Ambarella (NASDAQ:AMBA) used its CES briefing to outline an expanded edge AI strategy, introduce new silicon and software milestones, and describe a shift in go-to-market execution aimed at scaling beyond the company’s historically direct-sales model. Management emphasized that the presentation was not a financial update, noting that fiscal 2026 ends in a few weeks and that financial guidance would be addressed on the company’s earnings call in late February.
Edge AI positioning and market dynamics
CEO Dr. Fermi Wang framed edge AI as fundamentally different from data center AI, citing requirements including low power consumption, low latency, privacy, limited bandwidth (including DRAM bandwidth), and demand influenced by both enterprise CapEx and consumer spending. He said Ambarella believes edge AI remains in the early stages of commercial development, creating an opportunity for the company’s silicon and software platform.
New silicon milestones: 4nm sampling and 2nm tape-out
Wang highlighted five focus areas for the company, including platform updates, a new 4-nanometer SoC, a taped-out 2-nanometer chip, a new go-to-market strategy, and an evolution of the Cooper software platform.
On silicon, Ambarella said it is sampling its first 4-nanometer AI SoC, CV7, built at Samsung Foundry. Wang said Ambarella received samples and brought up two demonstrations within 48 hours: an 8K P60 video plus AI processing demo and a four-channel 4K P30 video input demo, each combining video processing and AI. He said CV7 is expected to enter production by the end of the year in certain applications.
Ambarella also confirmed it has taped out its first 2-nanometer chip at Samsung Foundry. Wang described it as a semi-custom design with a first customer helping fund the chip, with additional updates to come later. Separately, management reiterated its earlier announcement of the N1 family targeting edge infrastructure, noting the first design win is expected to reach mass production in the first half of the year.
Cooper Developer Platform and expanding software tooling
Wang and other executives repeatedly pointed to a common software foundation—the Cooper Developer Platform—shared across Ambarella products. Wang said this common stack allows the company to support multiple applications with limited incremental R&D, describing the additional cost of enabling new applications as primarily customer-facing engineering support for application-level software.
In a panel discussion, executives described how the company’s toolchain evolved with changes in AI workloads. Software executive Malhar said Ambarella’s existing “CNN gen” tool has been used for a decade and that the company added support for transformer workloads by incorporating matrix multiplication into its third-generation architecture. He said Ambarella developed “VLM Gen” tooling for autoregressive models and is working on challenges such as inference-time scaling, quantization, and other optimization techniques. Malhar also said Ambarella is embracing Hugging Face workflows, enabling model conversion and deployment across products sharing a common platform.
Muneyb described Cooper as supporting multiple operating systems—real-time operating systems, ROS, embedded Linux, and QNX—under a common SDK and toolchain, and said Ambarella is leaning more into open source frameworks and “agentic” layers to accelerate application development.
Applications: security, automotive, robotics, and edge infrastructure
Management said the company has broadened beyond enterprise security into additional edge AI markets. Wang cited fiscal 2026 as a record revenue year in Ambarella’s history and said growth was supported by enterprise security and two newer markets: portable video and telematics. He also highlighted three large opportunities the company is pursuing: robotics (with emphasis on aerial drones), edge infrastructure, and autonomous driving.
In autonomous driving, VisLab General Manager Alberto Broggi discussed trends toward AI-driven, end-to-end approaches and described Ambarella’s stack as multiple connected networks rather than a single large model. He said tooling around data acquisition, automatic annotation, dataset selection, and evaluation is needed to scale the operational design domain (ODD). Broggi cited a Level 4 autonomous trucking project with Omovio (formerly Continental) that he said is targeted for SOP next year.
During Q&A, Wang said Ambarella does not intend to “do the robot” itself, citing the variety of robot types. He said the company’s strategy is to focus on volume markets such as autonomous driving and aerial drones, and to make software IP available for licensing to robotics customers.
Go-to-market shift and developer-focused expansion
Management said Ambarella is expanding from a predominantly direct-sales approach toward a channel-oriented model that includes global system integrators (GSIs), independent software vendors (ISVs), and other partners. Muneyb described the need to add a “pull motion” to complement traditional “push” design-win dynamics, arguing that software applications with affinity to Ambarella silicon can accelerate adoption. Ambarella announced a “developer zone,” a “Model Garden” of optimized models, and “agentic blueprints.” Muneyb said the company launched with two ISV partners, Cogniac and MelCX, and said the company aims to have dozens of ISVs by year-end, followed by onboarding of channel and system integration partners, with revenue impact expected in future years.
In response to an investor question on timing, Wang said the need for a new go-to-market model became clear as the company engaged more physical AI and robotics opportunities, which often involve many startups and a broader customer base than Ambarella’s historical focus on large accounts.
Operations and semi-custom interest
COO Chan described Ambarella’s tape-out history and silicon execution, saying the company has taped out more than 40 production SoCs since its first 130-nanometer device. He said 97% of taped-out SoCs have gone into production with a single spin or less, and 70% reached production with no spin. Chan also emphasized Ambarella’s long relationship with Samsung Foundry, saying the companies have worked together for 17 years and that the relationship supports early access to advanced nodes and stable wafer supply.
Chan said customers are showing “very strong interest” in custom or semi-custom SoCs, citing rising development costs at advanced nodes, scarcity of advanced SoC design talent, and the desire for silicon differentiation. Wang said the company’s first semi-custom chip is funded primarily through NRE payments and that the agreed-upon ASP is “close enough” to Ambarella’s gross margin profile, adding that future business models will be evaluated as additional projects progress.
About Ambarella (NASDAQ:AMBA)
Ambarella, Inc is a global semiconductor company headquartered in Santa Clara, California, specializing in video compression, image processing and computer vision technologies. The company designs low-power, high-definition system-on-chip (SoC) solutions that enable the capture, processing and streaming of video in a variety of embedded applications. Ambarella’s platforms combine advanced video encoding, multi-core central processing units and hardware accelerators to deliver high-resolution imaging with low power consumption.
Ambarella’s product portfolio caters to multiple markets, including security and surveillance, automotive vision, wearable cameras, drones and robotics.
