Open in interactive viewer → charts, metric popovers & call review

CEVA: A 44% Crash Into the Best Quarter It Has Ever Guided

Record Q4 revenue and doubled EPS sit on the other side of a stock that just fell nearly in half — the market is betting the licensing slip and the memory excuse are the real story
CEVA · Earnings Call · 2026-11-09

The tape and the guide are screaming different things

CEVA walked into its November 9 print with no benefit of the doubt. The stock is down roughly 44% over the last 90 trading days and sits about 47% below its July 9 peak of $46.81 — a brutal, one-directional slide into the number. Then management delivered the opposite of that tape: fourth-quarter guidance of $34–38 million, roughly 24% sequential growth at the midpoint, which Yaniv Arieli framed plainly. “And obviously, with that said, we're still guiding to the best licensing revenue and royalty revenue ever for CEVA, which is the $34 million to $38 million, which is a very nice new place to be at, if all goes well.” — Yaniv Arieli, Chief Financial Officer · 2026-11-09 Amir Panush was even more direct: the midpoint is “the highest quarter ever for CEVA.” — Amir Panush, Chief Executive Officer · 2026-11-09 Non-GAAP EPS is guided to more than double sequentially from $0.13. So the central tension of this report isn't the numbers — it's the gap between a company promising its best-ever quarter and a shareholder base that has spent three months selling it down. Somebody is wrong, and the transcript gives us the tools to figure out who.

The recurring excuse that stopped sounding like a one-off

Start with the miss. Licensing came in at $16.3 million, up only 2% year over year, because a large multiyear connectivity renewal with an existing customer slipped out of Q3. It has since been signed in early Q4. Fine, a timing quirk — but the explanation for the quarter's softness has a long tail now. The same theme keeps reappearing. In August, Yaniv told the room on memory pressure: “right now, we haven't heard anything specific from our customers.” — Yaniv Arieli, Chief Financial Officer · 2026-05-11 In May, Amir had already caveated that “definitely the memory shortage has an impact on the wireless handset industry, and it's hard to quantify exactly how that will make an impact in the second half.” — Amir Panush, Chief Executive Officer · 2026-08-10 Now in November it's stated as fact: “memory market headwinds tempered volume across compute-intensive consumer devices such as smartphone, PCs and TVs.” — Amir Panush, Chief Executive Officer · 2026-11-09 That is the fourth consecutive quarter of memory market headwinds as an explanatory crutch, and the language has hardened from "haven't heard anything" to "disciplined assessment" baked into guidance. When a company trims its full-year outlook and simultaneously claims its best quarter is one quarter away, the memory narrative is doing a lot of load-bearing work in convincing investors. It is also why reasonable people sold first and waited for proof second.

As you pointed out, overall, if we look at the market, all the compute-intensive markets are under stress right now... a disciplined assessment of how we look at the quarter, also considering that this quarter [indiscernible] day, we haven't executed to our expectation, we came with that guidance.

Amir Panush, Chief Executive Officer · 2026-11-09

NeuPro Ultra: the one genuinely new theme in this call

Here is where CEVA breaks new ground rather than repeating itself. For several years the AI story has been "edge NPUs, 18–24 months to royalty, be patient." This quarter introduced NeuPro Ultra and a deliberate reach beyond the edge into AI infrastructure — a genuinely fresh thematic axis, and one that maps directly onto a global market conversation. Amir named two new classes of buyer: “there are definitely the memory suppliers and the high-speed connectivity suppliers.” — Amir Panush, Chief Executive Officer · 2026-11-09 That is a company-unique vocabulary appearing for the first time — memory supplier and connectivity suppliers are brand-new terms for CEVA, and management confirmed a new customer already signed for an AI infrastructure application in Q3. The pitch is credible in principle (accelerators attached to memory and interconnect reduce decode latency and improve power efficiency in inference racks), and it also cross-validates against the broader market's relentless data-center build. But note the honest tell: management admits time-to-royalty is the same 18–24 months. So this is a 2028 royalty story dressed as a 2027 inflection. AI licensing at a record and >35% of licensing is impressive; it is still licensing revenue, not the annuity investors ultimately pay for.

The smartphone modem that keeps arriving next quarter

Meanwhile, the major U.S. customer modem ramp — the thing that has been hyped as the swing factor since 2025 — has again been deferred to the following quarter. In August the language was second-half; now it is "Q3 contribution was quite minimal" and “we expect a very significant ramp in Q4 and as we go into 2027.” — Amir Panush, Chief Executive Officer · 2026-11-09 That is now a multi-quarter pattern of the inflection sitting one print away. To be fair, mobile royalty revenue still grew 20% year over year and 30% sequentially before the ramp kicked in — that is real. But the market has heard "Q4 is the ramp" before, and combined with the licensing slip, it explains why a stock at 11.5x trailing sales (per the latest reading, before the 44% drawdown) could still lose nearly half its value. Ten-year revenue trend: up just 75% over a decade, peak $33M in 2022Q4 The underlying business is more resilient than the price action suggests — Gross margin near 87% and $221M of cash on a $763M market cap gives room. But operating income remains negative, and the free-cash-flow line swings around zero. The bull case still rests entirely on converting design wins into royalties.

Bottom line

Nothing here is fraudulent and nothing is obviously broken — the connectivity franchise is #1 by a wide margin, Wi-Fi 6 mix jumped from 24% to 57% of units, and the Q4 setup is real if the modem ramp and the signed connectivity deal land. But the reason this is a name in motion rather than a quiet quarter is the inversion: a 44% drawdown against the most bullish single-quarter guide in company history. Either the market over-extrapolated a licensing slip and a tired memory excuse, or management is again leaning on "next quarter is the record" to buy time. The NeuPro Ultra infrastructure expansion is the one item that could genuinely reset the story — but at 18–24 months to royalty, it won't rescue 2026. The proof point arrives in Q4. There is no more room to defer the inflection.