Earnings Highlights: Semtech (SMTC) Q2 FY2027 — Data Center Tops $100M, Guided Up Another 45%: An Analog Veteran Rewritten by 1.6T

2 days ago

10 min read

Semtech delivered revenue of $341.9M (YoY +32.7%) and adjusted EPS of $0.71 (YoY +73.2%) this quarter, with EPS growing more than twice as fast as revenue. The real signal is not the headline total but the mix: data center revenue topped $100M in a single quarter for the first time (YoY +91%), and the company guided another 45% jump next quarter, implying YoY +160%. A company long filed under "analog + LoRa" is being redefined by 1.6T-generation TIAs, drivers and linear equalizers. At the same time, it is selling its lowest-margin cellular module business, lifting adjusted gross margin excluding that business to 59.7%, with next quarter guided at 63.9%.

1. Three Key Signals

Signal one: data center has gone from "one of the growth drivers" to "the main engine." Data center revenue reached $100M this quarter, YoY +91% and QoQ +39%, about 29% of total revenue. The company guided data center up another 45% in Q3, implying roughly +160% YoY, and management said outright that Q4's annual growth rate will "accelerate" further. This is a slope with no plateau in sight.
Signal two: profit leverage is outpacing revenue growth, and it is structural. Adjusted operating margin was 24.4%, with Q3 guided at 31.0% (QoQ +660bps, YoY +1,040bps); adjusted EPS jumps from $0.71 this quarter to $1.05 guided. There are two layers here: product mix (1.6T FiberEdge, CopperEdge and LoRa are all high-margin), and the structural uplift from carving out the cellular module business, which the company itself quantifies at 500bps or more.
Signal three: management framed "capacity," not demand, as the main constraint for FY2028. Targeted orders for this fiscal year are fully booked, and more than 70% of next fiscal year is already covered. CEO Hong Hou said plainly that currently secured capacity "may not be enough to support FY2028, especially the second half." The company is adding test equipment, expanding foundry and OSAT capacity in parallel, and is prepared to trade prepayments and CapEx for allocation. When a company starts talking about prepayments, its order visibility has already outrun its production lines.

2. Revenue and Financials

Breakdown by the three end markets:

Infrastructure $124M: QoQ +25%, YoY +69%. Within it, data center $100M, QoQ +39%, YoY +91%, driven by continued 800G strength, 1.6T CopperEdge shipments, and the start of the 1.6T FiberEdge ramp.
Industrial $179M: QoQ +16%, YoY +25%. Within it, LoRa-related revenue $58M, QoQ +31%, YoY +58%, an all-time high; IoT systems and connectivity $98M, up 11% both QoQ and YoY.
High-end consumer $39M: QoQ +2%, YoY −5%. TVS held on to sequential growth despite rising memory prices squeezing customers' BOMs, the only bright spot in this segment.

Cash flow also strengthened: operating cash flow $69M (QoQ +90%, YoY +55%), free cash flow $61M (QoQ +119%, YoY +48%), with CapEx at only 2% of revenue. Ending cash was $204M, debt principal $503M, and net leverage 1.1x. CFO Mark Lin said the CapEx ratio will rise with capacity expansion but will stay below 5% of revenue overall, with individual quarters possibly running slightly above depending on equipment delivery timing.

Semiconductor product gross margin was 62.8% (QoQ +210bps), above the top of guidance. This is the single most important number for gauging the contribution of the 1.6T product mix, and far cleaner than the consolidated gross margin.

3. Data Center: CopperEdge Lays the Base, FiberEdge Takes Over, Photonics Bets on 3.2T

The most worthwhile thing to unpack this quarter is the market's ingrained perception that "Semtech's data center = CopperEdge," which management spent a long time dismantling. Last quarter, in Earnings Highlights: Semtech (SMTC) FY2027 Q1, we already saw data center firing on all cylinders; this quarter's change is that momentum has spread from a single product line to three.

CopperEdge (linear equalizers): already a done deal. Management called its linear equalizer solution the industry's "de facto standard"; 1.6T products are ready for volume deployment, and design-ins go as high as 3.2T. More importantly, the applications are diverging: ACC cable adoption is tied to new platform design cycles (customers won't rip out AECs already in use to swap in ACCs), so volume ramps starting in Q4; but design activity for onboard linear equalizers is very intense, and that is volume that doesn't need to wait for a platform transition. For the division of labor between copper and optics in scale-up, see our full breakdown in SemiAnalysis: Copper vs. Optics Is a False Dichotomy.

FiberEdge (TIA + driver): the market-share story is worth more than the revenue story. 800G TIA share rose from about 18% to over 50% within two years; for 1.6T, share is expected to exceed 50% before the end of this fiscal year (next January). And the denominator is expanding at the same time: management relayed that industry expectations for annual 800G module demand have been revised up from about 50 million units at the start of the year to 80–90 million. Rising share multiplied by an expanding market is the formula behind this line's revenue explosion this year. For the BOM changes the 1.6T generation brings, compare with 800G vs. 1.6T Optical Module BOM Teardown.

Notably, driver revenue "hasn't started yet": management said the products are under customer evaluation and "will be contributing very meaningfully." This is an increment not yet reflected in the numbers. MACOM is also going after the same TIA/driver pie; see Earnings Highlights: MACOM (MTSI) FY2026 Q3 for a sense of how crowded this race is.

Photonics (post-HieFo acquisition): only gain chips are currently shipping in volume. High-power CW lasers and semiconductor optical amplifiers (SOAs) are expected to begin sampling and customer qualification "in the coming months," and CW laser revenue contribution from transceivers won't arrive until the first half of FY2028. The company has also hired an industry veteran to lead a photodiode (PD) design team, aiming to build PD arrays co-optimized with its own TIAs.

On capacity: the HieFo line has secured an adjacent, ready-to-use building with facilities already completed, and fab capacity can be raised 3–4x by year-end.

Management laid out the content-value math clearly: from 800G to 3.2T, Semtech's content per optical module is set to rise from "high single-digit dollars" to "high double-digit." An analyst misread this on the call as the teens, and Hong Hou clarified on the spot that he meant $80–90. That is a roughly tenfold jump in content value, but it is realized only at 3.2T.

4. NPO/XPO: The Next Position Semtech Wants to Lock Down

Management and analysts spent considerable time in the Q&A on NPO (near-packaged optics). The logic is straightforward: total bandwidth is exploding, but chassis shoreline is fixed. The only answer is higher bandwidth density, and high-density packaging squeezes component spacing and makes crosstalk and signal integrity harder. That is exactly why "electrical (TIA/driver) and optical (laser/PD) must be co-designed." Semtech said it is already in 10–15 NPO-related projects, some engaging end customers directly.

Its positioning assets: it is already a leading TIA array supplier, its linearized laser arrays perform well, and PD arrays are being added. On top of that, LPO's track record has built industry confidence in linear architectures, and that confidence is spilling over to NPO and some forms of CPO; XPO itself includes linear, non-retimed architectures. For the full picture of this route battle, see our breakdown in OCP APAC 2026: CPO/NPO/XPO Panel.

It's not a battle of routes; it's about who can put electronics and optics in the same package, and still mass-produce it, before the shoreline runs out.

5. LoRa's Second Engine: Three Pillars and Amazon Sidewalk

LoRa hit a record $58M this quarter, with Q3 guided another +15% QoQ, +65% YoY. An analyst directly challenged the old "20% annual growth" benchmark; Hong Hou replied that they had "certainly smashed the 20% ceiling" and that annual growth above 20% "is sustainable." This is one of the few passages on the call with the tone of a medium-term commitment.

Status of the three pillars (per management):

LoRaWAN (industrial/commercial): still the bulk of revenue.
LoRa Plus (smart home/security, multi-protocol): about 20–25% of LoRa revenue. The Gen 4 platform supports dual-band operation and raises throughput to 2.6 Mbps while retaining low power and long range.
Amazon Sidewalk (mass consumer): currently still nominal, expected to be a high single-digit share this year. Ring has launched a LoRa sensor product line, and Sidewalk is expanding from the US to Canada and Mexico, followed by Europe, Australia and Japan. A dose of calm is warranted here: Sidewalk offers "Amazon-scale" upside, but its actual contribution this year is small. LoRa's current growth comes mainly from LoRaWAN and LoRa Plus, not Sidewalk. Conflating the two will overstate this line's near-term slope.

6. Management Outlook: Q3 Guidance and Language Analysis

Q3 FY2027 guidance (non-GAAP):

Item	Guidance	Implied change
Revenue	$410.0M ± $5.0M	QoQ +20%, YoY +54%
Adjusted gross margin	58.3% ± 100bps	QoQ +380bps, YoY +530bps
Adjusted gross margin (excl. module business held for sale)	63.9%	+560bps vs. consolidated basis
Adjusted operating expenses	$112.0M ± $3.0M	Higher R&D, lower SG&A ratio
Adjusted operating margin	31.0%	QoQ +660bps, YoY +1,040bps
Adjusted EBITDA	$134.3M (32.8% margin)	QoQ +620bps
Adjusted diluted EPS	$1.05 ± $0.03	QoQ +48%, YoY +119%

Segment guidance: data center QoQ +45% (about +160% YoY), LoRa QoQ +15% (YoY +65%), and high-end consumer growing on seasonality and TVS share gains. Tax rate assumption 18%, weighted share count 99M.

Certainty grading of management language:

Confirmed: "Targeted orders for this fiscal year are fully booked," "very high confidence and conviction in Q3 and Q4," "customers want the parts tomorrow" — statements backed by backlog.
Expected: "1.6T FiberEdge share above 50% by year-end," "FY2028 more than 70% covered" — backed by orders but still to be verified.
Target/vision: "content value rising to high double-digit dollars," "3.2T in about 18 months, meaningful deployment in about 2 years" — long timelines with many variables; a roadmap, not a commitment.

The CFO also explicitly anchored the "starting gross margin after the module sale" at 64%, and stressed that 800G is seeing "no price erosion." Hong Hou added more detail: new products can command a premium, but it fades; right now customers value supply availability over price, rising costs can be negotiated and passed through to customers, but the company deliberately won't use its seller's leverage to "gouge" customers. The subtext: price is their chip to trade for long-term agreements and allocation, not a tool to juice margins this quarter.

7. Supply Chain and Customer Clues

Full coverage of module makers: management said it is "designing to every module provider" in its target markets, with sole-source positions at several.
More direct hyperscaler engagement: ACC has clear visibility with "a leading hyperscaler," ramping from Q4; NPO has 10–15 projects, some tied directly to end customers.
Two capacity moves: adding test equipment in the back end, and qualifying new manufacturing partners to diversify geopolitical risk; expanding capacity in the front end with "a leading partner," using tools including prepayments and CapEx.
Geopolitical risk explicitly flagged: one of the few self-disclosed risks on this call, and worth adding to the watch list.
Interoperability is not yet an issue: for ACC, customers care most about availability, interoperability second; the company said it "hasn't seen a lot of activity requiring us to interoperate with other vendors." That favors Semtech near term, but it also means once standardization kicks in, the moat will be repriced.
Sale of the cellular module business: a definitive agreement has been signed, closing in Q4, neutral to non-GAAP EPS. Management classified IoT routers, gateways and platforms as "a portfolio we like for now," but said explicitly that portfolio optimization is a "continued journey" — further moves are fairly likely.

8. Risks and Counterarguments

First, the slope of the guidance is itself a risk. Q3 data center +45% QoQ rests on a "record backlog," but converting that backlog depends on wafer, test and packaging capacity all arriving at the same time. Management itself said capacity for the second half of FY2028 "may not be enough." If supply gets stuck, growth is deferred rather than lost, but that's not usually how the stock price reads it.

Second, the tenfold content story is tied to 3.2T, 18–24 months out. CW laser revenue from transceivers starts only in the first half of FY2028, and PD arrays even later. In between there is a stretch where "spending has happened but revenue hasn't arrived," which will show up in steadily rising R&D.

Third, the 1.6T share figures are company estimates. "Above 50% by year-end" has no third-party validation yet, and competitors in 1.6T TIAs/drivers won't stand still.

Fourth, high-end consumer is still declining year over year. At −5% YoY, with rising memory prices squeezing smartphone BOMs, there's no near-term reason to expect this segment to reaccelerate.

Fifth, the gap between non-GAAP and GAAP is large. GAAP EPS of $1.59 this quarter exceeded non-GAAP $0.71 due to tax and one-time items, the opposite of what most people expect; if those items reverse next quarter, the GAAP numbers will look ugly. Always read Semtech on a non-GAAP basis, and keep your own eye on what makes up the adjustments.

Conclusion

Semtech accomplished one thing this quarter: it rewrote itself from a company "propped up by LoRa and noticed for CopperEdge" into an optical interconnect chip supplier spanning TIAs, drivers, linear equalizers, gain chips, CW lasers, SOAs and PDs. The financial evidence: data center topping $100M in a quarter, 62.8% semiconductor product gross margin, and a structural 500bps margin lift from selling the module business.

From next quarter on, what this company has to prove is no longer "is there demand" but "can it secure capacity."

Three metrics to track over the next three months:

1. Actual 1.6T FiberEdge market share: the company says above 50% by the end of this fiscal year. The October 15 investor day (San Jose) will provide a TAM and share breakdown, and that multi-year model is the single most important public reconciliation for validating this line. 2. The concrete form of capacity expansion: whether wafer prepayments, long-term agreements or a sharp upward CapEx revision show up. If they do, FY2028 growth is "locked in" rather than "expected." 3. The first meaningful revenue from drivers and ACC: management has placed both after Q4, and they are the first hard evidence for the "more than just CopperEdge" thesis.

*This article is for technology and industry trend analysis only and does not constitute investment advice.*