CEVA · CEVA, Inc. · Stock research

Current Thesis

Ceva licenses DSP/NPU compute and wireless-connectivity IP — Bluetooth, Wi-Fi, UWB, cellular IoT — into other companies' silicon and collects per-unit royalties, a picks-and-shovels model levered to edge-AI chip volumes. The fundamental story has not broken, but the price structure has. The June reclaim that carried shares to within ~3% of the $51.60 52-week high failed to hold, and the mid-July semiconductor rout — triggered by Chinese startup Moonshot AI unveiling a model that narrowed the gap with top US offerings and reignited AI-capex fear — dragged the stock back through the ~$44 June breakout base to the mid-$30s (July-25 close ~$36.10, intraday low $35.85), roughly 30% off the high. Sell-side is still climbing (TD Cowen raised to $55 on 2026-07-13), yet the tape is rolling over into a confirmed 2026-08-10 Q2 print. That combination — broken momentum, a binary catalyst two weeks out — makes this a stand-aside rather than a fresh momentum entry, whatever the licensing narrative deserves on a multi-year horizon.

Bullish and bearish views on CEVA, Inc.

The model's bull view on CEVA, Inc. (CEVA), in brief: Sell-side still raising targets. TD Cowen lifted its PT to $55 from $45, Buy, on 2026-07-13, tying Needham's $55 (2026-06-15 "Physical AI" initiation) as Street-high; Stifel $50 (2026-07-10) and UBS $48 sit beneath. After the drop those targets imply 30–50%+ upside from ~$36.… The bear view: The breakout base broke. Price sliced through ~$44 — the level the June structure needed to hold — and now trades in the mid-$30s, ~30% under the $51.60 high set in June. A failed breakout has turned into a downtrend of lower highs. Macro AI-capex scare hit the group. Moonshot… Both cases follow in full.

Bull Case

• Sell-side still raising targets. TD Cowen lifted its PT to $55 from $45, Buy, on 2026-07-13, tying Needham's $55 (2026-06-15 "Physical AI" initiation) as Street-high; Stifel $50 (2026-07-10) and UBS $48 sit beneath. After the drop those targets imply 30–50%+ upside from ~$36.
• Licensing inflection is dated and real. Q1 2026 (reported 2026-05-11): licensing & related revenue $17.8M, +18% YoY — best in three years — with 14 new license agreements; total revenue $27.0M, +11% YoY, beat the $26.12M consensus; AI was >20% of licensing revenue.
• Fresh AI design win, 2026-07-06. Custom AI-silicon licensing deal with a US software/AI-platform company for computing devices — seeds a future royalty tail and keeps AI a live licensing line.
• Connectivity franchise depth. Needham cites ~68% global wireless-connectivity IP share (2026-06-15); the Ceva AI DSP + accelerator designed into the Renesas R-Car V4 automotive platform converts to per-unit royalties as volume ships.
• Guide raised, H2-weighted. FY2026 revenue guided to the top of the 8–12% range with a 40–50% non-GAAP profit-growth target (2026-05-11); royalty strength skews to H2 on Wi-Fi, cellular IoT and automotive AI.
• Valuation objection has partly reset. Consensus target sat near $41 (2026-07-03); after the slide to ~$36 the stock trades below the mean target rather than above it, removing one bear talking point.

Bear Case

• The breakout base broke. Price sliced through ~$44 — the level the June structure needed to hold — and now trades in the mid-$30s, ~30% under the $51.60 high set in June. A failed breakout has turned into a downtrend of lower highs.
• Macro AI-capex scare hit the group. Moonshot AI's mid-July model release (~2026-07-16) narrowed the US lead and sparked a fresh semiconductor rout; the VanEck Semiconductor ETF (SMH) fell ~9% over the period (third weekly decline in four weeks). High-beta CEVA underperformed, down ~13% month-to-date in July.
• A cautious new voice. Benchmark's Gary Mo initiated Hold on 2026-07-16, calling the shares richly valued after more than doubling off the $17.02 52-week low.
• Thin earnings quality. Non-GAAP EPS $0.04 in Q1 2026 vs $0.06 a year earlier — EPS fell YoY on +11% revenue; GAAP net loss $4.5M; TTM net income about -$11.77M. Royalty revenue was only $9.2M and is decelerating relative to the lumpy licensing pop.
• Multiple still full even after the drop. ~$1.03B market cap on ~$112M TTM revenue ≈ 9x sales; forward earnings remain richly priced while Needham dates Physical-AI adoption "likely still more than two years away" (2026-06-15).
• High-beta unwind risk. Beta ~2.0, ~1.5–1.9M ADV; both the June-5 reversal and the mid-July slide showed how fast it gives back gains on a risk-off semi tape.

Setup & Price Structure

The story held; the tape did not. June's reclaim to within ~3% of the $51.60 high demanded that ~$44 stay support, and the mid-July rout removed it. Shares now sit in the mid-$30s (July-25 close ~$36.10, intraday low $35.85) inside a 52-week range of $17.02–$51.60, printing lower highs off the June peak with price beneath the broken breakout base. This is the rolled-over structure the momentum playbook flags as a value trap: a cheaper multiple attached to a downtrending chart, with sell-side targets pointing one way while price goes the other. There is no valid momentum entry here. A recovery thesis first needs a base — a higher-low hold in the low-$30s followed by a volume reclaim back over ~$44 — ideally confirmed by AVGO/ARM/SOX/SMH turning up, since a small-cap IP licensor does not lead the group off a bottom. Below the mid-$30s, the next references are the low-$30s and the high-$20s pre-June-run congestion.

Catalyst Calendar (next 30 days)

• 2026-08-10 (confirmed): Q2 2026 earnings, before market open, 8:30 a.m. ET call. Binary event. Watch royalty ramp (Wi-Fi / cellular IoT / automotive AI, H2-weighted per the May guide), licensing pipeline durability, and AI mix versus the $9.2M royalty deceleration. The 3-trading-day pre-print window opens ~2026-08-05; entering an unconfirmed base ahead of a binary print on a broken tape is a gamble, not an edge.
• No FDA/PDUFA or other dated company events in the 30-day window beyond the earnings print.

Elapsed catalysts

• Ongoing — semiconductor tape / SMH. The mid-July AI-capex fear (Moonshot AI model, ~2026-07-16) is the live overhang; stabilization in AVGO/ARM/SOX is the precondition for any CEVA base. No dated macro release is the trigger — it is group sentiment on edge-AI silicon spend. (passed 24d ago)

What Would Change Our Mind

• Bullish re-fire: a weekly close back above ~$44 on expanding volume reclaims the broken breakout base and revives the momentum leg; a Q2 print (2026-08-10) showing royalties re-accelerating off the $9.2M base alongside continued double-digit licensing growth would convert the $50–$55 sell-side targets from hope into trend.
• Bearish confirmation: a weekly close below $34 breaks the July support shelf (July-25 low $35.85) and opens the low-$30s / high-$20s, confirming the failed-breakout unwind.
• Theme decay: the edge-AI / physical-AI licensing narrative flipping from cooling to dead if design-win headlines and target revisions stall while the AI-capex scare keeps the semi complex under distribution.

Correlation Notes

• CEVA reads off AVGO / ARM / SOX / SMH, not off itself: a ~2.0-beta IP-licensing small cap amplifies the semiconductor tape in both directions. Mid-July proved it — SMH -~9% over the period, CEVA -~13% month-to-date.
• Fundamentally levered to Bluetooth / Wi-Fi / UWB / cellular-IoT unit volumes and to automotive AI (Renesas R-Car V4); consumer, IoT and auto end-market unit cyclicality drives the royalty line.
• The shared risk factor that hit the group in mid-July is AI-capex fear — China model parity (Moonshot AI) questioning US AI spend; CEVA moves with sentiment on edge-AI silicon demand more than on its own news.
• Narrative comps: ARM (IP licensing), AVGO / QCOM (connectivity), and the broader edge-AI silicon complex; treat cluster confirmation as a requirement before any re-entry, given the name does not set its own trend.