Ambarella (AMBA) Q2 2027 Earnings Call Transcript

Ambarella reported fiscal Q2 revenue of $108.1 million, up 13.2% year over year and slightly above guidance midpoint, with non-GAAP EPS of $0.18 and record automotive revenue. Q3 revenue guidance of $115 million-$124 million implies a $119.5 million midpoint, supported by Physical AI and IoT demand, while gross-margin guidance remains 59%-60%. Management lifted its fiscal 2032 SAM forecast to $22.9 billion from $8.5 billion in fiscal 2027, introduced the X7 edge AI accelerator, and sees $1 billion of potential seven-year revenue from CapGemini and Macnica partnerships; however, memory scarcity and rising supply-chain costs pose a Q4 volume risk.

Analysis

AMBA's near-term rerating hinges less on the reported quarter than on whether it can convert a silicon-design-win narrative into revenue without diluting its gross-margin model. The new indirect channels expand customer access but insert partners into economics; management's unchanged margin framework is an aspiration, not evidence, until channel mix and partner compensation are disclosed. The more immediate earnings sensitivity is memory availability at AMBA customers: constrained DRAM can cap unit shipments even as AMBA's own bill of materials is insulated, creating a Q4 revenue-risk asymmetry that consensus may underappreciate.

X7 creates a credible wedge into edge inference, but it moves AMBA from differentiated vision SoCs into a crowded accelerator field versus NVDA, QCOM and numerous private ASIC vendors. Its low-power and memory footprint matter most where customers are power- or thermal-constrained; they do not automatically establish software portability, customer qualification, or pricing power. Initial sampling/design-win announcements over the next 1-3 months are likely sentiment catalysts, while material revenue should be discounted until FY2028 given the stated ramp cadence.

The non-obvious positive is that memory scarcity may accelerate substitution away from GPU-heavy edge designs toward purpose-built inference silicon, improving AMBA's design-win funnel. Conversely, elevated inventory days combined with distributor concentration can obscure true end-demand versus channel stocking; a distributor inventory correction would pressure revenue and valuation simultaneously. The small buyback authorization is not a meaningful downside floor relative to development spend and does not offset execution risk on semi-custom programs.

Key Decisions for Investors

• Maintain a tactical long AMBA only through the next design-win and Q3 guide catalysts (1-3 months), preferably entered after any post-earnings pullback; target upside requires evidence of X7 customer adoption or raised FY2028 visibility. Exit if Q4 commentary identifies memory-driven order cuts or if gross-margin guidance falls below 59%.
• For a lower-beta expression, pair long AMBA / short QCOM in equal dollar amounts for 3-6 months: AMBA has greater upside torque if edge inference shifts toward ultra-low-power custom silicon, while QCOM provides a liquid hedge against broad edge-AI enthusiasm. Stop the pair if AMBA fails to show incremental design wins by the next earnings call.
• Do not underwrite the channel partnership revenue targets in valuation until AMBA discloses booked backlog, committed purchase obligations, partner economics, and at least one named enterprise deployment. Treat these as watch-list catalysts rather than base-case estimates for the next 12 months.
• Monitor WT Microelectronics inventory and AMBA days-inventory trends at the next 10-Q. A further rise in inventory days alongside revenue growth below guidance would indicate channel loading risk and support reducing exposure or initiating a short-term AMBA short.