Lattice Semiconductor posted record Q2 2026 revenue of $201M, up 62% YoY, with its stock surging 50% as mid-range FPGAs (MachXO3, iCE40 UltraPlus) face 52-week lead times driven by AI datacenter and industrial test equipment demand. This report maps the supply picture, catalog-grounded part options, and buyer action plan.
Table of Contents (6 sections)
By Procurement Priya · icboms supply-chain desk · data through September 11, 2026
A semiconductor test equipment maker in South Korea cannot get the MachXO3 family FPGA it needs to bring its next-generation ATE platform online. Lead times are stretched past 52 weeks at franchised distributors. The equivalent Intel Cyclone IV E part is also on allocation. The buyer's engineering team has two choices: qualify a fallback now, or watch an NPI deadline slip by a quarter. This is the mid-range FPGA squeeze of Q3 2026, and it is being driven by a convergence of AI datacenter build-out and industrial automation demand that has outpaced every major supplier's capacity planning.
Lattice Semiconductor — the world's largest low-to-mid-range FPGA supplier — reported record second-quarter 2026 revenue of $201 million on September 8, 2026, representing 62% year-over-year growth and pushing the company's annualized revenue run rate past $1 billion following the closing of its AMI acquisition. Lattice's first-quarter 2026 revenue had already grown 42% year over year. The company's stock has surged approximately 50% in 2026 as investors price in the AI datacenter opportunity. This is not a niche story. Lattice makes the FPGAs that go into industrial protocols, edge inference modules, display bridges, and timing controllers — the unglamorous parts that every system actually needs. When Lattice is sold out at 52-week lead times, it is a supply chain event that touches every industrial OEM.
This report covers the live mid-range FPGA supply picture for Lattice MachXO3 and iCE40 UltraPlus families, the Intel Cyclone IV E alternative, what these parts actually do in AI datacenter and industrial systems, and the concrete steps a buyer should take today to avoid an NPI roadblock.
What the numbers say: Lattice's record quarter and the supply chain consequence
Three data points anchor this squeeze, all independently verifiable and dated.
Lattice Semiconductor reported record second-quarter 2026 revenue of $201 million on September 8, 2026, a 62% increase year over year, with management citing strong demand across communications, compute, and industrial end markets. The company's annualized revenue run rate has crossed $1 billion following the formal closing of its AMI acquisition, which adds new AI datacenter board management and security FPGA capabilities to Lattice's portfolio. The Futurum Group noted in its Q4 2025 earnings analysis that Lattice's AI datacenter-adjacent revenue was growing at 85% year over year, a rate that has continued into 2026.
Lattice's first-quarter 2026 revenue had already grown 42% year over year, which means the Q2 beat represents an acceleration, not a one-time event. The company guided Q3 conservatively but acknowledged that lead times on certain families remain extended. TIKR.com reported in September 2026 that Lattice stock had surged approximately 50% year-to-date, reflecting investor confidence in the AI datacenter tailwind.
The downstream consequence arrived in South Korea's semiconductor test equipment sector, where multiple equipment makers reported in September 2026 that FPGA lead times had stretched to 52 weeks — effectively a full-year wait for a critical programmable logic component. Semiconductor test equipment makers rely heavily on mid-range FPGAs for the pattern generation, protocol bridging, and timing control functions inside ATE platforms. When a $50 FPGA is the gate for a $5 million test system, procurement desks pay attention.
What this means at the desk level: the MachXO3LF family (which includes the LCMXO3LF-9400C-6BG256C) is experiencing allocation at major franchised distributors. The iCE40 UltraPlus family (including the ICE40UP5K-SG48I) is flowing but with lead times extended from the 8-to-12-week baseline to the 20-to-30-week range. Intel Cyclone IV E (Altera) alternatives are similarly tight, as the broader FPGA supply pool has been drawn down by AI datacenter infrastructure orders.
Why mid-range FPGAs are the sweet spot for AI datacenter and industrial systems
Mid-range FPGAs — typically defined as devices with 1,000 to 100,000 logic cells — occupy a specific design niche that AI datacenter and industrial automation providers cannot easily replace with higher-end devices. An AMD/Xilinx Artix-7 or Kintex-UltraScale part running $500-plus is overkill for a 10GbE timing bridge or a display serialiser. A $3 CPLD lacks the logic density for a machine vision preprocessing pipeline. The mid-range FPGA fills that gap, and it is the segment where Lattice competes most effectively against Intel (Altera) and, increasingly, against internal ASIC designs.
In AI datacenter infrastructure, mid-range FPGAs handle several non-compute functions that are critical to system reliability: BMC (baseboard management controller) assistance, PCIe retimer bridging, timing synchronization across accelerator cards, and security policy enforcement. As AI server clusters scale from hundreds to thousands of nodes, the aggregate demand for these bridge FPGAs grows linearly with cluster size — it is not a function of compute performance but of system management complexity. Every additional GPU node requires its own BMC assistance FPGA.
In industrial automation, the MachXO3 family is widely used for motor control bridging, industrial Ethernet protocol termination (PROFINET, EtherCAT, EtherNet/IP), and safety circuit monitoring. These are not "exciting" applications in the AI sense, but they are high-volume, long-lifecycle designs that expect 10-year part availability — which means buyers are deeply sensitive to EOL risk and second-source qualification costs.
The structural risk: as AI datacenter build-out accelerates, the "borrowed" FPGA capacity that industrial buyers have been relying on is being reclaimed. The Lattice AMI acquisition specifically targets AI datacenter board management applications — a signal that Lattice itself expects AI datacenter demand to continue drawing capacity away from the industrial base.
The catalog: three confirmed parts procurement desks should know right now
All three parts below are confirmed in the icboms catalog with active product status and full specifications on file. These are the reference parts our sourcing desk uses for RFQ on these families.
The Lattice MachXO3LF is the highest-logic-density option in the catalog for this family. The reference part is the LCMXO3LF-9400C-6BG256C — Lattice, MachXO3LF series, 206 user I/O, 256-CABGA package, 2.375V supply voltage, 442 Kbit embedded block RAM, active status. This part targets industrial bridge, protocol conversion, and mid-range embedded vision applications. The CABGA package is surface-mount, and the 2.375V supply voltage is standard for legacy industrial power rails. Lead times on this family at franchised channels are in the 30-to-40-week range as of September 2026, up from the 12-to-16-week baseline in 2024.
The Lattice iCE40 UltraPlus is the lowest-power option for battery-backed and edge inference applications. The reference part is the ICE40UP5K-SG48I — Lattice, iCE40 UltraPlus series, 39 user I/O, 48-QFN package, 1.14V supply voltage, 1.17 Mbit block RAM, active status. This is an ultra-low-power FPGA designed for mobile, battery-powered, and edge AI applications where power budget is the primary constraint. The 48-QFN package is compact (7mm × 7mm) and suitable for space-constrained designs. Lead times on this family are in the 20-to-30-week range, up from the 8-to-12-week baseline.
The Intel Cyclone IV E is the broadest industrial FPGA family and the most common Xilinx/Intel Altera alternative. The reference part is the EP4CE6E22C8N — Intel (formerly Altera), Cyclone IV E series, 91 user I/O, 144-EQFP package, 1.15V supply voltage, 276 Kbit embedded memory, active status. Cyclone IV E is the default drop-in recommendation when a MachXO3LF is not available, and engineering teams frequently qualify both families in the same BOM as a dual-source strategy. The EQFP package format is tray-based, which affects MOQ and reel handling.
How to buy in this market: seven actions for Q4 2026 and Q1 2027
First, audit every mid-range FPGA in your active BOMs. List every Lattice MachXO3, iCE40, or Intel Cyclone IV/III/II part with package, density, and current lead time. Flag anything with a lead time above 16 weeks. That audit is the starting point for every conversation with your supplier.
Second, qualify an alternative family now — before you need it. The MachXO3LF and iCE40 UltraPlus families have pin-compatible and functionally-equivalent alternates in the Intel Cyclone IV E (Altera) family and, for lower-density designs, the Lattice LCMXO2 family. Qualification takes 8 to 12 weeks in a well-run engineering team. Waiting until the primary source is on allocation means that qualification run happens under schedule pressure, which increases the risk of a field failure.
Third, pre-book on confirmed 2027 builds. The Lattice AMI acquisition formally closed in Q3 2026, bringing Lattice's annualized revenue run rate to $1 billion and giving the company capital to invest in capacity — but new capacity takes 18 to 24 months to come online. For buyers with Q1 2027 or Q2 2027 production builds, now is the time to place pre-book orders on the MachXO3LF family. The MOQ on CABGA parts is typically 1 tray (minimum 84 pieces for a 144-EQFP; verify with your distributor), and the Q4 2026 window for locking allocation is closing.
Fourth, watch the MOQ conversation carefully. Our catalog reference on the LCMXO3LF-9400C-6BG256C is tray MOQ, which means the minimum order quantity is tied to the manufacturer's pack count. For prototype builds or low-volume production, partial-reel or cut-tape availability varies by distributor and changes week to week. Talk to us about breakout options for smaller quantities — we carry selective allocation on Lattice MachXO3LF and iCE40 UltraPlus families and can usually find flexibility for confirmed engineering builds.
Fifth, do not assume last-time-buy language means end-of-life. Lattice has not issued an EOL notice on the MachXO3LF family as of September 2026, and the AMI acquisition is expanding Lattice's FPGA roadmap rather than consolidating it. What the 52-week lead time signal means is allocation, not retirement. A procurement desk that treats this as an EOL situation will over-order and create a secondary inventory problem in 2027.
Sixth, verify every lot when parts arrive. Mid-range FPGAs are among the most counterfeited logic families in the spot market, and demand pressure increases counterfeit risk. Date-code consistency, top-side marking font and depth, and lead finish appearance should be checked against the Lattice datasheet. If the lot does not match manufacturer specs, quarantine it and contact the distributor immediately.
Seventh, talk to our sourcing desk before you commit to a contract. We cover Lattice, Intel/Altera, and AMD/Xilinx mid-range FPGA families with dual-channel franchise and spot allocation. Our Shenzhen sourcing desk can quote tray, reel, and cut-tape quantities on MachXO3LF, iCE40 UltraPlus, and Cyclone IV E families for confirmed production and prototype builds.
Data notes
Data cutoff: September 11, 2026. Market figures in this article come from Lattice Semiconductor Q2 2026 earnings results (Business Wire, September 8, 2026, $201M record revenue, 62% YoY), Lattice Semiconductor Q1 2026 earnings results (Business Wire, April 2026, 42% YoY growth), Lattice Semiconductor Q4 2025 earnings analysis (The Futurum Group, January 2026, AI datacenter +85% YoY), TIKR.com (September 2026, Lattice stock +50% in 2026), and South Korea semiconductor test equipment FPGA supply reports (finance.biggo.com, thelec.net, September 2026). Catalog parts referenced (LCMXO3LF-9400C-6BG256C, ICE40UP5K-SG48I, EP4CE6E22C8N) are confirmed in the icboms catalog at the time of publication with active product status; lead times and MOQ are quoted at desk level and are subject to change without notice. Treat all figures as directional planning data, not quotes. Verify lead times, date codes, and product status with your distributor before issuing a production purchase order. ICBOMS is an independent semiconductor distributor sourcing through multiple authorized channels in Shenzhen, China; we are not an authorized Lattice Semiconductor, Intel (Altera), or AMD (Xilinx) distributor, we do not speak for the manufacturers, and we do not republish their product change notifications.
Contact
ICBOMS — independent semiconductor sourcing desk, Shenzhen, China. Brand coverage: Lattice Semiconductor, Intel (Altera), AMD (Xilinx), and adjacent FPGA and CPLD families. Services: catalog pricing, lot verification, BOM kitting, dual-source qualification support, pre-book for confirmed production builds, partial-reel breakout. Applicable buyers: hardware engineers, EMS procurement teams, ODM sourcing desks, semiconductor ATE test equipment buyers, industrial automation PLC and HMI purchasing teams, AI datacenter infrastructure hardware teams.