AMBARELLA INC Q2 FY2025 earnings call

• New product ramps: CV5, CV7, and CV3-AD families are 5-nanometer, integrating third-generation AI inference accelerator and commanding above-average ASPs. CV5 family is ongoing and ramping, expecting over 1 million units shipped this year. CV7 family expected to enter production end of fiscal '25. CV3-AD family for L2+ autonomy. - Customer activity: Rivian's R1S SUV and R1T pickup use Ambarella's CV5 AISoC; Samsara's CM33/CM34 AI Dash cameras use CV22 SoC; BAIC, Luxeed, Honda, AXIS, i-PRO, Insta360, Moultrie use Ambarella's SoCs in various products. - Focus on AI-powered solutions: Rising demand for AI inference at edge, with new products enabling more advanced AI networks like CLIP and Vision Language Model.

Ambarella's second quarter revenue was $63.7 million, near the high end of the guidance range, up 17% sequentially. IoT represented about 70% of total revenue, with auto revenue growing slightly sequentially. Non-GAAP gross margin for fiscal Q2 was 63.3%, slightly above the midpoint of prior guidance. Receivables day sales outstanding decreased, and inventory days also decreased.

• Fiscal Q3 revenue guidance: $77 million to $81 million, with sequential growth in both IoT and Auto. - Fiscal Q3 non-GAAP gross margin expected 62.5% to 64%. - Fiscal Q3 non-GAAP OpEx expected $49 million to $51 million. - Fiscal year 2025 revenue expected mid to high teens growth versus last year, with new product ramps driving visibility in the second half.

• Global auto production forecasted down, EV OEM shakeout underway. - Mixed enterprise and consumer IoT spending. - Pricing pressures in the China auto market.

Q: You referenced vision language models. Is this a new opportunity or an opportunity that you've been addressing for some time here? And how your architecture might be different from GPU or an ASIC here. And is this a CV3 opportunity?

A: For vision language Model, we demoed at CES this year with N1 chip derivative of CV3 family. We attract customer interest for using vision language model with cameras. Can use CV72 to run smaller models for VRM.

Q: As I kind of look at your guidance for October, which was very strong, and then reconcile it with the full year guidance that you gave. It implies maybe a significantly weaker Q4 than traditional seasonality. Can you talk about maybe the moving parts here into January?

A: When we do the calculation, current guidance between Q3 and whole year reflects normal seasonality for Q4, with Q4 typically down sequentially.

Q: You talked about the outlook for the quarter, being driven by new products. Is that kind of content increases because you're migrating people to CV5? Are there new end markets or applications?

A: IoT enterprise and other IoT categories saw growth, with customers cleaning up inventory and going back to regular ordering pattern. Some customers taking CV5 into production.

Q: Can you talk about the product cycles of CV5, CV72, CV75?

A: CV5 will continue strong next year. CV72 expected to start ramping end of this year, shipping in IoT enterprise next year. CV75 ramps as mid and low-end of CV7 family. CV72 is significant ASP jump from CV22 family.

Q: How do you think the puts and takes on the growth rate of IoT versus automotive would be for next year?

A: In Q3, IoT and automotive growth similar. Next year, IoT expected to have better growth as CV72 contributes, with CV3 family to help automotive growth later.

Q: Comparison of this product cycle to past cycles?

A: Current cycle is sticky due to higher AI performance without excessive power consumption increase, differentiating from competitors.

Q: On the 10% customers, Chicony wasn't mentioned. Should we imply that means consumer is getting to be less percentage of your overall revenue?

A: IoT home category (including Chicony design) is weak due to low-end models with limited performance requirement, but gross rate slower than other segments.

Q: On the traditional video processor side of the business. Should we sort of think about that as a potential another potential lever going forward?

A: Market moved more towards AI, and Ambarella hasn't taped out video processors in recent years, so video processor sales not expected to increase.

Q: Margin profile of new products considering smaller process nodes and foundry pricing?

A: Don't anticipate margin profiles changing significantly with smaller process nodes; long-term gross margin range 59% to 62% expected.

Q: Growth in fiscal second quarter was mostly IoT. Which applications came back or did inventory burn and ordering resume in the IoT segment to help with the fiscal second quarter?

A: IoT enterprise stabilized inventory quickly and returned to regular ordering pattern, with few customers taking CV5 into production helping.

Q: Visibility into program ramps for CV3 in the '25 or calendar '26 timeframe?

A: CV3 expected to have first full year of production in calendar year 2026, with Leapmotor as initial customer.

Q: Ambarella's passenger OEM win last quarter. Has that provided a halo effect for current passenger OEM engagements?

A: Any design win helps, but each OEM design win competes on technology, pricing, and support; first win helps but not deterministic for future wins.

Q: China auto market aggressive pricing. Any negative impacts?

A: Pricing pressure in China market exists, but gross margin profile not expected to change significantly in next few quarters.

Q: How do you view seasonality?

A: Q4 down sequentially, Q1 can be down sequentially, strongest quarters Q3 and Q2 based on five-year or 10-year averages.

Q: Regarding non-industrial IoT in guidance. Is there anything regarding consumer IoT segment that helped with the guidance and the strength of the guidance in 3Q?

A: Consumer IP cam market (IoT home) is weak, but waiting to see if new functions and neural network requirements enable better fit for Ambarella's products in this segment, which didn't help Q3 guidance.