Ambarella Lifts Five-Year Market Forecast to $22.9 Billion as Edge Infrastructure and New Accelerator Chip Reshape Growth Story
Q2 FY2027 earnings call, September 3, 2026
Ambarella used its fiscal second-quarter call to unveil a meaningfully larger long-term opportunity, raising its rolling five-year serviceable addressable market forecast to $22.9 billion in fiscal 2032 from $8.5 billion in fiscal 2027, a roughly 20% compound annual growth rate. That compares with the prior forecast issued in May 2025, which called for an 18% CAGR through fiscal 2031. The revision comes with a notable mix shift: IoT markets now represent about 70% of the terminal year figure, up from a forecast where automotive carried a slightly higher proportion. CEO Fermi Wang attributed the bulk of the upward revision to a new product category the company is entering: standalone AI accelerators and AI systems-on-chip aimed at on-premise "edge infrastructure" deployments, a market Ambarella had not previously broken out as a distinct growth driver.
A New Chip Category: X7 Accelerator Targets On-Premise Inferencing
The headline product news was the preview of Ambarella's first standalone AI accelerator, referred to internally as X7, which is sampling now and expected to land initial design wins in edge infrastructure applications. Unlike Ambarella's traditional camera-centric SoCs, X7 is designed to function as a coprocessor alongside host processors such as ARM or x86 chips, giving customers a way to add AI performance without redesigning their boards. Wang explained the distinction the company is drawing internally: an edge AI SoC integrates camera perception, AI acceleration, CPUs and encoding into a single chip, whereas an AI accelerator is not camera-specific and targets a broader range of digital and physical modalities, which management views as essential for enterprise operational-efficiency use cases.
The competitive framing here matters for investors. Wang acknowledged that roughly 50 startups are chasing this space alongside Nvidia and Qualcomm, calling it "a crowded space." Ambarella's pitch to customers rests on power efficiency rather than raw throughput, with Wang noting the accelerator needs only 4 megabytes of memory to run large language model workloads within a 4-to-5-watt power envelope in current design wins. Given the memory-driven cost inflation squeezing the entire industry, that footprint argument is becoming a more central part of the company's sales pitch, though it remains unproven at scale.
Memory Shortage Reframed as a Structural Tailwind, Not Just a Headwind
Management spent considerable time addressing the industry-wide memory supply crunch, driven by vendors prioritizing AI data center demand, which Wang said is "impacting the entire industry" and forcing Ambarella itself to pass rising supply chain costs on to customers to protect its long-term gross margin target of 59% to 62%. CFO John Young reported non-GAAP gross margin of 59.3% for the quarter, below the midpoint of prior guidance. Importantly, Wang clarified in response to a question from BofA's Liam Pharr that memory pricing does not directly hit Ambarella's own gross margin since the company neither buys nor resells memory. The real risk is secondary: if customers raise end-product prices to absorb memory costs, unit volumes and therefore Ambarella's order volumes could suffer. Management said it saw only modest impact in the second and third quarters but flagged genuine uncertainty heading into fiscal fourth quarter, with Wang saying the team is "continuing to monitor this situation very closely, by talking to customers all the time." Notably, Wang is positioning the shortage as a demand catalyst for Ambarella's lower-memory-footprint architecture, arguing that customers who initially prototyped with GPU-based solutions are now reconsidering given both cost and power constraints, a dynamic that could accelerate design win conversion if it plays out.
Two New Channel Partnerships Aim at $500 Million Each Over Seven Years
Ambarella disclosed two multiyear go-to-market partnerships designed to build out an indirect sales channel that essentially did not exist before, since virtually all revenue today comes through direct sales. The company signed a strategic partnership with Capgemini targeting large enterprise customers seeking to move edge AI and Physical AI deployments from pilot to scale, and a seven-year distribution agreement with Macnica aimed at smaller, more fragmented customers across verticals including drones, retail, robotics and industrial IoT. Wang quantified the ambition directly, telling analysts the company is expecting "the range of revenue we are expecting from this collaboration is $0.5 billion with each one of them" over the seven-year term. Executive Muneyb Minhazuddin characterized the two partnerships as complementary but structurally different: Macnica is a volume play aggregating many small design wins that were previously inaccessible to Ambarella's direct sales model, while Capgemini is a value play built around fewer, larger and more complex enterprise deployments. Both partners will build on Ambarella's existing Cooper software platform rather than developing competing stacks, which Wang described as the primary reason the partners chose to work with Ambarella in the first place. Management was candid that this channel build-out will take time, guiding investors to expect meaningful revenue contribution in two to three years rather than an immediate inflection, even though some early design wins are already generating revenue.
Semi-Custom Silicon Adds a Third Growth Vector
Alongside organic product development and channel expansion, Ambarella confirmed its first semi-custom chip project, a 2-nanometer CV8 SoC expected to generate initial production revenue in fiscal 2028, with additional semi-custom discussions underway with other companies. Wang described the model as building customer-specified chips while retaining the right to sell into non-competing accounts, an approach that leverages Ambarella's existing IP blocks for AI acceleration, image signal processing, encoding and CPU functions. CFO John Young clarified that a $9 million reduction in GAAP R&D expense this quarter, tied to a canceled customer development project, was unrelated to the semi-custom pipeline and involved a legacy automotive autonomy program the company had been negotiating to unwind for some time.
Robotics Pipeline Advances but Remains Early-Stage
Ambarella reiterated that its robotics design win pipeline, previously sized around 15 wins worth roughly $100 million, continues to grow, though management declined to provide an updated breakdown. A new design win with a CV72-based quadruped robot was cited as evidence of expanding traction. Wang drew a direct analogy to the early days of autonomous driving roughly seven years ago, when automotive customers cobbled together individual modules before eventually consolidating around integrated domain controllers, arguing robotics customers are at a similar inflection point today, currently favoring standalone perception modules but with clear intent to eventually adopt integrated domain controller architectures. Ambarella says it is building products for both approaches, but the more lucrative domain controller opportunity remains a future rather than current revenue driver.
Near-Term Guidance Solid, Automotive Hits a Record
Underlying the strategic announcements, the quarter itself was a modest beat, with revenue of $108.1 million landing slightly above the midpoint of prior guidance, up 13.2% year-over-year, driven by a new record in edge AI revenue and automotive revenue reaching an all-time high on commercial vehicle adoption strength. Non-GAAP EPS came in at $0.18. Guidance for fiscal third quarter calls for revenue of $115 million to $124 million, with gross margin guided to 59% to 60%, essentially seasonal performance that management said assumes no material deterioration from the memory situation. The company's board also authorized a new $50 million share repurchase program through June 2027, though no shares were repurchased in the just-completed quarter.