Ambarella

Founded in 2004 by Fermi Wang and Les Kohn, Ambarella built its franchise on video compression silicon before pivoting to AI inference at the edge. Its CV3-AD automotive AI domain controllers integrate an image signal processor, ASIL-D safety islands, and a neural-network accelerator in a single SoC — targeting the ADAS and autonomous-vehicle market as a low-power alternative to discrete GPUs. Its N1 series, introduced in 2024, repurposes the same CV3-HD architecture for on-premise generative AI: the N1-655 decodes 12 simultaneous 1080p30 streams and runs multiple vision-language models (Phi, Gemma, Llama) under 20 W, enabling edge AI boxes, autonomous mobile robots, and smart-city cameras. Ambarella reported FY2026 revenue of $390.7 million — a 37% year-over-year increase — with edge AI accounting for approximately 80% of that total. The company has been listed on NASDAQ (AMBA) since its IPO in September 2012 and remains independent.

Ambarella has carved out a durable position at the intersection of computer vision and edge inference, shipping silicon into mass-market automotive and security camera programs that require real AI at milliwatt budgets. Its FY2026 37% revenue growth — driven mostly by edge AI — shows the automotive ADAS and on-premise GenAI markets are pulling its products into volume. The N1 GenAI SoC line extends that moat into the fast-growing category of on-premise multimodal inference, where power constraints make Ambarella's architecture a structural advantage over GPU-based alternatives.

ambarella.com Edge AIUSApublic · AMBA

At a glance

Total raised

Undisclosed

no public funding on record

Headcount

—

no public estimate on record

Headquarters

Santa Clara

USA

Founded

2004

22 years active

Status

Public

Ticker · AMBA

Subsector

Edge AI

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Founding

50% off

/month

91 of 100 spots left

Full

$10

/month

Available once all 100 Founding Member spots are claimed.

Get full access

Already subscribed? Sign in →

Funding history

Total raised

—

Rounds tracked

Latest stage

IPO

Date	Stage	Amount	Valuation	Lead investors
Oct 10, 2012	IPO	$36M	—	—
Feb 1, 2006	Series C	$15.4M	—	Walden International, Benchmark Capital
Feb 17, 2004	Series A	—	—	Benchmark Capital, Walden International

Stock performance

How they make money

Hardware (one-time)Pattern

Ambarella is a fabless semiconductor company that generates revenue exclusively from the sale of SoCs to OEMs and ODMs — primarily in automotive ADAS, security cameras, robotics, and industrial machine vision. Chips are designed in-house and manufactured at third-party foundries (TSMC on leading-edge nodes), then sold to device makers who embed them in cameras, domain controllers, and edge AI boxes. There is no recurring subscription or service layer; each chip sale is a one-time transaction. Asia-Pacific OEMs accounted for approximately 88% of FY2026 revenue, and the company's top distributor, WT Microelectronics, represented a large share of total shipments. Average selling prices have risen as the product mix has shifted toward higher-complexity edge-AI SoCs — the CV3-AD and N1 series carry notably higher ASPs than the legacy HD video encode chips that built the company's early revenue base. Gross margins run in the high-50% to low-60% non-GAAP range, consistent with a differentiated fabless model. Ambarella does not monetize its CVflow SDK or developer toolchain separately; those are provided to OEM customers as part of the chip engagement to accelerate design cycles and deepen platform lock-in.

Generated by editorial-sourced · June 4, 2026Analytical summary · not investment advice

Founder backgrounds

Feng-Ming (Fermi) Wang

Co-founder, President, CEO & Chairman

Fermi Wang holds a B.S. in Electrical Engineering from National Taiwan University and M.S./Ph.D. degrees in Electrical Engineering from Columbia University. He began his career at C-Cube Microsystems, where he contributed to development of the world's first MPEG CODEC chip, rising to senior engineering and business management roles. He co-founded Afara Websystems — a throughput computing startup — serving as CEO, and Sun Microsystems acquired Afara in July 2002 for its server architecture. Wang co-founded Ambarella in January 2004 and has served as President, CEO, and Chairman since inception, steering the company from a digital video compression specialist to a leading edge-AI SoC vendor with FY2026 revenue of $390.7 million. He also serves on the Board of Visitors at Columbia Engineering.

Les Kohn

Co-founder & CTO

Les Kohn studied physics at Caltech alongside coursework in electrical engineering and computer science, then began his career as a microprocessor architect at National Semiconductor in 1977, co-architecting the National 32000 family. He went on to serve as chief architect of the Intel i860, then as a Distinguished Engineer at Sun Microsystems where he led the design of the UltraSPARC I and II — Sun's first 64-bit processors. At C-Cube Microsystems he was chief architect for three generations of digital video CODECs, including the first single-chip MPEG-2 and HD CODEC. He co-founded Afara Websystems as CTO, and after Sun's 2002 acquisition he remained at Sun as a Fellow developing Afara's throughput-computing technology that became the core of Sun's UltraSPARC roadmap. In 2004 he co-founded Ambarella, where he has served as CTO, shaping the CVflow neural-network processing architecture that underpins all of Ambarella's SoC product lines.

AI bull / bear

Bull case

Ambarella has built a structurally differentiated position at the intersection of computer vision and edge AI inference — a market where power envelopes, not raw compute, determine design wins. Its CV3-AD automotive domain controllers and N1 GenAI SoC series address two fast-growing adoption curves (ADAS and on-premise multimodal AI) with silicon purpose-built for single-digit-watt budgets, a constraint that rules out GPU-based alternatives in most automotive and camera deployments. FY2026 revenue of $390.7 million represented 37% year-over-year growth with edge AI accounting for roughly 80% of the mix, and management has cited approximately $1 billion in cumulative edge AI revenue across an installed base of 45 million SoCs — indicating real volume, not paper design wins. The competitive moat deepens over time through the CVflow architecture and its accompanying SDK/model-optimization toolchain: customers that port their neural networks to CVflow face meaningful switching costs, and the proprietary ISP integration (image signal processing fused with AI inference on the same die) is difficult to replicate quickly. If Level 2+ ADAS adoption continues its current trajectory and on-premise GenAI boxes gain traction as enterprises avoid cloud inference costs, Ambarella's addressable market expands meaningfully without requiring a technology pivot.

Bear case

Ambarella operates in a segment where every major semiconductor incumbent — Nvidia, Qualcomm, and Intel's Mobileye — has publicly stated edge AI and automotive vision as strategic priorities. These rivals have deeper balance sheets, broader OEM relationships, and the ability to bundle ADAS compute into larger platform contracts that Ambarella cannot match. Customer concentration is a structural vulnerability: WT Microelectronics accounted for approximately 63% of total revenue in recent periods, meaning a single distributor relationship mediates the majority of the company's sales, and any inventory correction or customer defection flows quickly to the top line. The automotive revenue thesis also carries long lead times and timing uncertainty. Ambarella wins design decisions years before a vehicle enters production, so a slowdown in OEM platform commitments — or delays in Level 3 regulatory frameworks — can create multi-year gaps between design-win announcements and revenue recognition. Additionally, roughly 88% of revenue is sourced from Asia, creating meaningful exposure to geopolitical friction, export controls, and currency volatility that could disrupt both manufacturing (leading-edge foundry concentration) and customer demand simultaneously.

Generated by editorial-sourced · June 4, 2026Refreshes quarterly · see disclosures

In the news

Ambarella · last 30 days

No articles ingested yet for Ambarella. Once the hourly news pipeline is live, every article the classifier tags as mentioning this company appears here with its one-line AI summary and sentiment.

Watch

Leadership

Feng-Ming (Fermi) Wang

Co-founder, President, CEO & Chairman

—

Les Kohn

Co-founder & CTO & Chief Technology Officer

—

John A. Young

Chief Financial Officer (appointed February 2024)

—

Products

CV3-AD Automotive AI Domain Controllers
A family of automotive-grade SoCs that integrate an image signal processor, ASIL-D safety islands, and a neural-network accelerator in a single low-power package. CV3-AD targets Level 2+ through Level 4 ADAS and autonomous driving applications and is designed as an energy-efficient alternative to discrete GPU-based compute platforms in vehicles.
N1 Edge GenAI SoC Series
Introduced in January 2024 and expanded in January 2025, the N1 series repurposes Ambarella's CV3-HD architecture for on-premise generative AI. The flagship N1-655 decodes 12 simultaneous 1080p30 video streams while concurrently running multiple vision-language models (VLMs) — including Phi, Gemma, LLaVA-OneVision, and Llama — under 20 W. Targeted at AI boxes, autonomous mobile robots, and smart-city security cameras.
CV5 Series — Edge AI Vision SoCs
Mid-range edge-AI SoCs for security cameras, drones, and industrial vision. The CV5 family integrates Ambarella's CVflow neural-network inference engine with advanced ISP and H.265/H.264 encoding, enabling real-time object detection and classification at very low power. A core volume product for the security-camera OEM market.
CVflow Neural-Network Processing Architecture
Ambarella's proprietary AI inference engine embedded across its SoC families. CVflow is a vector-processor architecture optimized for convolutional and transformer networks, providing the shared compute fabric that underpins the CV3-AD, CV5, and N1 product lines. Customers deploy CVflow-based chips with Ambarella's SDK and model-optimization toolchain.

Open roles

We don't have a live feed for this company's ATS. Their careers page has every open role.

View all careers ↗