We use cookies to understand how our site is used so we can improve it. Necessary cookies are always on. You can accept all, reject all, or choose what to allow. Learn more
FPGA & Semiconductor DistributionThe channel is the moat, not the chip.
FPGA & Semiconductor Distribution
The channel is the moat, not the chip.
In 2024–25 the world bought somewhere between US$10 billion and US$14 billion of field-programmable gate arrays — a small, fast-growing corner of the semiconductor world where one vendor, AMD/Xilinx, holds more than half the market and the top five hold 80–86%. What makes the FPGA market structurally distinct is geography: the chips are designed in California but nearly half of demand has migrated to Asia-Pacific, and the silicon makers outsource the design-in support that converts a chip into a Korean OEM's product to an authorized regional channel. That channel — not the chip — is the scarce, defensible asset in any single country. This is a study of that market through Korea's only listed pure-play FPGA design-in distributor, Macus.
An FPGA on a circuit board. The chip is reprogrammable silicon designed in the US by AMD/Xilinx, Altera and Lattice; turning it into a Korean drone, base station or storage controller takes circuit design, firmware and field-engineering support the silicon makers do not supply at home. That support — the design-in channel — is where Macus lives, and where the market's defensible margin sits.
In 2024–25 the world bought somewhere between US$10 billion and US$14 billion of field-programmable gate arrays — reprogrammable logic chips that engineers configure after manufacture for 5G base stations, radar and avionics, data-center acceleration, industrial control and storage. The range is wide because the research houses disagree on scope and base year: Mordor Intelligence sizes the 2025 market at US$9.93 billion, MarketsandMarkets at US$11.73 billion, GMInsights at US$13.8 billion, Fortune Business Insights at US$13.92 billion. They converge on direction even where they diverge on level — a compound growth rate clustered around 9–12% a year, carrying the market toward roughly US$17–28 billion by 2030–32, with the most aggressive houses reaching US$37–41 billion by 2034–35.
What makes the FPGA market structurally distinct is not its size but its shape. One vendor, AMD/Xilinx, holds more than half of it; the top five hold 80–86%. The chips are designed in California, yet 44–55% of demand has migrated to Asia-Pacific. That gap between where FPGAs are designed and where they are consumed is the reason an entire channel exists: the silicon makers outsource the work of turning a generic chip into a specific Korean product — circuit design, firmware, field-engineering support — to authorized regional distributors. This is a study of that channel, its margin structure and the silicon cycle it rides, read through Korea's only listed pure-play FPGA design-in distributor, Macus, and of why the franchise to sell the chip, not the chip, is the defensible asset in any single geography.
Global FPGA market — a band, not a point
~$10–14bn
The 2024–25 FPGA TAM across eight independent houses (Mordor $9.93bn; MarketsandMarkets $11.73bn; GMInsights $13.8bn; Fortune Business Insights $13.92bn; Polaris $13.63bn 2024). Consensus CAGR ~9–12%; terminal ~$17–28bn by 2030–32, extreme houses to $37–41bn by 2034–35. The spread is scope-driven (some houses fold adaptive-SoC/Versal in) — preserved as a range, not reconciled. Sources: Mordor Intelligence; MarketsandMarkets; Fortune Business Insights; GMInsights; Polaris Market Research.
Two clocks: the chip and the channel that carries it
Macus does not live in the FPGA chip market. It lives one layer down, in semiconductor distribution — and the two grow at different speeds. The global semiconductor-distribution channel is sized at US$53.5 billion in 2025, rising to US$80 billion by 2035, a roughly 4.1% compound rate. The FPGA product itself compounds at around 10%, more than twice as fast. That divergence is the central structural finding of the whole market: as FPGA content value per unit rises faster than the distribution fee on it, value migrates within the channel — toward technical, design-in distribution and away from box-moving fulfillment. The forward thesis of the channel is compression at the bottom, premium at the top.
The premium is measurable, and it is large. A broadline fulfillment distributor — Avnet, with US$23.8 billion of FY2024 revenue, around 80% of it semiconductors — runs at a 3.6% operating margin. A Korean technical distributor like Uniquest, at ₩730 billion of revenue, runs near 5.0%. Macus, on ₩227.24 billion of FY2025 consolidated revenue, reports a 12.80% operating margin. For the same nominal function — moving a chip from a vendor to a customer — the design-in tier earns roughly 3.5 times the margin of broadline fulfillment. That gap is the entire economic case for the value-added segment, and the reason a small Korean firm can exist profitably in a market its global peers measure in tens of billions.
Distributor
Model
Revenue
OP margin
Avnet
Broadline / fulfillment
$23.8bn (FY24)
3.6%
Uniquest
Technical (auto/industrial)
₩730.0bn
~5.0%
Macus
Technical / design-in (FPGA)
₩227.24bn
12.80%
The design-in premium: operating margin by distribution model (latest fiscal year). Broadline players move scale and logistics; the technical / design-in tier embeds circuit design and field-engineering support into the customer's product. The ~3.5x margin gap between Macus and a broadline distributor is the entire economic case for the design-in segment. Sources: Avnet 10-K FY2024 (SEC); Uniquest FY2025 (FnGuide A077500; DART); Macus FY2025 consolidated (DART 20260318000498).
Broadline distribution is a scale-and-logistics business: warehouse, inventory, fulfillment, a margin in the low single digits. Design-in distribution is the opposite trade — narrow scope, engineering-heavy, embedded in the customer's product from planning to mass production, and a margin several times higher. The two compete on different axes and rarely for the same dollar.
What is pulling demand — and the bands at the end of the decade
The demand case underneath the FPGA market is broad and durable. AI inference is the newest driver: FPGAs win on latency and performance-per-watt against CPUs and GPUs for network, storage and edge acceleration, and AMD's data-center Versal Premium Gen2 samples in early 2026. The single largest application slice is still IT and telecom at around 35% of 2024 demand, sustained by programmable 5G base-station hardware. Aerospace and defense is among the fastest-growing segments — secure, adaptable, long-lifecycle silicon for radar and avionics, with a specifically Korean tailwind as Xilinx FPGAs serve as drone main-processors amid the build-out of Korean drone units. Underneath all of it sits sovereign-semiconductor policy: Korea's target to lift the non-memory share of output from 30% toward 50% by 2030 enlarges the localized demand pool the channel feeds.
Where it gets unreliable is the forecast, and the disagreement is the story rather than a footnote. Group the houses into a fan and they will not converge on a single curve: the conservative case (Mordor, US$9.93bn in 2025 at 9.35%) reaches about US$17.2 billion by 2031; the central cluster (MarketsandMarkets US$19.34bn by 2030, Grand View US$23.34bn by 2030, Fortune Business Insights US$27.51bn by 2032) sits in the low-to-high twenties; the aggressive houses (GMInsights, Polaris) carry the market to US$25.7–36.7 billion by 2031–34. Each house holds its own base year, terminal and rate, and they should not be reconciled. The honest answer is the band — roughly US$17–28 billion by 2030–32 — not a point inside it.
The Korean FPGA-channel demand cycle, read through MacusExhibit · Macus consolidated revenue (₩bn, bars) against operating margin (%, line), FY2021–FY2025. Cited as a demand proxy for the Korean FPGA channel, not as a company performance metric: the FY2023 −9% dip tracks the FPGA-vendor inventory correction (AMD Embedded −33% in CY2024) one period ahead, confirming the channel is a levered read on the silicon cycle. Revenue and operating profit are company-reported; OP margin re-computed (OP ÷ revenue). Source: DART 20260318000498 (consolidated).
Read that series as a market signal, not a scorecard. Macus's revenue ran from ₩86.9 billion in 2020 to ₩140.7 billion (2021), ₩193.8 billion (2022), then dipped to ₩175.4 billion in 2023 before recovering to ₩199.6 billion (2024) and a record ₩227.2 billion (2025). The 2023 dip is the tell: it falls one period ahead of the AMD Embedded segment's CY2024 trough, when that business — essentially the Xilinx FPGA line — fell 33% on inventory normalization. A regional channel absorbs the correction earlier, through sell-through to Korean OEMs, then recovers as vendor inventory normalizes. The first quarter of 2026 printed revenue down 16.6% against an unusually strong Q1-2025 — base-effect softness on the evidence, one data point, not yet a new down-cycle.
Forward FPGA market — the fan, not the curve
$17–28bn
Terminal global FPGA TAM by 2030–32 across the house fan: conservative ≈$17.2bn (Mordor, 2031, 9.35%); central $19.3–27.5bn (MarketsandMarkets 2030, Grand View 2030, Fortune BI 2032); aggressive to $25.7–36.7bn (GMInsights 2031; Polaris 2034), $41.1bn by 2035. Consensus CAGR ~9–12%. Each curve is a house's published base compounded at its published rate — the band is the forecast, not any single line. Sources: Mordor Intelligence; MarketsandMarkets; Grand View Research; Fortune Business Insights; GMInsights; Polaris Market Research.
The structural divide: a duopoly upstream, two axes downstream
Strip the market down and it organizes around two structures, one upstream and one down. Upstream is a duopoly-plus, re-cut in 2025. AMD/Xilinx holds more than 50% — roughly US$1 billion of revenue a quarter — after absorbing Xilinx in a ~US$49 billion deal in 2022. Altera is the clear number two, made independent again in 2025 when Intel sold 51% of it to Silver Lake at an US$8.75 billion valuation, repositioning as the largest pure-play FPGA company. Lattice, at 8.6% share, is the fast-rising mid-range insurgent with its Avant line; Microchip holds 6.2% in flash and antifuse for industrial and aerospace. The top five together hold 80–86%. With one vendor controlling more than half the market globally, the authorized-distributor franchise — not the chip — is the scarce, defensible asset in any single country.
The supply side after 2025 is two private-equity- or strategically-backed pure-plays — AMD/Xilinx above 50%, Altera/Silver Lake at number two — plus the mid-range insurgent Lattice. When the silicon is this concentrated, the geographic asset is the authorized channel, not the part number. Macus has held AMD/Xilinx authorization since 1997.
Downstream, the channel competes on two orthogonal axes that rarely trade against each other. The first is scale and logistics — the broadline distributors. WT Microelectronics, after acquiring Future Electronics, is number one by revenue; WPG Holdings runs around US$28–32 billion and about 30% of Asia; Arrow Electronics around US$27.8 billion; Avnet US$23.8 billion at a 3.6% margin. They dwarf Macus's roughly US$0.17bn (₩227bn) externally, but they move boxes. The second axis is technical, design-in distribution — circuit design and field-engineering support embedded in the customer's product from planning through mass production, at 5–13% operating margins. Macus sits at the high-margin, narrow-scope corner of that second axis. It does not compete with Avnet for the same dollar; its small scale is by design, and its margin is the evidence of where channel value concentrates.
Distributor
Revenue (USD)
Axis
OP margin
WPG Holdings
~$28–32bn
Broadline / scale
n/a
Arrow Electronics
~$27.8bn
Broadline / scale
~3–4%
Avnet
$23.8bn
Broadline / scale
3.6%
Uniquest
~$0.53bn
Technical / design-in
~5.0%
Macus
~$0.17bn
Design-in / FPGA
12.80%
The downstream channel mapped on its two axes: revenue scale (broadline) against operating margin (design-in). The two clusters — high-scale/low-margin and low-scale/high-margin — describe two different businesses; they do not trade against each other. Won converted at ~₩1,375/USD for a common axis; WT Microelectronics (#1) omitted as not on a comparable basis. Sources: SEC (Arrow, Avnet); WPG Holdings / MarketScreener; FnGuide A077500 (Uniquest); DART 20260318000498 (Macus).
The case in point: a pure FPGA design-in bet
Macus is worth studying because it has taken the design-in side of this divide all the way. It is a non-memory semiconductor solution company — the Korean authorized partner of AMD/Xilinx for FPGAs, MACOM for RF and Renesas for MCUs and analog — with no production facilities of its own. It describes itself explicitly as not a simple reseller: it joins the customer's design from product-planning and circuit design through to supply at mass production. Listed on KOSDAQ since 2007 and spun from Coacross in 2006, it earns its living entirely on the technical-distribution axis. The 96.25% of FY2025 consolidated revenue that is parent merchandise — ₩218.74 billion of FPGA, analog, RF and MCU components — is the company's participation in the non-memory market; a small storage-server line through Macus Systems and an investment line make up the rest.
One disclosure boundary matters for honest reading. Within that ₩218.74 billion merchandise line, the company reports no split between FPGA, analog, RF and MCU — its filing states the revenue carries no business-segment breakdown. So the precise FPGA share of the core line is not disclosed, and is not invented here. What is visible is the concentration: the storage-server line shrank from ₩14.9 billion in FY2024 to ₩5.0 billion in FY2025, leaving the consolidated story almost wholly the parent component-distribution business. The other visible fact is that Macus reports no long-term supply contracts — pricing is spot, per customer order — which is exactly why the channel transmits the FPGA demand cycle so directly, and why the 2023 dip tracked the silicon trough.
The design-in moat made concrete. Macus provides high-difficulty technical support from a customer's early-development stage — circuit design, firmware, field-application engineering — embedding itself in the product before mass production. That engineering content, not the chip itself, is what a broadline distributor cannot replicate and what carries the ~12.8% margin against a broadline's ~3.6%.
The comparison that frames Macus's position is with Uniquest, the other Korean listed technical distributor. Macus is roughly a third of Uniquest's revenue but carries more than twice its operating margin — the FPGA design-in niche is structurally richer than broad auto and industrial distribution, even within the value-added tier. Against a broadline like Avnet, the differential is about 3.5 times. It is worth flagging the limit of the headline claim: no published FPGA-distribution market share exists for Korea, and the description of Macus as the country's only listed pure-play FPGA distributor omits private and in-house import channels — it is company- and press-asserted, not a measured share.
“An authorized FPGA distributor is a franchise wearing the costume of a reseller. The decisive question in this market is never how advanced the chip is — it is who holds the right to design it into a customer's product in a given geography, and how richly that design-in work is paid against the box-moving alternative.”
— Nathan Research Group, Semiconductor Distribution Series N°01
Where the market goes from here
Project the FPGA market forward and two things hold at once. The chip compounds in the high single to low double digits, plausibly toward US$17–28 billion by the early 2030s, pulled by AI inference, 5G, defense and sovereign-localization demand. The distribution channel that carries it compounds far more slowly, at 4–5%, so the dollar growth and the margin both migrate to the design-in tier rather than to broadline fulfillment. The supply side stays concentrated — two pure-plays plus the Lattice insurgent — which keeps the authorized-distributor franchise scarce and geographic. The constraints on the channel are a normal silicon cycle (a second inventory correction would echo the FY2023/CY2024 trough) and the pace at which Korea's non-memory localization actually pulls design wins onshore.
The lesson generalizes past one company and past FPGAs. In any market where a concentrated set of upstream producers sells through regional intermediaries — and where the producer outsources the customer-specific engineering rather than performing it — the durable economics belong to whoever owns the franchise to do that engineering in a given geography, not to whoever moves the most volume. The FPGA channel is a vivid case: broadline distributors win scale and surrender margin; design-in distributors win margin and accept small scale, because the authorized franchise plus the engineering content is the moat. Which model compounds is decided in customers' design rooms and in the vendor's channel-authorization decisions, geography by geography — not in any single year's revenue line. Our full study of the FPGA market — the eight-house sizing band, the channel value-chain economics, the upstream concentration map and the forward fan — is available to download with this article.
Working With Nathan Research Group
Partner With Nathan Research Group
Public filings and market forecasts establish the shape of the FPGA channel — the eight-house TAM band, the vendor-concentration structure, the design-in margin premium visible in disclosed accounts. What they cannot show is how an authorized-distributor franchise actually renews, how design-in wins convert to repeat silicon orders across a customer's product life, and where inside the undisclosed merchandise line the FPGA, analog, RF and MCU dollars really sit. That intelligence lives with the field-application engineers, fabless customers and channel operators who design these chips into Korean hardware — and reaching them, compliantly, is what we do.
Korea’s first dedicated expert network — Seoul, since 2013
Who we put in the room
FPGA design-in / field-application engineer
Engineers who have placed AMD/Xilinx, Altera and Lattice silicon into Korean defense, 5G and storage hardware — on what makes an authorized channel sticky and what a design win is actually worth over a product's life.
Korean fabless / non-memory customer
The OEMs and ODMs the channel feeds. They read how the K-Semiconductor non-memory 30%→50% push and the drone/defense build-out translate into real component demand.
Semiconductor-distribution channel structure
Operators who know the broadline-versus-design-in divide first-hand — why a ~3.6% broadline margin and a ~12.8% design-in margin describe two different businesses, not one.
Insiders to the AMD/Xilinx and Altera channel programs read how authorized status is granted, renewed and lost — the asset at the center of the geographic moat.
Korea semiconductor policy & defense electronics
The localization tailwind is policy-driven. Specialists pressure-test how sovereign non-memory targets and drone-unit creation pull FPGA design wins onshore.
Enterprise storage / Pure Storage channel
Macus's smaller storage-server line participates in a separate channel. Operators read where that segment is heading after it shrank from ₩14.9bn to ₩5.0bn in a year.
How an engagement works
1Scope
We turn your thesis into a precise expert profile and question set, mapped to the decisions you need to close.
2Source & vet
We screen and compliance-clear each expert — relevance, recency, and the absence of conflicts — before any call.
3Convene & synthesize
We arrange interviews on your timeline and, where useful, deliver written synthesis tied back to your questions.
If your team is evaluating Macus, the FPGA distribution channel, or the broader semiconductor sector, tell us the decision you’re trying to make.
A direct reply from a partner, not an intake form.
Based solely on public sources. Nathan Research Group does not request or facilitate material non-public information, and runs every engagement through a documented compliance protocol.
