CEVA · CEVA, Inc. · Stock research

Current Thesis

The 2026-08-10 Q2 print resolved the binary in the company's favour and the tape sold it anyway. Revenue $29.0M, +13% YoY; licensing and related revenue $18.2M, +21% YoY and the highest in three years; non-GAAP diluted EPS $0.08 against a $0.07 consensus; FY2026 revenue growth guidance raised to 13–15% from 12%, with non-GAAP operating income guided to grow roughly 70% YoY. Investing.com recorded the shares down 10.45% on the day of the release. On 2026-08-11, TD Cowen cut its target to $45 from $55 while keeping Buy, Oppenheimer cut to $38 keeping Outperform, and Rosenblatt raised to $49 keeping Buy. The reference close of 2026-08-14 is $33.28, 33.5% below the $50.06 52-week high set in June, with RSI(14) at 46.8 and a three-month return of -8.8%.

The leg an investor buys here is no longer the June momentum leg — that structure went in the mid-July semiconductor drawdown and the print did not rebuild it. What remains is a conversion story: a three-year-high licensing book turning into per-unit royalties, with Bluetooth High Data Throughput royalties guided on the 2026-08-10 call to begin ramping toward the end of 2026 and to deliver "meaningful growth in 2027 and 2028." That leg is dated and specific, but it has no company catalyst inside the next 30 days; the next scheduled resolution point is the Q3 print in early November.

Bullish and bearish views on CEVA, Inc.

The model's bull view on CEVA, Inc. (CEVA), in brief: Licensing accelerated again. Q2 licensing and related revenue $18.2M, +21% YoY, the highest in three years, on 10 agreements signed including 2 first-time customers and 2 direct OEM deals (Q2 2026 press release, 2026-08-10). Royalties turned sequentially. Royalty revenue $10.8M… The bear view: Royalty growth is flat where it counts. $10.8M was only +1% YoY. The sequential bounce is one quarter; the YoY line does not yet show the conversion the thesis rests on. The market rejected a beat-and-raise. A 10.45% decline on 2026-08-10 into a raised guide is a demand signal… Both cases follow in full.

Bull Case

• Licensing accelerated again. Q2 licensing and related revenue $18.2M, +21% YoY, the highest in three years, on 10 agreements signed including 2 first-time customers and 2 direct OEM deals (Q2 2026 press release, 2026-08-10).
• Royalties turned sequentially. Royalty revenue $10.8M, +17% QoQ, attributed on the call to wireless connectivity shipments, automotive AI ramps and smartphone share gains. Ceva-powered device shipments were 567M units in the quarter, +16% YoY.
• The guide went up. FY2026 revenue growth raised to 13–15% from 12%; non-GAAP operating income guided ~+70% YoY and non-GAAP net income ~+50% (2026-08-10 call). Q3 revenue guided to $30.5M–$34.5M, midpoint $32.5M, at ~88% non-GAAP gross margin.
• Operating leverage is showing. Non-GAAP operating margin 11% in Q2 2026 versus 3% in Q2 2025.
• Balance sheet carries the wait. Cash and equivalents $44.3M plus marketable securities and short-term bank deposits $176.4M as of the Q2 release — no financing pressure while the royalty ramp develops.
• A dated new royalty vector. Ceva-Waves Bluetooth High Data Throughput powers Actions Technology's ATS296X-series audio chips (announced 2026-08-04), extending a platform the company describes as having shipped in 100M-plus SoCs.
• The Street has not left. Rosenblatt raised to $49 on 2026-08-11; TD Cowen and Oppenheimer maintained Buy/Outperform while trimming targets.

Bear Case

• Royalty growth is flat where it counts. $10.8M was only +1% YoY. The sequential bounce is one quarter; the YoY line does not yet show the conversion the thesis rests on.
• The market rejected a beat-and-raise. A 10.45% decline on 2026-08-10 into a raised guide is a demand signal about the marginal buyer rather than about the quarter itself.
• Targets came down from the sell-side's own bulls. TD Cowen $55 → $45 and Oppenheimer to $38 on 2026-08-11; TD Cowen's $55 had itself been a raise from $45 only on 2026-07-13. Prior coverage in this dossier recorded JPMorgan Neutral at $30 in mid-July.
• GAAP still loses money. GAAP diluted loss of $(0.10) per share in Q2 2026, improved from $(0.15) a year earlier but negative. The $0.08 headline is non-GAAP.
• The marquee AI deal pays late. Management said on the 2026-08-10 call that the custom AI-silicon engagement with a leading platform company tapes out "in quarters ahead," with production 1.5–2 years after that — a 2028 royalty event.
• Analyst questions on the call clustered on timing risk: whether platform-level engagements sustain growth, when new AI and connectivity programs ramp, GAAP profitability, FX headwinds, and smartphone uncertainty tied to memory constraints.
• Structure broke before the print and stayed broken. The June breakout shelf in the mid-$40s gave way in the mid-July semiconductor rout that followed Moonshot AI's model release (~2026-07-16); no reclaim of it has printed since.

Setup & Price Structure

Reference close 2026-08-14: $33.28. That is 33.5% below the $50.06 52-week high (June 2026) in the split- and dividend-adjusted series graded here. RSI(14) 46.8 is mid-range — no oversold extreme to mean-revert from — and the three-month return of -8.8% brackets both the mid-July break and the post-print decline.

The narrative is saturated, dated by 2026-08-10 and 2026-08-11. The definition that fits is thin new bid rather than failed fundamentals: the company delivered a revenue beat, a licensing three-year high and a raised full-year guide, and the shares fell 10.45% on the day, after which two of the three brokers who revised on 2026-08-11 cut targets. The momentum phase that ran into the June high is over; the re-rating that carried the stock from the $17.02 low is now being marked down by desks that raised into it. The label is not dead because the operating numbers accelerated rather than rolled — revenue +13% YoY, non-GAAP operating margin 11% versus 3% — but no base has formed to date the next leg from.

Crowding and positioning observables, stated as observables: three price-target revisions landed on a single day (2026-08-11); retail-facing screen coverage appeared on 2026-08-14 (Zacks technical/rank commentary) after the drawdown rather than before it; price sits below its June structure rather than extended above it; the recent filings record shows no insider Form 4 activity in this window; and the next earnings date is roughly three months out, so no imminent print is compressing the setup in either direction.

Structural reference points: the mid-$40s June shelf is the level a genuine trend repair would have to reclaim on volume; the low-$30s is where price is currently resolving; below that, prior coverage identified high-$20s congestion as the next area of prior trade.

Catalyst Calendar (next 30 days)

• Through 2026-09-15: no confirmed company-dated event. The Q2 release, call and quarterly filing are done (2026-08-10/11). The most recent Ceva conference-participation release located in this research is dated 2026-02-19 and covers March events only; no September 2026 investor-conference date is confirmed.
• ~2026-11-09 (est.): Q3 2026 results. Tests the $30.5M–$34.5M guide, whether royalties hold above the $10.8M Q2 level, and whether the raised 13–15% FY guide survives contact with H2.

Elapsed catalysts

• ~2026-Q4 (est.): Bluetooth High Data Throughput royalties begin ramping, per management's 2026-08-10 framing of "toward the end of 2026." First observable in reported royalty revenue at the Q3 or Q4 print. (passed 16d ago)
• Undated but live: the custom AI-silicon tape-out described on the 2026-08-10 call as "quarters ahead." No announced date; an announcement would arrive as unscheduled news. (passed 16d ago)

What Would Change Our Mind

The conversion leg is what is being underwritten, so it breaks at the royalty line first: royalty revenue failing to grow sequentially off Q2's $10.8M at the ~November Q3 print, or FY guidance being walked back from the raised 13–15% range, would mean the licensing strength of the last two quarters is not converting on the timeline management described on 2026-08-10.

On price, a weekly close below $31 loses the low-$30s area held since the print and opens the high-$20s congestion that preceded the 2026 re-rating; that would move the label from saturated to dead.

The other direction is equally specific. A higher low in the low-$30s followed by a weekly close back above the mid-$40s June shelf, on expanding volume and with SMH/SOX turning up, would re-date the narrative as accelerating and put Rosenblatt's $49 target back in play as a live reference. Absent one of those two, the name sits between a broken structure and an unproven royalty ramp, which argues for waiting on a base rather than chasing the guide.

Correlation Notes

• Group beta dominates single-name news. The mid-July drawdown was macro: Moonshot AI's model release (~2026-07-16) reignited AI-capex fear, SMH fell roughly 9% over that stretch, and the stock broke its June shelf with the group. Prior coverage in this dossier cited beta near 2.0 on ~1.5–1.9M shares average daily volume — direction confirmation reads better off SMH/SOX, ARM and AVGO than off the name itself.
• Smartphone royalties tie to the memory cycle. Management flagged smartphone uncertainty linked to memory constraints on the 2026-08-10 call, and the entry-level recovery runs through UNISOC, so DRAM/NAND pricing and Chinese handset volumes feed the royalty line with a lag.
• Automotive AI is a Renesas-linked ramp. The Ceva AI DSP and accelerator content in the R-Car V4 platform converts to per-unit royalties on that program's shipment schedule, making auto-semi order commentary a leading read on one royalty component.
• The licensing line does not track the group. Deal timing drives it (10 agreements in Q2 2026), so quarter-to-quarter revenue stays a poor real-time demand proxy even when royalties are behaving.