SEMIFIVE Accelerates Growth with 97% YoY Revenue Surge in H1 2026, Powered by ASIC Turnkey, Mass Production, and IP

SEMIFIVE reported H1 2026 bookings of KRW 118.9B and revenue of KRW 94.7B, reaching 71% of full-year 2025 bookings and 78% of full-year 2025 revenue within six months. Profitability improved sharply: gross margin rose 18.7pp to 30.6% and operating margin improved 47.0pp to -11.5% as ASIC turnkey revenue more than doubled to KRW 54.1B and IP revenue rose to KRW 32.5B (already above full-year 2025 at KRW 30.5B). Management expects growth momentum to strengthen in H2 as backlog and production volumes convert into revenue and the new 3D-IC AI ASIC pipeline expands.

Analysis

This reads less like a one-off earnings beat and more like evidence that custom AI silicon is moving from “design intent” to repeat production. The second-order winners are the ecosystem names that monetize complexity: EDA/IP licensors and advanced packaging/test suppliers should benefit as more projects require mask sets, validation, and 3D-IC content, while merchant accelerator vendors face incremental substitution risk as OEMs internalize differentiated silicon.

The key market risk is that bookings quality may be better than visibility, but not necessarily linear conversion. Over the next 1-3 quarters, the real test is whether backlog turns into cash with continued gross-margin expansion; if margins stall or operating loss reduction reverses, the market will likely re-rate this as a scaling story with execution risk rather than a durable compounder. Over 6-18 months, the biggest threat is customer concentration: hyperscalers and large OEMs can pause custom projects, pull development spend in-house, or standardize on a smaller set of silicon partners.

Contrarian takeaway: the headline focus on growth may miss the more investable signal, which is a broader capex transfer toward the semi supply chain. The move is probably underdone in packaging/IP proxies and overdone in assuming every custom-ASIC win becomes a sticky annuity. I would watch for any commentary on node migration, 3D-IC yields, and whether reuse of validated IP actually widens margins or eventually commoditizes pricing.

Key Decisions for Investors

• Long SNPS/CDNS over 1-3 months on pullbacks: custom ASIC proliferation should increase front-end design starts, verification, and IP attach rates; prefer buying weakness rather than chasing momentum. Falsifier: if hyperscaler custom-silicon spend narrows to a few internal teams and design-cycle growth slows materially next quarter.
• Long AMKR or ASE on a 3-6 month horizon: the ramp from development into mass production should lift advanced packaging/test content, especially with 3D-IC mention as a secular mix driver. Best entry is after any post-rally consolidation; reduce if OSAT utilization commentary softens or packaging lead times compress.
• Pair trade: long SMH / short QQQ for 1-3 months as a cleaner way to express broader AI silicon adoption beyond NVDA beta. The thesis works if custom silicon broadens the beneficiary set; exit if AI capex growth becomes narrowly concentrated or the semis index fails to outperform on the next data-center spend update.
• Watchlist, not immediate trade: if a listed Korean semicap/AI-ASIC pure-play becomes liquid or tradable in your venue, use this as confirmation of ecosystem demand rather than a standalone signal. The missing data is customer concentration and forward booking conversion; do not size aggressively until those are disclosed.