Cymetica
Specialized AI
PAPER
Menu
Dark Mode
Plain English Mode
PAPER TRADING MODE — Enable real trading on your Account page
Back

AI Adaptation Research Report — CEVA (CEVA INC) · WhiteSwan constituent

AI Adaptation Research Report — CEVA (CEVA INC) · WhiteSwan constituent

EventTrader research desk · generated 2026-11-05 04:01 UTC · report ai-adapt-ceva-20261105 · narrative: built from the evidence pack by the platform

Company CEVA — CEVA INC · SIC 7370 Services-Computer Programming, Data Processing, Etc.
Book AMF-WHITESWAN (long) — WhiteSwan
Composition algorithm Swan Composite v1.1.0 — 50% AI Adaptation Score + 25% AI Story Score + 25% market verdict, conviction-weighted
Anchor filing 10-Q filed 2026-08-10 (period 2026-06-30)
Adoption level 4 — native · model-confirmed
AI Adaptation Score 71.8 / 100 · run 2026-11-05 · sector software percentile 97.9
Price 28.02 USD
Place in the book rank 8 of 10 · composite 66.0 · story 50.1 · market verdict 70.5 · weight 6.6%

1. Verdict

  • CEVA (CEVA INC) is a WhiteSwan constituent: its own filings document AI inside the product or operations with investment behind it, and its AI Adaptation Score ranks at the top of the software companies (SIC 7370–7379) on the 2026-11-05 run [S5].
  • The score rests on the company's 10-Q filed 2026-08-10, read in full by the platform's scorer [S1].
  • Adoption level 4 — native on the 0–4 rubric, confirmed by a model read with credibility 7.0/10 [S1].
  • AI Adaptation Score 71.8 / 100, percentile 99.7 of 6503 scored filers and 97.9 within its sector (software, 585 filers) [S5].
  • What would move it: a next report that withdraws the deployments, investment or outcomes it now states [S1].

2. The company

CEVA INC files with the SEC under CIK 1173489, SIC 7370 (Services-Computer Programming, Data Processing, Etc.), listed on Nasdaq [S4]. The filing the score reads is the 10-Q filed 2026-08-10 for the period ended 2026-06-30, effective for the books from 2026-08-11 [S2].

3. What the filing says about AI

The scorer found 30 AI-related sentences in the 10-Q: 30 in the substantive sections (business, management's discussion) and 0 in the risk factors [S1]. Signals counted in the substantive text: proprietary-model language 0, deployments 6, measured outcomes 4, investment 0, third-party vendor tools 0, exploratory language 1 [S1].

Verbatim from the filing:

"Royalty revenues were $10.8 million, up 17% sequentially, reflecting improving smartphone royalties, continued strength in wireless connectivity and the ongoing ramp of automotive AI programs." — Item 2. Management's Discussion and Analysis [S1] "ABI Research projects that by 2030, over 50% of TinyML shipments will be powered by dedicated embedded AI hardware such as NeuPro‑Nano. ● Inference layer – AI DSPs for perception and sensor intelligence: Our SensPro2 AI DSP family addresses demand for efficient, high‑performance AI signal processing across sensor‑rich applications, from smartphones and drones to automotive ADAS and industrial IoT." — Item 2. Management's Discussion and Analysis [S1] "BUSINESS OVERVIEW We enable Physical AI, the artificial intelligence embedded in billions of devices that connect, sense and infer data in the real world." — Item 2. Management's Discussion and Analysis [S1] "Our transformative semiconductor IP and embedded software offerings are incorporated by customers into application-specific integrated circuits (ASICs) and application-specific standard products (ASSPs) to enable power‑efficient, intelligent, secure and connected devices that connect, sense and infer – the three critical pillars of the rapidly evolving era of AI‑enabled smart edge technology." — Item 2. Management's Discussion and Analysis [S1] "We believe our portfolio of technologies comprised of connectivity, sensing and inference – the three foundational layers of Physical AI – positions Ceva at the center of the most important megatrends shaping the semiconductor industry, including 5G expansion, generative and embedded AI, industrial automation and vehicle electrification." — Item 2. Management's Discussion and Analysis [S1] "These engagements demonstrate the growing adoption of our AI technologies across semiconductor, OEM and platform customers. ● Inference layer – Embedded AI and TinyML: Our NeuPro‑Nano NPUs bring cost‑ and power‑efficient AI to microcontrollers and systems-on-chips (SoCs), enabling artificial IoT (AIoT) devices for sound, vision, vibration, and health monitoring." — Item 2. Management's Discussion and Analysis [S1] "We are also seeing continued contributions from AI-driven royalties, highlighted by the ongoing ramp of automotive AI programs and a ramping AI SoC for surveillance, marking the initial phase of what we expect to become a meaningful long-term growth driver." — Item 2. Management's Discussion and Analysis [S1] "While Edge AI refers to running AI workloads locally on devices rather than in the cloud, Physical AI extends this concept further: it unifies connectivity, sensing and inference layers into a single fabric that allows devices not only to process data at the edge, but also to interact intelligently with their physical environment and the cloud." — Item 2. Management's Discussion and Analysis [S1] "Based on market research, we believe these sectors will represent a $170 billion total addressable market for Physical AI and Edge AI by 2030." — Item 2. Management's Discussion and Analysis [S1] "AI-related licensing represented more than 20% of total licensing and related revenues in the quarter, highlighting the accelerating adoption of edge AI across multiple end markets." — Item 2. Management's Discussion and Analysis [S1]

4. The AI Adaptation Score

The AI Adaptation Score weighs the adoption level (0–4) at 45%, building with AI from the ground up (greenfield) at 20%, the credibility of the claims at 15%, the trajectory against the previous filing at 12% and the investment signal at 8%, with trajectory and investment measured against the company's own sector; the greenfield reward counts only from level 3 upward [S5]. CEVA's score of 71.8 decomposes into level 45.0, greenfield 8.0, credibility 10.5, trajectory 4.8 and investment 3.5 points [S5]. Its cluster label for the run is builder with a margin of 40.9 points from the sector's other cluster [S5].

The model read of the filing returned level 4 (native) with confidence 0.72 and sub-scores product 9.0, operations 4.0, investment 4.0, data moat 6.0, credibility 7.0, greenfield 4.0 (each 0–10) [S1]. Its summary: "AI is CEVA's core product: proprietary NPU/DSP IP (NeuPro-M, NeuPro-Nano, SensPro) licensed as the basis of Physical AI, with measurable traction—AI licensing >20% of revenue and ramping automotive AI royalties. Native. Not founded AI-first (evolved from connectivity/DSP), so limited greenfield origin." [S1].

"AI-related licensing represented more than 20% of total licensing and related revenues in the quarter, highlighting the accelerating adoption of edge AI across multiple end markets." — Item 2. Management's Discussion and Analysis [S1] "We enable Physical AI, the artificial intelligence embedded in billions of devices that connect, sense and infer data in the real world." — Item 2. Management's Discussion and Analysis [S1] "The quarter included one of the most strategically significant AI licensing agreements in our history, with a leading global AI and computing platform company selecting our NeuPro-M NPU IP for its next-generation custom AI silicon." — Item 2. Management's Discussion and Analysis [S1] "Our NeuPro‑M AI NPU family delivers efficient, high‑performance architectures for generative and classic AI across devices from gateways and notebooks to AR/VR and smartphones." — Item 2. Management's Discussion and Analysis [S1] "We are also seeing continued contributions from AI-driven royalties, highlighted by the ongoing ramp of automotive AI programs and a ramping AI SoC for surveillance" — Item 2. Management's Discussion and Analysis [S1]

5. Trajectory

The previous periodic report, the 10-Q filed 2026-05-11, scored level 4 (native) with 29 AI sentences [S3]. The level is unchanged between the two reports [S1].

6. Recent developments

  • wire:prnewswire 2026-11-04T14:32+00:00: Ceva Welcomes Introduction of Bluetooth High Data Throughput, Accelerating Next-Generation Wireless Connectivity [S7]
  • wire:prnewswire 2026-11-02T05:00+00:00: Ceva Extends Wireless Connectivity IP Leadership as AI Moves to Connected Devices [S8]
  • wire:prnewswire 2026-11-02T05:00+00:00: Ceva and Mobilint Advance Physical AI with REGULUS Edge AI SoC [S9]
  • wire:prnewswire 2026-10-28T11:00+00:00: Ceva Unveils Next-Generation NeuPro NPU Portfolio and Announces Expansion into AI Infrastructure [S10]
  • wire:prnewswire 2026-10-14T11:00+00:00: Ceva, Inc. Schedules Third Quarter 2026 Earnings Release and Conference Call [S11]
  • wire:prnewswire 2026-09-25T20:35+00:00: 2026 Drive for Life Raises $1.5 Million for Charity with help from Carson Hocevar & Pat Fitzgerald [S12]
  • wire:gnews-ma 2026-09-22T12:35+00:00: STIM to Acquire Norwegian Advisory Firm OCEVA, Splits Company in Two - Seafoodnews [S13]
  • wire:gnews-ma 2026-09-21T10:33+00:00: STIM to acquire OCEVA, combine it into new aquaculture advisory - Undercurrent News [S14]

7. What would change the verdict

CEVA leaves the long book if a later report drops the deployments, investment or outcomes it now states, if the model read judges the claims uncredible, or if higher-scoring filers displace it on a later run [S5]. The book is recomposed on every scoring run from the ten highest scores that pass the liquidity screen [S5].

8. Risks and caveats

The score reads what the company discloses, not what it does: a company can adopt AI without describing it in a periodic report, and can describe it without adopting it [S5]. Risk-factor language is counted separately from substantive text and never credited as adoption; boilerplate that resembles generic risk disclosure is excluded [S5]. Only the sections the scorer parses are read (business, risk factors, management's discussion for a 10-K or 10-Q; the earnings release of an 8-K is scored but never anchors a level) [S5].

9. Watch list

  • The next periodic report of CEVA (expected around 2026-11-08) and its AI sentences [S4].
  • The model read of the current filing (done) [S5].
  • The next scoring run's AI Adaptation Score and the book's recomposition [S5].

Place in the book

  • Rank 8 of 10 in AMF-WHITESWAN (Swan Composite v1.1.0, scoring run 2026-11-05): the book holds the highest composites [S5].
  • Composite: 0.50 × 71.8 (AAS) + 0.25 × 50.1 (story) + 0.25 × 70.5 (market) = 66.0 [S5].
  • Market verdict 70.5: over the last 90 days CEVA rose more and fell less than the software universe's median daily return (down-capture 0.86, up-capture 1.68, L = 0.82; 32 down / 19 up days) — the Failure-To-Adapt fund's selection signal, 50 = moves with the software universe's median daily return [S5].
  • AI Story Score 50.1: no AI story or a neutral one in the AI stories written about the company [S5].
  • Weight: rank factor 1.22 (rank 8 of 10) × story factor 1.00 × market factor 1.21, normalised across the book and clamped to 5–20% → 6.6% [S5].

Sources

Method and disclaimer

The AI Adaptation Score reads each company's own periodic SEC reports (10-K, 10-Q, 20-F, 40-F): deterministic features (AI sentences by section, proprietary-model, deployment, outcome, investment and vendor language, boilerplate and negation excluded) set an adoption level on a 0–4 rubric; a model read of the filing's AI sentences confirms the level and scores credibility and greenfield (the business built with AI from the ground up rather than retrofitted); the score combines level, greenfield, credibility, trajectory against the previous report and investment, measured against the company's sector, with greenfield carrying the highest reward. The books are composed by Swan Composite v1.1.0: each company's composite is 50% AI Adaptation Score + 25% AI Story Score + 25% market verdict — the AI Story Score is the sentiment of the AI stories written about it, the belief network of companies named in the same stories and the price reaction to those stories (no story scores neutral); the market verdict is the Failure-To-Adapt fund's selection signal, the signed down/up capture ratio of the stock against the software universe's median return over 90 days (a stock that falls more and rises less than its peers scores below 50). AMF-BLACKSWAN holds the ten lowest composites that pass the liquidity screen and the new-listing gate and AMF-WHITESWAN the ten highest, weighted by conviction (rank, story agreement and market agreement, each leg 5–20%), recomposed on every scoring run. Every quote in this report is the filing's text verbatim; every factual sentence names its source; a source the platform could not reach is reported as missing, never filled in.

This report is information, not investment advice. Figures come from the cited sources as of their dates. Nothing here is a recommendation to buy, sell or hold any asset.