Fermi Wang
Analyst · Morgan Stanley
Thank you, Louis. And good afternoon. Thank you for joining our call today. We achieved significant milestones during fiscal 2023 in our ongoing transformation to an edge AI company. On the technology and product execution front, we broadly sampled the first of our 3rd generation edge AI SoCs, CV3-AD, commenced mass production of our first 5nm SoC, CV5, introduced our centralized 4D imaging radar, and took an important step towards the commercialization of our automotive software stack IP. On the business front, we announced strategic partnerships with global automotive leaders such as Bosch and Continental, and customer interest and activity with our edge AI products continues to grow. As of the end of the fiscal year, we have cumulatively shipped more than 13 million computer vision SoCs to more than 325 unique customers, with more than 230 customer products in production, an increase of more than 50% from a year ago. CV represented more than 45% of our fiscal 2023 revenue and was a factor driving our firmwide SoC average selling price up. In fiscal 2024, we expect CV to continue to grow in absolute dollars to represent about 60% of total revenue. This is expected to continue to drive our ASP higher. Fiscal 2023 was not without its challenges, and as the year progressed, cyclical and economic headwinds became more significant, negatively impacting our revenue results and outlook. Despite the challenges in the second half of fiscal 2023, we reported record annual revenue of $337.6 million, up about 2% year-over-year. With our technology leadership and the scalability of it increasingly recognized by customers globally, we don't believe the current softness has anything to do with the technical competitiveness of our new technology and products, and we remain committed to build-out a complete edge AI portfolio of SoCs and software. I would like to highlight some of the customer engagements that cause us to be so encouraged about our long-term outlook. During Q4, we took two important steps toward the commercialization of our software IP. First, and building upon Continental's November announcement that it would offer ADAS solutions based on Ambarella's CV3 AI central domain controller, on January 4th, we announced Continental and Ambarella were extending their cooperation to include full software stack development for ADAS and autonomous driving applications. The two companies will jointly develop scalable, end-to-end hardware and software solutions, with Ambarella software IP embedded in Continentals' full stack. Second, we announced in December the world's first centralized 4D imaging radar architecture. Combining Ambarella's Oculii adaptive AI radar software and the efficient CV3 AI domain controllers enables the central processing and fusion of raw 4D imaging radar data with other sensor inputs, including cameras, lidar and ultrasonics. We believe this breakthrough architecture provides greater environmental perception and safer path planning in AI-based L2 to L4 autonomous driving systems. Business development momentum with our CV3 family was strong in the quarter too. In December, Bosch announced it plans to adopt the CV3 family of AI central domain controllers for the realization of next generation ADAS functions. Bosch requires efficient and flexible high performance computing, as well as scalability to enable software stacks to be re- used from L2 to L4 applications. Bosch stated that they are adopting Ambarella's CV3 family because it is the perfect fit to accelerate time-to-market and to create scalable solutions for their customers. During CES, we announced CV3-AD685, the first production SoC in the CV3 family of AI central domain controllers for L2 to L4 vehicles. The third generation CVflow AI engine in CV3-AD685 includes neural network processing that is 20x faster than the previous generation of CV2 SoCs, along with additional general vector processing capabilities to provide the overall performance required for full autonomous driving stack processing, including computer vision, 4D radar, deep fusion and planning. We are proud to note next generation transformer networks have been successfully deployed by customers onto CV3, which highlights the flexibility and efficiency of the architecture for autonomous driving and potentially other applications in the near future. We had a highly successful CES 2023 with more than 200 customer meetings and we demonstrated a wide range of automotive and IOT solutions, including many joint demonstrations with our automotive tier 1 and software partners. Our CV3 won a CES 2023 Innovation Award in the Embedded Technologies category. We offered demo rides in our latest EVA cars featuring a single CV3 central domain controller performing all perception, classification and path planning functions, and processing both camera and radar data. The EVA car included 15 cameras and 9 radars with industry leading power efficiency. And together with Continental, we demonstrated a single CV3 connected to six 8 megapixel cameras and four 3 megapixel cameras, while running multiple neural networks on each video stream, utilizing about 10% of the CV3-AD processor. Other CES highlights included Kodiak Robotics, which exhibited its self-driving truck integrating Ambarella's CV2AQ AI vision processor in the sensor pods on both sides of the truck for the camera perception processing. Autobrains and Seeing Machines demonstrated a combined front ADAS and driver monitoring solution running on a single CVflow SoC. The companies' joint offering provides automakers with a streamlined, single-box, multi-camera solution, including up to a 8 megapixel forward-facing camera and 5-megapixel in-cabin camera. In February, China-based Hyperview, an autonomous driving technology company, announced that it had selected Ambarella's CV3-AD family to develop high performance computing autonomous driving platforms. The development will pair CV3 with Hyperview's software stack to provide production-ready perception, automated driving and parking solutions. Hyperview has previously achieved mass production with multiple Chinese OEMs, and we are pleased they have chosen to incorporate CV3 into their new products. I will now discuss some of our customers' product introductions during the quarter. We are extremely excited to announce our first 4D imaging radar project will soon be entering mass production at Geely's Lotus Technology unit. Geely intends to leverage the Lotus technology and brand into the market for higher volume passenger vehicles. Lotus announced its Eletre pure electric SUV which includes two 4D imaging Radars based on Ambarella's Oculii radar technology, one in the front and one in the back, providing high resolution imaging capable of detecting and tracking vehicles within 300 meters and pedestrians within 100 meters, and accurately measuring the speed and distance of objects. After announcing the CV3 relationship with Bosch Mobility, we were excited to extend the relationship to the CV2 family of SoCs. During CES, Bosch introduced its new connected RideCare platform and the companion fleet camera. Based on our CV25 AI processor, the product was honored as a CES 2023 Innovation Awards Best of Innovation honoree in the In-Vehicle Entertainment & Safety category. Amazon Ring at the show also introduced its Ring Car Cam based on our H22. The Car Cam features dual-facing cameras and can monitor the car's interior as well as the road. Garmin also announced the CV25 based Dash Cam Live camera at CES. The camera provides 140-degree, 1440p HD views and includes live view, theft alerts and location tracking. In the enterprise security camera market, European market leader Axis introduced a number of cameras based on our CV25. The M3088-V compact dome camera is an 8-megpixel design with WDR processing and support for analytics with deep learning at the edge. And the new small form factor M1055 and M1075 box cameras target small stores and residential care applications and feature edge-based deep learning and 1080P. Korean market leader Hanwha introduced three new enterprise class models in its 9000 cameras series all based on our CV2 AI vision processor. The 4K models include box, bullet and panoramic designs and feature AI-based analytics, including object classification, license plate recognition and car make and model recognition. Korea-based IDIS also introduced its 6281HX speed dome camera based on our S3L vision processor and featuring 30X optical zoom capability. These representative engagements illustrate how we are successfully leveraging our state-of-the-art video processing heritage into larger, more diverse and higher quality markets for high bandwidth processors and software for machine sensing applications. A majority of the early growth we have seen with our edge AI products to-date has been for new product cycles in existing IoT markets, like the Axis and Hanwha Techwin enterprise security camera engagements I just discussed, but we are encouraged with our early success in critical new areas, like the CV3 family of central domain controllers for L2 to L4 mobility applications. CV3, in a single SoC, synergistically leverages the functionality Ambarella has created over the years – camera and radar perception, deep learning AI, and automotive software stack IP. While we are experiencing near-term pressure on our revenue from cyclical and economic factors, we firmly believe our secular growth opportunity remains very positive, and we remain focused on leveraging our leadership position in edge AI. AI is just beginning to transform so many industries, and we expect AI adoption to accelerate over the coming years, in particular in the AI edge inference market that is our area of focus. Now, I will now hand it over to Brian. Thank you.