安霸 2025年度报告

May 2026 Dear Shareholders: Continuing the momentum seen in fiscal 2025, we established a new revenue record of $390.7 million in fiscalyear 2026. This represents a 37% increase over the prior year and outpaced the growth of the overallsemiconductor industry for the second consecutive year. Edge AI remained our key revenue driver, withapproximately 80% of our fiscal 2026 revenue attributed to edge AI applications. We have shipped a cumulativetotal of approximately 45 million edge AI SoCs into a range of end markets, and we believe we are wellpositioned as the edge AI market continues to accelerate. The trend of AI processing at the edge continues to evolve and expand While AI processing in data centers has received much attention, we see a growing shift toward performing AIinference processing at the edge - closer to the sensor and the point of action. The edge AI market is still in theearly stages, but there is meaningful progress in models, tools and applications that can run efficiently on-device.Techniques such as distillation and mixture-of-experts (MoE) are helping enable models that are “smaller yetsmarter,” which we believe can broaden adoption from early use cases to more mainstream deployments. Overtime, we expect more edge and physical AI systems to operate in real time, act proactively and support closed-loop decisions for end users. Our highly efficient processor, when implemented within the edge of a computing network, reduces latency,improves power consumption, reduces data communication expense, and improves privacy and security. Webelieve this gives us significant advantages in the edge AI market. We are establishing Ambarella as an edge AI platform leader To efficiently address the expansion in the number of edge AI use cases, and to help our customers reusesoftware and scale their business, a broad edge AI platform is required. We have an installed base ofapproximately 45 million edge AI SoCs with more than 370 unique customer AI products having reachedproduction and approximately $1 billion in cumulative edge AI revenue. The breadth of our platform is alsoexpanding, with 12 edge AI SoCs already available, delivering performance levels supporting AI models with10’s of thousand to 34 billion parameters. Our edge AI software development platform, used by our customers tocreate software applications that run on our AI SoCs, is a critical development tool and has reached productionsupporting more than 200 different AI model architectures. We expect growth to be driven by expanding use-cases for edge AI and higher ASP products AI inference SoCs now represent a majority of both our automotive and IoT revenues and have contributed to ahigher blended average selling price (“ASP”). We expect continued revenue growth in our edge AI inferenceSoC business driven by (1) expansion of edge applications that are implementing AI functionality and (2) animproving product and customer mix as demand increases for higher-performance AI processing. The edge AI inference market commenced in the enterprise market (e.g. physical security) and is now extendinginto both new enterprise and consumer markets. We intend to expand our edge AI business from the endpoint(the terminal device in a network which represents a vast majority of our revenue today) into the edgeinfrastructure market, which will allow us to address incremental and higher value market opportunities. Our corporate SoC ASP was approximately $15 in fiscal 2026, and we expect ASP to continue to trend upwardover time, driven primarily by increased AI inference content in our SoCs and growing demand for higher-performance solutions. This demand is being supported by several factors, including: workloads shifting from the cloud/data center to the edge; larger and more capable models (including transformer-based networks); the movefrom serial to parallel (AI) processing; incremental processing to support both CNN and transformer networks;and an increase in both the number of sensors per system and the increasing resolution of those sensors, whichrequires more data to be processed at the edge. Introducing new products continues to be critical Our 5nm products represented our key growth driver in fiscal 2026. Sales of our CV5 SoC family grew well infiscal 2026. A new wave of revenue, from our CV75 and CV72 AI SoCs, reached production status for the firsttime in the fourth quarter of fiscal 2025, posted strong growth in fiscal 2026, and is expected to deliver stronggrowth in fiscal 2027. Our recently introduced 4nm CV7 edge AI SoC provides superior perception (datacollection) capabilities with a corresponding increase in AI inference horsepower, while supporting bothadvanced CNN and the more advanced transformer AI networks, which enable GenAI. CV7 commands anaverage selling price well above our current ASP, and we expect first revenue from this new product toward theend of fiscal 2027. Two important new go-to-market strategies Today, our reve