Second-order map
Exploratory — reasoned, hypothetical relationships for research, not investment advice.
- NVDACatalyst
NVIDIA Corporation
Headline AI-chip catalyst: new US export controls plus first earnings under interim CEO Colette Kress concentrate market attention on GPU shipment volumes
- CEVASpotlight subject
CEVA Inc.
Spotlight primary: licenses foundational processor and connectivity IP embedded by chip designers upstream of hyperscalers; royalty model is insulated from export-control shipment volatility because IP ships wherever silicon ships, and forced-domestic sourcing could broaden the SoC-designer licensee base
- ARM
Arm Holdings plc
If more geographies must build AI-edge SoCs domestically, demand for foundational CPU/NPU IP architectures that CEVA DSP IP complements could broaden across new licensees
- CDNS
Cadence Design Systems
Each new IP license hypothetically feeds a downstream SoC tape-out, increasing demand for EDA tools and design flows used to integrate licensed blocks
AI-edge device OEM / IoT module maker
Proliferation of domestically-sourced AI edge devices is the end-market that pulls CEVA-licensed connectivity IP into volume silicon
- TSM
Taiwan Semiconductor Manufacturing
Licensed IP ultimately becomes fabricated silicon; a wider base of SoC designers could translate into broader foundry demand regardless of where chips ship
- AVGO
Broadcom Inc.
Direct operational beneficiary: custom-ASIC and connectivity franchise supplies accelerator interconnect to the hyperscaler capex chain the catalyst drives, capturing demand adjacent to GPUs
- ANET
Arista Networks
Second-order supply beneficiary: AI-cluster buildout tied to the GPU catalyst requires high-speed switching fabric regardless of which accelerator ships
- CRDO
Credo Technology Group
If AI clusters scale, demand for high-speed connectivity components like active electrical cables and SerDes could rise as a non-obvious interconnect beneficiary
- CIEN
Ciena Corporation
Expanding distributed AI data-center capacity could lift demand for optical transport linking clusters across sites
- VRT
Vertiv Holdings
Denser AI networking and compute raises power and thermal-management requirements, a lateral infrastructure read-through
- CLS
Celestica Inc.
Hyperscaler networking hardware buildout could flow to contract manufacturers assembling switching and server systems
- LSCC
Lattice Semiconductor
Direct beneficiary of forced-domestic AI-edge sourcing: low-power programmable logic and control-plane silicon is embedded alongside AI accelerators in edge and infrastructure systems
- MCHP
Microchip Technology
Broader domestic edge-device production hypothetically pulls more microcontrollers and mixed-signal components into the bill of materials
- ONTO
Onto Innovation
A wider base of chip designers and more diversified fabrication could lift demand for inspection and metrology equipment, a lateral tooling link
- SITM
SiTime Corporation
Proliferating AI-edge and networking silicon needs precision timing components, an easily-overlooked downstream dependency
Take it further
Copy the analysis below into your own AI tool to pressure-test the reasoning and push it further.
The market's full attention today is on the AI chip trade — the confluence of sweeping new US Commerce Department export controls and Nvidia's first earnings print under interim CEO Colette Kress is pulling every dollar toward the direct GPU names, the accelerator plays, and the hyperscaler capex chain. That reflexive crowding into the headline GPU trade is understandable, but it misses where the durable structural signal actually sits. The more interesting read-through belongs to the silicon and software IP licensing layer — specifically the company whose processor and connectivity IP gets embedded by the chip designers the market is laser-focused on today, before those chips ever reach a hyperscaler.
CEVA reported Q3 2026 results on November 9, signing ten IP licensing agreements — six of them for AI technologies, including one with a new customer targeting AI infrastructure.
IPnest ranks CEVA first in wireless interface IP with approximately 68% market share. The export-control tightening that is rattling direct GPU shipment volumes does not reach CEVA's licensing model the same way — IP royalties flow from the silicon that ships wherever it ships, and a world in which more geographies must source AI edge devices domestically is structurally a world in which more SoC designers license foundational IP, not fewer.