 consensus, and a Q2 guide of $323–333M versus roughly $301M consensus at the time.
• Data center is the growth engine: Q1 FY27 data center revenue ~$71.6M, about 25% of total, guided +35% sequentially and +85% YoY, with FY27 data center reiterated at roughly +50% YoY across CopperEdge ACC, LPO, near-package optics and FiberEdge.
• No negative pre-announcement drift in estimates: the Zacks EPS consensus of $0.62 for the July quarter was unchanged over the trailing 60 days as of the 2026-08 preview, and implies +51% YoY.
• Benchmark reiterated Buy at $230 on 2026-08-20, four sessions before the print, with analyst Cody Acree writing that "October data-center mix, margin quality and backlog conversion should matter more than a small July variance," and citing 800G optical demand and improved bookings/backlog visibility.
• Coverage is still expanding, not contracting: BMO Capital initiated on 2026-08-21 at Outperform, framing Semtech as positioned to take share in emerging ACC/LPO networking as hyperscaler capex stays elevated. Incumbent targets as of 2026-08-20: UBS $225, TD Cowen $210, Needham $200 (reiterated 2026-08-14), Stifel $188.
• The divestiture removes the lowest-strategic-fit line for cash: the 2026-08-13 definitive agreement transfers substantially all assets, IP, customer relationships and personnel of the cellular module business to Compal for $62M, closing expected fiscal Q4 FY2027. That unit arrived with the 2023 Sierra Wireless acquisition, a $1.2B transaction.
• CopperEdge carries a measured power argument: the GN8234 redriver ran live 1.6T active copper against NVIDIA 224G/lane SerDes at OFC 2026 at under 2W per cable end and sub-100ps latency; the GN8304 previewed 3.2T at 448G/lane.
• Balance-sheet headroom added 2026-07-06: a $360M Morgan Stanley-led revolver maturing 2031-07-06, undrawn, replacing the prior JPMorgan facility.

Bear Case

• Two failed attempts at the $140 shelf since mid-July. The 2026-08-14 close of $140.35 was surrendered by 2026-08-20 ($124.39 per Benchmark's note) and the 2026-08-21 reference close of $124.09 is below the 2026-07-17 flush close of $124.96. The July closing floor did not hold.
• The marginal new analyst arrived with the lowest number in the band. BMO's $155 initiation on 2026-08-21 sits $20 under the previously reported low target of $175 in the 14-analyst S&P Global poll of 2026-08-14, and roughly $50 under that poll's $205.25 average. A new house underwriting the same ACC/LPO thesis at a materially lower valuation is a datapoint about what the story is worth, not about whether it is true.
• Consensus offers no cushion. $328.4M versus a $328M ±$5M guide leaves no room for an in-line print to be read as a beat. The asymmetry sits entirely in the October guide against Benchmark's $360.1M.
• The divested unit's revenue contribution has still not been disclosed as of 2026-08-21. Until management quantifies it — 2026-08-25 is the first natural venue — the go-forward revenue base and the FY28 dilution are unpriced.
• Adjusted EPS is doing all the work. TTM GAAP net income was -$33.2M (EPS -$0.38) as of the 2026-07-19 review; every target in the $155–$230 band rests on non-GAAP figures.
• CopperEdge 1.6T revenue concentrates in one named hyperscaler per the FY26 10-K. A design change at that account is a step down in the line, not a soft quarter.
• Insider supply into the June–July strength: roughly $5.4M of insider selling over the three months to 2026-07-19, with no open-market purchases.

Setup & Price Structure

Reference close 2026-08-21: $124.09. That is 29.0% under the $174.73 52-week high and a three-month price change of -20.9%. RSI(14) at 52.3 is the unusual number here — momentum has not gone oversold despite price returning to the July lows, which is what a controlled distribution out of a range looks like rather than a flush.

Structure: the $140 area, defended on 2026-08-14, is now overhead. The 2026-07-17 weekly close of $124.96 was the prior floor and the latest weekly close undercuts it by 87 cents. Below that there is no dated closing shelf in the recent record to point at, which is why the level that matters for a break is set with a buffer rather than at the money.

The narrative is maturing. Dating it: coverage is still being initiated (BMO, 2026-08-21) and the business is still compounding into the print (+27.5% YoY expected revenue on 2026-08-25), so the narrative has not failed — but price has made no net progress since the 2026-05-25 guide-raise and the marginal bid has twice refused $140. It is not accelerating: there is no fresh high, no expanding participation, and the 2026-08-15 Benzinga "Stock Whisper" mention is the quiet end of retail attention rather than a broad bid. It is not dead: the twelve-month gain Benchmark cited on 2026-08-20 was 159%, and no house has downgraded through the drawdown.

Crowding and positioning observables, stated as observables: published targets span $155–$230 against a $124.09 close, a spread that has widened rather than converged since 2026-08-14; short interest was 6.01% of float (5.56M shares) as of 2026-07-19; beta 2.34; insiders sold roughly $5.4M in the three months to 2026-07-19 with no open-market buys; and an earnings date sits two sessions away, which compresses any pre-print positioning decision into a coin-flip window.

Catalyst Calendar (next 30 days)

• 2026-08-26 — first full session of reaction; the sell-side revision window reopens and the $155–$230 target band gets re-marked or defended.
• ~2026-09-08 (est.) — Form 10-Q for fiscal Q2. The company has not dated this;
• ~2027-01-31 (est.) — expected close of the Compal transaction in fiscal Q4 FY2027, subject to regulatory approval. Outside the 30-day window, but it is the event that actually converts reported revenue to a data center plus LoRa mix.

Elapsed catalysts

• 2026-08-25 — Q2 FY27 results, after the close; conference call 1:30 p.m. PT / 4:30 p.m. ET. Marks the $328M ±$5M revenue guide and $0.61 ±$0.02 adjusted EPS, the October-quarter outlook against Benchmark's $360.1M model, the FY27 data center +50% YoY framing, and the first management quantification of the Compal divestiture's revenue and margin effect. (passed 1d ago)

What Would Change Our Mind

The structure has already bent: the July closing floor at $124.96 gave way on 2026-08-21, so the question is no longer whether the August rally failed but whether the June–August range is a base or a distribution top. A weekly close below $118 answers that in favour of distribution — it would put price roughly 32% under the 52-week high with no dated closing shelf beneath it and the $140 area twice rejected above.

On the fundamentals, the specific things that would break the leg: a 2026-08-25 revenue print under $323M; an October-quarter guide below the ~$360M the sell-side is modelling; FY27 data center growth walked below the +50% YoY framing set on 2026-05-25; or a disclosed cellular-module revenue contribution large enough that FY28 consensus resets lower once the divestiture is modelled out. New Form 4 filings showing sales executed into any post-print recovery would extend the insider-supply pattern observed through 2026-07-19.

The other direction is equally specific. A weekly close back above $140 with an October guide above $360M would re-establish the leg that failed twice since mid-July, and would date the July–August drawdown as multiple compression on a sector move rather than a change in the business. Absent both, the print comes and goes and the name is left trading on the semiconductor tape.

Correlation Notes

• CRDO is the higher-beta pure-play on the same interconnect theme; relative strength between the two separates sector beta from company-specific moves. MRVL, ANET and Amphenol are the adjacent reads on ACC-versus-optics attach rates inside the rack.
• Beta 2.34 means the name amplifies semiconductor-group moves in both directions; the 22.7% decline into 2026-07-17 had no company-specific trigger.
• A divergence worth watching: on 2026-08-17 the AI-memory complex ripped (SanDisk +10%, Micron +6%) while SMTC was trading from $140.35 toward the $124.39 level Benchmark referenced on 2026-08-20. Strength in the AI-semis tape did not transmit to this name in that window.
• Hyperscaler capex commentary is the upstream driver for the CopperEdge and FiberEdge lines; any top-three customer guiding 2027 capex down reads through directly to the +50% FY27 data center framing.