Theme: Compute (FPGA) | Price: $146.30 | Market Cap: ~$20.0B | Date of note: 2026-07

1. What it is

Lattice Semiconductor designs low-power, small-footprint FPGAs (field-programmable gate arrays) — programmable logic chips used for control, connectivity, and now increasingly edge-AI inference tasks in industrial, automotive, communications, and consumer systems. Product families include Certus-NX, ECP, Mach, iCE40, and CrossLink, plus video-connectivity ASSPs and an IP-licensing business. Lattice is a niche specialist, not a scale player: at ~$20B market cap and ~$500M in annual revenue, it is an order of magnitude smaller than AMD (which owns Xilinx, the FPGA high-density/data-center leader) and Intel (which owns Altera/Agilex, the other FPGA giant). Lattice deliberately competes in the low-power, low-density segment those two largely don't optimize for, rather than head-to-head in large data-center/networking FPGAs.

2. The exponential thesis

The quant screen's +45% forward revenue growth and +34% acceleration need context: they are being measured off a cyclically depressed base (see Section 3), not off a steady-state run-rate. That said, the underlying secular drivers cited by the screen are real and identifiable in the business:

  • Edge AI inference: Lattice has positioned its newer Avant platform (its next-generation FPGA architecture, launched to succeed the aging ECP5/CrossLink families) around low-power AI inference at the edge — cameras, sensors, industrial automation — where full GPU/data-center compute is impractical on power/cost/latency grounds. This is a genuine new product cycle, not just a re-hash of legacy parts.
  • Industrial/automotive/communications content growth: the profile explicitly cites OEM customers concentrated in communications and computing, consumer electronics, and industrial/automotive — end markets where FPGA content per unit (sensor fusion, connectivity bridging, functional safety logic) has been rising structurally.
  • Analyst estimates support a multi-year ramp, not a one-quarter pop: consensus revenue estimates show ~$521M (FY2026) → ~$760M (FY2027) → ~$934M (FY2028) → ~$1.13B (FY2029), i.e., a step-up that assumes Avant-driven design wins convert into shipping revenue over several years, roughly doubling revenue from the FY2025 trough by FY2028.

Verification caveat: none of the local FMP data sources (profile, income statements) carry segment-level revenue breakouts (e.g., % of revenue from Avant vs. legacy, or industrial vs. comms vs. auto). The "Avant platform" and "edge AI" framing is the well-documented company narrative and is consistent with the product-family and end-market disclosures in the profile, but this note cannot independently verify the magnitude of AI-driven content growth from the data on hand — that would require the 10-K/10-Q MD&A segment commentary, which is not in the local dataset.

3. Financial health

This is the single most important interpretive issue in the name: Lattice just went through a real, sharp cyclical semiconductor downturn, and the "+45% forward growth" figure is measured off a depressed trough, not a stable base.

Revenue trajectory (annual, FMP income_annual): - FY2022: $660.4M (net income $178.9M, EPS $1.30) - FY2023 (cycle peak): $737.2M, gross margin ~69.8%, net income $259.1M, EPS $1.88 - FY2024 (downturn): $509.4M (-30.9% YoY), net income $61.1M, EPS $0.44 - FY2025 (trough): $523.3M (+2.7% YoY — essentially flat, not a recovery yet on an annual basis), gross margin 68.2%, operating income just $15.3M (2.9% operating margin), net income only $3.1M, EPS $0.02

Quarterly cadence confirms the trough-and-recover pattern the prompt flagged: - Q1 FY2024: $140.8M → Q4 FY2024: $117.4M (bottom of the initial decline) - FY2025 quarters bounced along a low plateau: Q1 $120.2M, Q2 $124.0M, Q3 $133.3M, Q4 $145.8M (net loss of -$7.6M, EPS -$0.056 — a negative quarter) - Q1 FY2026 (most recent): $170.9M revenue, gross margin 68.8%, operating income $26.7M, net income $21.8M, EPS $0.16 — this is the strongest quarter since the FY2023 peak and the clearest sequential inflection in the dataset.

Interpretation: the +45% forward growth is a bounce off a cyclically depressed base, not yet proven durable share gain. Revenue is still ~29% below the FY2023 cycle peak ($737M) even after the Q1 FY2026 acceleration. The pattern (peak 2023 → nearly -31% decline in 2024 → flat/bottoming 2025 → sequential re-acceleration into 2026) is a classic semiconductor inventory-correction cycle (channel destocking following COVID-era overbuild), not new secular demand layering on top of a stable base. Some of the forward growth is real design-win conversion (Avant ramping), but a meaningful portion is simply comping against depressed 2024-2025 quarters. Treat the growth rate as directionally positive but mechanically inflated by base effects.

Balance sheet: confirms the historically low-debt/net-cash profile. - FY2025: cash $133.9M, total debt $78.2M, net debt -$55.7M (net cash); total equity $714.1M; current ratio 3.09x - FY2024: cash $136.3M, total debt only $15.3M, net debt -$121.0M (net cash) - Debt-to-equity is low (~11%) and interest coverage is not a concern (interest expense is negligible — $264K in FY2025 against $175M of operating cash flow)

Cash flow: durable and recovering faster than GAAP earnings. - FY2025 operating cash flow $175.1M, capex -$42.5M, free cash flow $132.6M (FCF margin ~25% of revenue) — notably, FCF held up much better than net income ($3.1M) because of $115.6M of stock-based compensation add-back and working-capital tailwinds. This is a real cash-generation business even in the trough year. - FY2024 FCF was $119.9M; FY2023 (peak) FCF was $249.5M. - The company continued to buy back stock through the downturn ($100M in FY2025, $67M in FY2024), which is a reasonable use of a net-cash balance sheet but also flatters per-share metrics somewhat during a period of depressed absolute earnings.

4. Competitive position / moat

Lattice's moat is a defensible niche, not scale leadership. It is the leading specialist in low-power, small-footprint, low-density FPGAs — a segment AMD (via Xilinx) and Intel (via Altera/Agilex) largely cede because their FPGA franchises are built around high-density, data-center/networking/aerospace-defense use cases with different power/cost/size tradeoffs. Lattice's advantage in its niche comes from: (1) years of accumulated IP and process optimization specifically for low-power operation, (2) a broad low/mid-density product portfolio (Certus-NX, ECP, Mach, iCE40, CrossLink) that covers many discrete design slots competitors don't prioritize, and (3) an IP-licensing/patent-monetization business that monetizes its technology stack beyond silicon sales.

The risk to this moat is scale asymmetry: AMD and Intel each have vastly larger R&D budgets, and if either decided to seriously contest the low-power segment (rather than staying focused on their core high-density franchises), Lattice's ~1,100 employees and sub-$1B revenue base would be structurally outgunned. To date this hasn't happened at scale, likely because the low-power segment is not large enough to be strategically interesting to either giant relative to their core businesses — but that is a competitive-forbearance dynamic, not a hard moat.

5. Valuation

The -15.76 PEG is broken and should be disregarded — flagging this explicitly as instructed. The mechanical cause is visible directly in the ratios_annual data: FY2025 priceToEarningsRatio is 3,495.6x (net income of only $3.1M against a market cap of ~$10.8-20B, depending on point-in-time price) and priceToEarningsGrowthRatio is -36.8x. A PEG ratio divides a trailing/forward P/E by an EPS growth rate; here, trailing EPS collapsed from $1.88 (FY2023) to $0.02 (FY2025), producing a deeply negative or meaningless EPS growth comparison and a triple-digit-to-quadruple-digit trailing P/E. Dividing a meaningless P/E by a meaningless (likely negative, base-effect-driven) growth rate produces a PEG that is pure statistical noise, not a signal about valuation richness. Do not use PEG for this name in its current earnings state.

Cleaner metrics, computed from key_metrics_annual/ratios_annual (FY2025, using enterprise value ~$10.7B per key_metrics — note this differs from the quote-page market cap of $20.0B, likely reflecting different price-date snapshots; directional conclusions below are robust to either): - EV/Sales (FY2025): ~20.5x — rich for a company growing revenue only ~2.7% YoY on a trailing basis, though more digestible if the forward $520M→$760M (FY2026→FY2027) ramp materializes. - EV/EBITDA (FY2025): ~191.7x — extremely elevated, but this denominator (EBITDA of only $56.0M) is itself trough/depressed, so the multiple is an artifact of the earnings trough, not necessarily "expensive" on a normalized basis. Using FY2023 peak EBITDA (~$255M) against a similar EV would imply a much more reasonable ~40-45x, and using FY2028 consensus EBITDA (~$237M) implies the market is pricing in a substantial recovery. - Price/FCF (FY2025): ~81x; FCF yield: ~1.2% — rich in absolute terms, though FCF ($132.6M) is a real, held-up-through-the-downturn number, which is more reassuring than the GAAP earnings picture. - Forward P/E using FY2026 consensus EPS ($1.05 avg): at $146.30/share, that's roughly ~139x forward P/E — still expensive, though it falls to roughly ~81x on FY2027 consensus EPS (~$1.81) and ~62x on FY2028 (~$2.36), i.e., the market is pricing a multi-year earnings recovery/growth curve, not current-year fundamentals.

Fair-value stance: rich, not cheap. Even setting aside the broken PEG, every clean multiple (EV/Sales ~20x, forward P/E in the 60-140x range depending on out-year assumed) prices in a fairly aggressive multi-year recovery and Avant-driven re-acceleration. The balance sheet and FCF generation are genuine quality supports, but the stock is not statistically cheap on any conventional metric — it is a "pay up for the recovery/re-rating story" situation, appropriate only for investors with conviction in the Avant ramp actually delivering the FY2027-2029 consensus numbers.

6. What the panel says

KB search for "LSCC" and "Lattice" returned a small, thin set of genuinely relevant hits — this is a real, not false-positive, dataset, but conviction is limited to one recurring source repeated across near-duplicate entries: - Three near-identical claims (jordi_visser / jordi_visser_ai / jordi_visser_m variants — likely the same underlying commentary re-ingested under different thinker-tags) tie Lattice together with Teradyne (and in one variant, ASML) as "analog-based semiconductor names [that] haven't participated yet and have a long road ahead," stance bullish, conviction 65-68, horizon "thesis," dated 2026-02-05. - This is a thin, single-source, rotation/relative-value call ("underparticipated names catching up"), not a fundamental deep-dive on Lattice specifically. It says nothing about Avant, edge AI, or the cyclical recovery directly — it is a generic "this basket hasn't run yet" thesis.

Data-quality flag: as anticipated, "Lattice" is a common English/generic word (crystal lattice, math lattice, lattice-based cryptography, etc.) and a naive search risks false positives. In this KB, the "Lattice" hits that surfaced were correctly scoped — all matches paired "Lattice" with "Lattice Semiconductor" explicitly or the ticker "LSCC," and no physics/crypto/generic-word false positives were found in this particular search. However, the volume of genuine evidence is very thin (3 near-duplicate mentions from what appears to be one commentator/source, not 38 independent claims). Note the quant profile's stated "panel conviction score: +18.8 (from 38 KB claims)" could not be reconciled against what this Grep search surfaced — only 3 genuinely on-topic claims were found, all from the same apparent source restated three ways. This is worth flagging upstream: either the 38-claim figure aggregates claims tagged to LSCC/Lattice under entity linkage not captured by a plain-text "Lattice" grep, or the panel-conviction pipeline is counting near-duplicate re-ingestions of the same source as independent claims, which would overstate breadth of coverage. Treat the "panel view" as effectively single-source, not a broad panel consensus.

7. Key risks

  1. Cyclical semiconductor demand: the FY2023→FY2025 collapse (revenue -31%, EPS from $1.88 to $0.02) demonstrates real cyclicality tied to industrial/comms/auto capex and channel inventory levels. The current "recovery" (Q1 FY2026 at $170.9M) could stall or re-correct if end-market demand softens again before Avant-driven share gains fully offset cyclicality.
  2. Competitive scale asymmetry: AMD (Xilinx) and Intel (Altera/Agilex) are vastly larger, better-capitalized FPGA franchises. Lattice's low-power niche has been relatively uncontested to date, but there is no structural barrier preventing either giant from targeting the segment more aggressively if it becomes strategically attractive (e.g., as edge-AI inference becomes a bigger, more visible market).
  3. Customer/end-market concentration and capex sensitivity: revenue is concentrated in industrial, automotive, and communications OEM customers — all sectors with lumpy, capex-cycle-driven ordering patterns, which is precisely what drove the 2024-2025 downturn and makes forward estimates (a near-doubling of revenue by FY2028) sensitive to a continued upcycle actually materializing.

8. Verdict

Watch

One-line thesis: A genuine best-in-niche low-power FPGA franchise with a credible Avant/edge-AI product cycle catalyzing a recovery off a real cyclical trough, but the "+45% forward growth" is meaningfully inflated by easy base-effect comps, and the stock already prices in several years of successful execution.

Valuation stance: rich on every clean metric (EV/Sales ~20x, forward P/E 60-140x depending on out-year) — not cheap, appropriate only for investors who believe the FY2027-2029 consensus ramp is achievable, not as a value entry.

Biggest risk: cyclical semiconductor demand reversing before the Avant-driven structural growth story is proven durable, given revenue is still ~29% below its FY2023 cycle peak.

Data flags: (1) PEG of -15.76 is mechanically broken/noise, driven by a near-zero FY2025 EPS base and depressed trailing P/E of ~3,496x — do not use it; (2) the "+45%/+34% acceleration" figures likely overstate structural growth because they are measured off a depressed cyclical trough (FY2025 revenue roughly flat YoY, still well below FY2023 peak); (3) KB evidence is thin — only 3 near-duplicate claims from what appears to be a single source, versus the quant profile's stated "38 KB claims," a discrepancy worth investigating upstream in the panel-conviction pipeline.