Ambarella, Inc. is positioning its chip portfolio for broader adoption of edge artificial intelligence across enterprise security, consumer video, wearables, automotive and edge infrastructure, Chief Executive Fermi Wang said at Citi’s TMT Conference.
The fabless semiconductor company is targeting fiscal 2027 revenue growth of 10%–15%, though memory availability and pricing remain risks. Automotive accounts for about 30% of revenue and enterprise security remains the largest addressed market.
Wang said customers have progressed from convolutional neural networks to transformer-based models and large language models, and are seeking more processing capability inside devices. He said the edge AI market does not yet have a single dominant application comparable with the major workloads driving data-center spending.
Founded in 2004, Ambarella first focused on digital-video applications before spending the past decade developing edge AI processing products. Its systems-on-chip combine AI inference engines with image-processing pipelines, CPUs and peripheral functions, spanning roughly 1 to 1,000 TOPS and supported by a unified Cooper software development kit that runs across the company’s 15 AI SoCs.
Wang identified power efficiency as the primary competitive advantage, describing performance per watt as central to design-win efforts. Customers may initially use GPUs because they are broadly available and easy to program, he said, but power consumption becomes a constraint for battery-powered products and systems with limited heat dissipation.
He characterized Qualcomm as the largest competitor because of greater overlap in offerings, said NVIDIA offers high performance but GPUs can be less suitable for energy-efficient needs, and described Synaptics’ edge AI coverage as concentrated in the roughly 1 to 10 TOPS range.
Security cameras are expanding beyond surveillance, Wang said, as retailers connect them to edge systems that use large language models to analyze store traffic, customer behavior and operational conditions. In consumer video, he cited demand for 360-degree cameras, sports cameras, stabilized handheld devices and drone cameras, estimating that market at roughly 10 million to 20 million units annually.
Wearable cameras, historically used by law enforcement, are drawing broader interest from retail and service businesses, though battery life and power consumption remain key constraints. Wang described the segment as growing faster than mature enterprise-security applications but from a smaller base.
Wang described robotics as high-growth but early-stage, with a small current revenue base. The company has a couple of humanoid-related design wins, he said, though he does not expect that business to materially change financial performance in the near term. Humanoid robotics is technologically difficult and likely will require multiple development cycles, he said.