Ambarella at Citi’s 2026 Global TMT Conference: edge AI push

On Wednesday, September 9th, 2026, Ambarella, a fabless semiconductor company, spoke at Citi's 2026 Global TMT Conference to promote its edge AI strategy. The company's chips are designed to incorporate inference, image processing, CPUs, and other functions into a single system. Ambarella highlighted the growing potential in security, consumer video, automotive, and robotics markets, while acknowledging the risks from supply-chain constraints and the absence of a dominant edge AI use case.

The company sees edge AI as a significant long-term growth market but noted that it still lacks a breakthrough application like cloud coding in data centers. Ambarella's strengths lie in its power efficiency, video processing expertise, and a unified software kit, the Cooper SDK, across 15 AI SoCs. A $800 million semi-custom chip agreement with Hanwha Vision underscores its growing semi-custom chip model.

Fiscal 2027 revenue growth is forecasted at 10% to 15%, with upside potential tied to accelerated edge AI adoption and downside risks from memory supply constraints. Automotive, security, and consumer video are the primary revenue drivers, while robotics and edge infrastructure are still in early stages. Ambarella's CEO, Fermi Wang, emphasized that the company has spent over a decade developing edge AI and physical AI applications.

Although it has expanded beyond digital video into a broader SoC platform, Wang stressed that Ambarella aims to be the system chip for the entire device, not just a single component. The company believes edge AI is at an inflection point, driven by a demand for more AI performance across various vertical markets. However, Wang admitted that investors often seek a single vertical to drive the business, which Ambarella does not yet have.

The company's strategy is built around three core pillars: performance per watt, video processing capabilities, and the Cooper SDK. Ambarella's AI portfolio ranges from 1 TOPS to 1,000 TOPS, offering a broader scope than some competitors. Comparatively, Ambarella is positioned between NVIDIA and Qualcomm in terms of raw performance and power efficiency.

In fiscal 2026, Ambarella reported $400 million in annual revenue, which grew by 20% year-over-year, with a gross profit margin of 58.52%. The company's revenue is largely driven by IoT (75-78%), enterprise security and consumer video (45%), wearables (10%), and automotive (30%). Its average selling price for SoCs is approximately $15, and management expects it to increase as AI performance requirements rise and larger die sizes become necessary.

Ambarella's gross margins are projected to remain relatively stable despite rising ASPs, as customers are anticipated to pay higher prices for enhanced AI capabilities. The company's AI SoC portfolio includes 15 chips, and it has an exclusive foundry relationship with Samsung, along with exclusive OSAT partnerships. Ambarella's first 2nm chip is currently in production, and the company has developed a flexible architecture that enables customers to upgrade their systems without starting from scratch.

As AI technology evolves rapidly, with shifts from CNN networks to transformers and large language models, customers express concerns about their AI chips not supporting future workloads. Ambarella's response is a scalable platform that can adapt to changing AI needs without requiring a complete redesign. In enterprise security, the company observed a transition from a simple security function to a comprehensive analytics platform.

With LLMs and other AI tools, the same hardware can now collect behavior data, marketing data, and operational metrics. In consumer video, Ambarella identified a "sweet spot" for its technology, estimating the category at 10 million to 20 million units annually, encompassing 360-degree video solutions.