Series: LoRaWAN Technical Reading · Issue #17 (Industry Trends #7) Published: 2026-09-25
On March 7, 2012, an American analog chipmaker called Semtech announced the acquisition of Cycleo, a Grenoble-based French startup. The consideration: $5 million in cash at closing, plus up to $16 million in earnouts tied to revenue and operating profit goals over four years — a total ceiling of roughly $21 million.
What Cycleo held was a patent on a long-range wireless modulation technique aimed at smart metering. That technology would later be named LoRa.
Fourteen years later, in 2026, that acquisition has produced: LoRa chip revenue of roughly $58 million per quarter (the quarter ended July 2026, up 58% year over year), 125 million LoRaWAN-connected devices, an ecosystem spanning 70 countries, and an ITU international standard adopted by more than 170 mobile network operators.
The first counter-intuitive fact is this: the highest-return acquisition in IoT history was made by a buyer who did not know what it was buying. Semtech's own announcement stated that Cycleo's IP "will be incorporated in future Semtech RF platforms" targeting "energy control, security and asset management markets." Not a single word about the Internet of Things, let alone LPWAN. The word LoRaWAN would not be invented for another three years.
The second counter-intuitive fact is told far less often: this "patent monopolist" nearly collapsed under its own next acquisition in 2024. In early 2023, Semtech spent roughly $1.2 billion acquiring the cellular IoT module maker Sierra Wireless, driving its net leverage ratio to 9.0x — in the semiconductor industry, a step away from debt restructuring. What saved the company were two engines: AI data-center optical interconnect silicon on one side, and the $21 million LoRa business on the other.
This article explains three things: how the LoRa ecosystem structure actually evolved over those fourteen years; where exactly the boundary between "open" and "proprietary" sits; and what that structure means, as of 2026, for companies building devices, gateways, and platforms.
The popular retelling of the 2012 acquisition contains several distortions.
Detail one: the price was not a king's ransom. It was pocket change.
Per Semtech's filing with the U.S. Securities and Exchange Commission: $5 million cash at closing (about ¥35.8 million), with Cycleo's shareholders eligible for up to $16 million more (~¥115 million) based on four-year performance milestones. The deal was funded from existing international cash reserves, and Semtech explicitly stated it would have "no material impact" on earnings that fiscal year.
In other words: the technology that would underpin a global ecosystem was acquired for a ceiling of about $21 million. For comparison, Semtech's 2023 acquisition of Sierra Wireless cost $1.2 billion — 57 times more.
Detail two: what was bought was not a product. It was a patent.
Cycleo was not a chip company; it was an IP company. It held the core patents on Chirp Spread Spectrum (CSS) modulation — patents formally granted two years after the acquisition, in 2014. The silicon implementations (the SX1272/SX1276 end-device transceivers, the SX1301 gateway concentrator) were taped out by Semtech with its own money after the deal. LoRa's journey from "a French patent" to "a chip you can buy" involved three years of Semtech's analog and mixed-signal engineering.
This detail is often overlooked, but it determined everything that followed: LoRa's value never lay in the patent alone, but in the combination of patent, silicon capability, and ecosystem operations. A patent without fabrication makes no business; a chip without a standard makes no ecosystem.
Detail three: the open ecosystem was not the plan in 2012. It was a forced pivot.
From 2012 to 2017, LoRa chips came from exactly one supplier: Semtech. That single-source model hit a ceiling around 2017 — large customers (Chinese metering and municipal projects especially) would not accept single-point supply risk, while the competing NB-IoT standard had an entire cellular industry pushing behind it. In September 2018, Alibaba Cloud IoT signed a LoRa IP licensing agreement with Semtech and, together with ASR Microelectronics, released the first licensed chip, the ASR6501. That is the context in which the pivot happened.
From that moment, LoRa stopped being "a Semtech product" and became "an IP licensing regime led by Semtech." It is the single most important turn in the ecosystem's evolution.
The most common mistake in discussing LoRa's openness is conflating two layers.
LoRa: the physical layer, proprietary. It refers to the CSS modulation itself — symbols represented by linear frequency-swept chirps, demodulated roughly 20 dB below the noise floor, with receiver sensitivity around -137 dBm and link budgets around 155 dB. The core patent family belongs to Semtech; third-party analyses generally estimate that the CSS modulation patents extend into the early 2030s. Any company that wants to make a LoRa chip either buys Semtech silicon or obtains a Semtech license. There is no third path.
LoRaWAN: the network protocol layer, open. Media access, network architecture, addressing, security, and regional channel plans are all published by the non-profit LoRa Alliance as open specifications that anyone may implement freely. In 2021, LoRaWAN was formally adopted by the International Telecommunication Union as an international standard (ITU-T Y.4480). By 2026, the LoRa Alliance counts more than 500 member companies, and more than 170 mobile operators have adopted LoRaWAN.
This two-layer structure — proprietary PHY collecting rent below, open protocol amplifying demand above — is Semtech's most important strategic design of the past fourteen years. Its cleverness:
The licensing system between the two layers has three tiers:
| Licensing tier | What it means | Representative licensees | Freedom |
|---|---|---|---|
| Chip purchase | Buy Semtech chips and build modules | The vast majority of module makers (RAK, Ai-Thinker, Hoperf, etc.) | Lowest: RF core fully dependent on the original supplier |
| Wafer / SiP license | Take Semtech wafers, package and integrate with your own MCU | Nationstech, AcSiP, SENSORO | Medium: product form is yours; the PHY core is still Semtech's |
| IP license | Design your own chips on top of LoRa IP | STMicroelectronics (STM32WL), Alibaba Cloud IoT with ASR (ASR6601) | Highest: only two such licensees worldwide |
This table answers a frequently asked question: "Are domestic Chinese LoRa chips truly domestic?" The answer is layered: chip design, packaging, modules, and applications are highly localized, but the root of the physical layer IP remains Semtech's — and that will not change before the CSS patent family expires in the early 2030s.
Laid out chronologically, the ecosystem's logic becomes much clearer.
| Time | Event | Significance |
|---|---|---|
| 2012-03 | Semtech acquires Cycleo for up to $21M | LoRa IP is born |
| 2014-2015 | SX127x / SX1301 ship; LoRa Alliance founded, LoRaMAC renamed LoRaWAN | From patent to product; from private protocol to open standard |
| 2018-09 | Alibaba Cloud IoT obtains IP license; ASR6501 released | Single-supplier era ends; licensing regime begins |
| 2021 | LoRaWAN becomes an ITU standard; 500M+ LoRa end devices, 6.9M gateways deployed | Ecosystem scale crosses critical mass |
| 2022-2024 | $1.2B Sierra Wireless acquisition; net leverage hits 9.0x; Hong Hou becomes CEO in June 2024 | The monopolist's own survival crisis |
| 2025 | Gen 4 platform (LoRa Plus, LR2021) launched; R&D spending +20% | From single protocol to multi-protocol single SKU |
| 2026 | 125M LoRaWAN devices; Amazon Ring's LoRa sensors ship; LoRa quarterly revenue $58M (+58%) | Consumer inflection + financial inflection |
Of the five turning points, the last three deserve elaboration.
The 2024 leadership change. Semtech's board moved decisively: on June 6, 2024, Dr. Hong Hou — a veteran of Intel and Emcore — took over as President and CEO. His mandate was explicit: deleverage, cut low-margin businesses, and concentrate R&D on high-value silicon. Over the following eighteen months, Semtech executed a textbook slim-down: an equity offering in December 2024 raised roughly $661 million for debt repayment; low-margin legacy hardware modules were divested in early 2025; and in 2026 the cellular module business was sold to Compal for approximately $62 million. Net leverage fell from 9.0x to 1.1x, and the divestiture is expected to structurally improve gross margin by more than 500 basis points.
The 2025 product pivot. The Gen 4 platform (marketed as LoRa Plus) changed the pitch: a single chip now supports LoRa alongside Z-Wave, Zigbee, Thread, and Matter, with data rates up to 2.6 Mbps, integrated TinyML edge AI, and a royalty-free SDK. Semtech's own positioning calls it "the Swiss Army knife of IoT." The backdrop: smart home and security markets demand a single SKU covering multiple protocols — an implicit admission that LoRa could not conquer the world on "long range, low power" alone.
The 2026 consumer inflection. Amazon and Ring launched a line of LoRa-based sensors running on Amazon Sidewalk — March 2026 in the U.S., followed by Canada and Mexico, with Europe, Australia, and Japan planned. LoRa entered a consumer market measured in tens of millions of units. Semtech management acknowledged this direction had a "false start several years ago"; the difference this time is Amazon's distribution.
Setting narrative aside, the numbers.
Revenue. For Semtech's fiscal 2026 (ended January 2026), LoRa-enabled revenue was $156 million, up 34% year over year. It accelerated through calendar 2026: in the quarter ended July 2026, LoRa revenue hit a record $58 million — up 31% sequentially and 58% year over year — with management guiding the next quarter to roughly +65%. LoRa now contributes about one-sixth of Semtech's revenue, but it is the best-margin piece of the company's mix.
Growth expectations. Semtech's stated long-term growth rate for LoRa is approximately 20% per year, with quarterly revenue in a $35-45 million band. That band reflects project-driven lumpiness rather than a ceiling — 2026 has run consistently above it, and management has publicly said it now expects growth "above the prior 20% benchmark."
Three demand pillars. In Semtech's own framing, LoRa now has three independent growth curves:
The geographic shift. One data point matters enormously for the Chinese market: Semtech's wireless product-line leadership stated publicly at The Things Conference 2025 that "China has re-emerged as a major growth engine for LoRa, with ecosystem players returning after several years of prioritizing NB-IoT," driven by the open ecosystem, deployment flexibility, and cost. This matches domestic reality: Chinese LoRa module shipments grew from roughly 110 million units in 2021 to nearly 290 million in 2023. The old judgment that "LoRa will be replaced by NB-IoT" has, at least in private-network scenarios in 2026, not materialized.
Issue #16's analysis of the platform retirement wave kept returning to one question: "Will this still exist in three years?" The same question applies to the chip layer — and Semtech's story supplies a fuller specimen: a monopolist is not automatically a survivor.
Semtech's 2023-2024 crisis was a triple stack:
For LoRa ecosystem participants, the meaning of that eighteen months is this: the entity holding your physical-layer IP has its own survival volatility. Had the worst case materialized — Semtech broken up, the LoRa patents landing in private-equity hands — the entire ecosystem's licensing terms, chip supply, and roadmap would have been repriced overnight. That is not a hypothetical risk: in the same period, ThingWorx was sold to private equity and promptly raised prices 20% (covered in Issue #16).
Semtech survived, and is in decent shape. But the crisis left a permanent awareness across the ecosystem: structural single-upstream risk must be hedged by the ecosystem's own open layers. Which leads to the next section.
Pulling the lens back from Semtech to the companies building products: within this "proprietary PHY + open protocol" structure, the implications differ by layer.
On the terminal side, choice has been broad since 2018: Semtech originals (SX126x/LR11xx), STMicroelectronics (STM32WL SoCs integrating MCU and LoRa radio on one die), ASR (ASR6601), plus SiP options from Nationstech and AcSiP.
The physical-layer lock-in certainly exists — every route ultimately pays Semtech a royalty. But it has one defining property: it is computable and transparent. The royalty is embedded in chip prices that are publicly comparable, and the LoRaWAN protocol layer is fully open — meaning you can swap physical-layer suppliers without touching a single line of your networking, data, or application code. Compared with the platform lock-in of Issue #16 (object models, rule logic, historical data formats — all proprietary), physical-layer lock-in is the mildest of the four kinds.
The gateway side is special: concentrator chips (SX1301/SX1302/SX1303) currently have no licensed alternative — Semtech is the sole source. But gateway differentiation was never about the chip — everyone can buy the chip. The difference is the entire software stack above it: backhaul protocols (UDP/MQTT/Basic Station), network-server integration (ChirpStack/TTN/self-hosted NS), regional compliance (CN470 channel masks), OTA mechanisms, and operations tooling.
That is precisely ManThink's product logic for the GD6: the SX1301/SX1302 concentrator is the industry's common substrate; the ESP32-S3 host and every line of firmware above it are open source; all three backhaul protocols are supported; and it connects to ChirpStack, TTN, or ThinkLink — any network server. Between a proprietary physical layer below and an open protocol stack above, a gateway vendor's value is to make the open segment as real as possible — so that the only thing a user is locked into is one irreplaceable chip, not an unportable software stack.
LoRaWAN's open specification is what allows ThinkLink, ChirpStack, and The Things Stack to exist legally and compete with each other — a direct fruit of protocol-layer openness. But platforms must be clear about the boundary: what can be done on LoRaWAN is bounded by the spec, while the pace of physical-layer evolution (new bands, new modulation modes like LR-FHSS, dual-band chips) is set by Semtech's silicon roadmap. Platforms need the agility to track new PHY capabilities quickly, or their position will be squeezed by the vendor's own cloud (LoRa Cloud).
Across the three layers, the "Semtech dependence" risk can be assessed structurally:
| Dependency type | Risk level | Hedge |
|---|---|---|
| PHY chip supply | Medium (single source, but Semtech's balance sheet is repaired) | Multi-supplier designs on the device side; inventory strategy on the gateway side |
| Royalty changes | Low (the licensing regime has been stable for 8 years; price hikes would push the ecosystem away) | The open protocol layer keeps an alternative path alive long-term |
| Change of patent ownership | Low-to-medium (net leverage 1.1x makes a hostile breakup far less likely) | Watch the CSS patent family's early-2030s expiry |
| Technology pivots | Medium (Gen 4's multi-protocol turn may divert resources) | Standard LoRaWAN deployments are unaffected |
Only one item genuinely belongs on a long-term watchlist: the CSS core patent family expiring in the early 2030s. On that day, the barrier to entering the LoRa physical layer shifts from "obtain a license" to "know how to build radios" — and unlicensed domestic alternatives and open implementations become materially possible. That is the single biggest variable in this ecosystem's next decade.
Misconception one: LoRa is an open technology.
No. LoRaWAN is an open specification; LoRa (the physical-layer modulation) is Semtech's proprietary, patented technology. Every claim of "open-source LoRa" applies only to the protocol layer and software stacks — never to the RF core.
Misconception two: Semtech just collects patent rent.
The patent is only the ticket. Semtech reinvests roughly 15% of revenue into R&D; the Gen 4 platform's multi-protocol fusion, TinyML integration, and dual-band design represent fourteen years of continuous investment post-acquisition. The patent moat protects pricing power; product iteration protects market share. Lose either, and the ecosystem flows to NB-IoT.
Misconception three: domestic LoRa chips mean escaping Semtech.
ASR's ASR6601 is a product of IP licensing; Nationstech and AcSiP products come from wafer/SiP licensing. What has been localized is chip design, packaging, and modules — the root of the PHY IP is still Semtech's. True "de-Semtech-ing" requires either patent expiry or a technology pivot. Those are two entirely different propositions.
Misconception four: Semtech is a monopolist, therefore it is safe.
The 2024 debt crisis just proved the opposite. Monopolists carry survival risk too, and their risk transmits to the whole ecosystem through supply, licensing terms, and roadmaps. Evaluate supply chains on financial structure, not market position.
Misconception five: LoRa and NB-IoT are locked in a winner-take-all war.
The 2026 reality is coexistence with a stable boundary: NB-IoT dominates carrier-network scenarios; LoRa dominates private networks. China's "return" to LoRa actually confirms that the boundary has stabilized. Semtech attributes the return to the open ecosystem, flexibility, and cost — fundamentally the private-network demand for "not being locked to a carrier," which does not conflict with NB-IoT's strengths at all.
Four sentences to close fourteen years.
First: a $21 million acquisition grew into a 125-million-device ecosystem not because of the patent, but because of the three-layer combination of patent, silicon, and open protocol. IP licensing alone would have made Cycleo a footnote; chip sales alone would never have made Semtech a standard-setter. Holding the physical layer private while opening the protocol layer outward is this ecosystem's core structural design.
Second: "openness" must be read in layers. LoRaWAN is an ITU international standard anyone may implement; LoRa's physical-layer patents extend into the early 2030s, and every chip route ultimately converges on Semtech. Without separating the two layers, you cannot assess your real dependence.
Third: monopolists carry survival risk too — and it just happened. The 9.0x-to-1.1x leverage rollercoaster reminds the ecosystem: upstream-risk assessment must look at financial structure and technology roadmap, not market position. Issue #16 asked "will the platform still be here in three years"; this issue asks the same of the chip layer.
Fourth: the early-2030s patent expiry is the biggest variable of the next decade. After it, the PHY entry barrier shifts from "obtain a license" to "know how to build radios." For device makers it means more suppliers; for Semtech it means moving value from the physical layer into software before then — via Gen 4, LoRa Plus, and ecosystem stickiness. That is visibly already the plan.
For companies building products, this article's single practical takeaway: of everything you depend on in LoRa, only the physical layer is irreplaceable; therefore everything else — firmware, protocol stack, platform, data formats — belongs on the open side. You cannot choose away the proprietary upstream; you can choose the middle layers. Making every choosable layer open is the cheapest hedge against single-upstream risk.
| Term | Description |
|---|---|
| LoRa | Long-range wireless physical-layer modulation based on Chirp Spread Spectrum (CSS); invented by Cycleo, patented by Semtech |
| LoRaWAN | The open network protocol specification above LoRa, maintained by the LoRa Alliance; ITU-T Y.4480 international standard |
| CSS | Chirp Spread Spectrum — the linear frequency-swept modulation underlying LoRa's physical layer |
| Cycleo | Grenoble-based French startup, inventor of LoRa; acquired by Semtech in 2012 |
| IP license | The highest licensing tier, permitting chip design on top of LoRa IP; held only by STMicroelectronics and Alibaba Cloud IoT (with ASR) |
| Wafer/SiP license | Licensing model providing LoRa wafers for packaging and integration into licensee-branded chips |
| LoRa Plus / Gen 4 | Semtech's fourth-generation platform fusing LoRa with Z-Wave/Zigbee/Thread/Matter in a single SKU, with edge AI |
| Amazon Sidewalk | Amazon's consumer network built partly on LoRa; Ring's sensors run on it |
| LR-FHSS | LoRa Long-Range Frequency Hopping Spread Spectrum, used for satellite (NTN) connectivity |
| ITU-T Y.4480 | The ITU international standard designation for LoRaWAN (adopted 2021) |