AMBA Semiconductors · Edge AI · Fabless chips · Auto tech · Thesis updated July 2, 2026

Diversification is improving, but growth is slowing

Ambarella is a fabless chip designer, which means it designs chips and pays outside foundries to make them.
Revenue was $100.4 million in Q1 FY27, up 16.9% from the same quarter a year earlier.
Management still expects FY27 revenue growth of 10% to 15%, a sharp slowdown from 37.2% growth in FY26.
The Hanwha long-term agreement could be worth more than $800 million over more than 10 years, but the near-term ramp is still unclear.
WT Microelectronics fell to 61% of Q1 FY27 revenue, down from 70% for FY26, which is a real but early sign of customer diversification.

01 Running thesis

A cleaner mix, not a clean win

Ambarella is trying to turn a lumpy chip design-win business into a more predictable edge AI platform. The key move is long-term agreements, or LTAs. These are multi-year customer deals that can tie Ambarella deeper into a customer's future products.

The bull case got better this period. WT Microelectronics, Ambarella's main distributor, was 61% of Q1 FY27 revenue, down from 70% for FY26. Automotive also hit a new revenue record, while enterprise security grew sequentially. That suggests Ambarella is not only leaning on consumer cameras and one distributor.

The bear case is still serious. Management reiterated FY27 revenue growth of 10% to 15%, far below the 37.2% growth reported in FY26. The Hanwha LTA is large, with potential revenue above $800 million over more than 10 years, but that does not answer how much arrives in FY27 or FY28.

Finn's view should stay mixed. The technology story is strong, and the revenue mix is improving. But the stock still needs proof that auto, enterprise IoT, and LTAs can offset slower growth, weak consumer IoT, margin pressure from advanced chips, and very high customer concentration.

Jun 2026▲The Q1 FY27 10-Q showed WT Microelectronics fell to 61% of revenue from 70% for FY26. That is early proof that automotive and enterprise IoT are helping diversify the business.

May 2026▲Management introduced a major Hanwha LTA with potential revenue above $800 million over more than 10 years. Automotive revenue also set a new record, but FY27 growth guidance stayed at 10% to 15%.

Mar 2026→The FY2026 10-K confirmed 37.2% revenue growth and stronger language around physical AI and transformer networks. It also named Insta360 as the largest end customer, adding a clearer concentration risk.

Feb 2026▼FY2026 was a record year, but management guided FY27 revenue growth to only 10% to 15%. The thesis became more cautious until new sales channels, semi-custom work, and auto growth show larger revenue impact.

Dec 2025→The Q3 FY26 10-Q confirmed strong AI-driven revenue growth. It also showed gross margin pressure from higher manufacturing costs and lower high-margin NRE mix.

Nov 2025▲Ambarella reported record Q3 FY26 revenue of $108.5 million, up 31.2% year over year. Management raised full-year growth guidance to 36% to 38% as edge AI products reached about 80% of revenue.

Sep 2025→The Q2 FY26 10-Q supported the AI chip growth story. WT Microelectronics was 67% of first-half revenue, keeping concentration as the main risk.

Aug 2025▲Management raised FY26 revenue growth guidance to 31% to 35% and said edge AI would be about 80% of revenue. Customer concentration worsened, with WT Microelectronics rising to 71% of revenue.

02 Business model

Design wins pay later

Ambarella sells system-on-a-chip products, or SoCs, to device makers, original design manufacturers, car suppliers, and equipment makers. A SoC is a full computer system on one chip. Ambarella focuses on low power AI at the edge, meaning AI that runs inside a camera, vehicle, robot, or sensor instead of in a distant data center.

The old model depends on winning designs years before the product ships. That can create strong revenue when a device sells well, but it can also make sales uneven. A lost design win, a delayed launch, or weak end demand can hurt revenue fast.

The new layer is LTAs. The Hanwha deal is the clearest example, with potential revenue above $800 million over more than 10 years. These deals could make revenue more visible and deepen co-development work, but investors still need the ramp schedule and margin terms.

Ambarella is fabless. That keeps the company focused on design, but it also means it depends on outside foundries, mainly Samsung, for manufacturing. Costs tied to advanced process technology have already pressured gross margin.

03 Product portfolio

Chips built for edge AI

Growth engine

CVflow architecture

CVflow is Ambarella's main AI engine for computer vision. The third generation can process transformer AI networks, which are used in advanced reasoning models.

Growth engine

CV7 family

CV7, CV72, and CV75 serve high-performance edge AI devices that need strong vision processing without using too much power.

Option

CV3 automotive family

CV3 targets L2+ to L4 autonomous driving and central domain control. It is meant to help cars perceive, plan, and act.

Option

N1 generative AI SoC

N1 is built for generative AI at the edge. Ambarella says it can run large language models up to 34 billion parameters on edge devices.

Option

Centralized 4D imaging radar

This radar architecture uses adaptive AI software and links with the CV3 family. It gives vehicles another high-resolution data stream for sensor fusion.

Steady

Legacy and consumer IoT chips

Consumer IoT includes products such as action cameras and other video devices. This business can still matter, but it showed a double-digit sequential decline in Q1 FY27.

04 Business segments

Still mostly IoT

IoT75%modest

Automotive25%growing fast

The mix is from Q1 FY27 management commentary: IoT applications were about three-fourths of revenue, with Automotive making up the balance. WT Microelectronics was still 61% of revenue in the same quarter, so end-market mix does not remove concentration risk.

05 Risk factors

What could break the story

WT Microelectronics concentration

High impact · High odds

WT Microelectronics was 61% of Q1 FY27 revenue and 70% of FY26 revenue. That means one distributor still touches a huge share of Ambarella's sales, even after the recent improvement. If orders through WT slow, the reported revenue hit could be large.

We watchWT Microelectronics as a share of quarterly revenue.

Insta360 end-customer exposure

High impact · Medium odds

Ambarella disclosed Arashi Vision, known as Insta360, as its largest end customer. That adds a second layer of concentration beyond the distributor. A weak action camera or drone cycle could hit demand before auto and enterprise IoT are large enough to offset it.

We watchInsta360 product launches, sell-through, and any filing updates on top end-customer concentration.

FY27 growth slowdown

High impact · High odds

FY26 revenue grew 37.2%, but management guided FY27 growth to 10% to 15%. That is a sharp slowdown. If investors were paying for a fast AI growth curve, slower sales could pressure the stock even if the company keeps winning designs.

We watchQuarterly revenue guidance and whether management keeps, raises, or cuts the 10% to 15% FY27 growth range.

LTA ramp and margin mystery

Medium impact · Medium odds

The Hanwha LTA could bring more than $800 million over more than 10 years. The missing piece is timing and profit quality. Volume pricing, nonrecurring engineering revenue, and co-development costs could change the margin story.

We watchManagement comments on Hanwha revenue in FY27 and FY28, plus any LTA gross margin or NRE details.

Auto competition

Medium impact · High odds

Ambarella competes with much larger chip companies, including NVIDIA, Qualcomm, and Mobileye. These rivals have deeper budgets and strong carmaker ties. Ambarella must prove that its low-power edge AI advantage can keep winning real vehicle programs.

We watchNew automotive design wins, lost programs, and customer names in telematics, safety, and autonomous driving.

Foundry and trade risk

Medium impact · Medium odds

Ambarella depends on third-party manufacturers, mainly Samsung, to make its chips. Advanced chips can cost more to build, and the company has already cited higher manufacturing costs. U.S.-China trade tension also matters because revenue is highly concentrated in Asia.

We watchGross margin commentary, purchase commitment charges, export controls, and China-related sales restrictions.

06 Quick answers

In one breath

What does Ambarella actually sell?

Ambarella sells low-power AI chips called SoCs. These chips help cameras, cars, robots, and sensors understand video and other real-world data near the device.

Why does customer concentration matter for AMBA?

A large share of revenue runs through WT Microelectronics, and Insta360 is the largest end customer. If either demand channel weakens, Ambarella may not have enough diversified revenue yet to fill the gap.

Is the Hanwha deal already in revenue?

The Hanwha agreement has potential revenue above $800 million over more than 10 years. The open question is how much revenue arrives in FY27 and FY28, and what margin that revenue carries.

What is the main bull case for Ambarella?

The bull case is that Ambarella becomes a key edge AI chip supplier for security, cars, robotics, and other devices. If automotive growth and LTAs keep expanding, the business could become less dependent on consumer IoT and one distributor.