Chip Design IP

Frankly speaking, this income statement is the kind of thing that makes you love the business and lose sleep over it at the same time. IP licensing is, at its core, a business that lives and dies by the calendar — one large order lands and quarterly revenue can jump to 1.2 billion, then one or two customer pushbacks and we're back down to a bit over 800 million. Strip that layer of volatility away, and the "normalized revenue" over the past four quarters has actually been stable in the 830 million to 900 million range, with gross margin settling back around 32%. Let me lay out two assumptions for you, and you tell me whether they hold up: **Assumption 1:** 2026Q2 revenue returns to roughly 890 million. Q1 is typically the tail end of the soft season for IP companies, and in Q2, as customer projects advance to the RTL stage, verification and integration licensing gets booked sequentially. Historically, our Q2 running 5–8% above Q1 is normal range. I'm not baking in another 2025Q3-style "one satellite mega-deal" scenario here — that was a one-off, not a base case to extrapolate. **Assumption 2:** 2026Q3 revenue climbs to 980 million, with **gross margin lifting from 32% to 33%**. The logic is that our NPU IP "tree-planting" from the past two years is starting to bear fruit. The new-generation NPU IP is ramping inside customer SoCs, and the per-unit royalty is a bit thicker than what we get on legacy DSP/GPU IP — so a blended product mix naturally nudges gross margin up by a notch. That said, operating profit remains in loss territory, because we haven't eased up on R&D spending — doing IP means constantly pouring money into the next-generation ZSP, the next-generation NPU. That's the cost of planting trees, and there's no way around it. On the **biggest risk**, I'll name one and call it enough: losses are **widening on a sustained basis**, from -100 million in Q2 to -340 million in Q1. If that slope continues, I'm going to need to sit down with the board and talk about how to balance "R&D investment" against "operating efficiency." The concrete approach would likely be more disciplined project portfolio management on the engineering side — trimming duplicate IP maintenance spend, and concentrating firepower on the NPU and Chiplet tracks. If we happen to land one or two large upfront payments from a major customer in any given month, the cash recovery can be fast; but on the flip side, if customer tape-outs slip by two or three quarters, the losses will keep compounding. I've said it before — "helping our customers tape out" is our craft. And what a craftsman fears most isn't too much work — it's not knowing when the next job is coming. I've already built some buffer into my forecast for that uncertainty, but honestly, I'm not fully confident either.