Ambarella's print passed the test. The tape still doesn't trust edge-AI silicon.

On the numbers, Ambarella did what the buy-side asked: Q2 FY2027 revenue of $108.1M came in above the $105–111M guide midpoint, non-GAAP EPS of $0.18 beat the $0.17 consensus, and the Q3 outlook of $115–124M implies a clean sequential step. The release called out record edge-AI revenue, balanced sequential growth in Auto and IoT, and 'very strong' ramp of the 5nm CV75 and CV72 SoCs — the products that took AMBA's gross margin target band from 59–62% and pushed 5-year serviceable market to $22.9B.

The disconnect is the story. The numbers say edge-AI silicon is shipping; the tape is pricing that the next two quarters won't accelerate the way the FY2032 SAM model implies.

Edge-AI silicon businesses live or die on production design wins, not reference boards. Ambarella is now sitting on the largest disclosed pipeline in its history, anchored by the Hanwha Vision long-term agreement signed May 28, 2026: a co-development and sourcing deal worth more than $800M over a period exceeding ten years, covering video security, robotics, industrial automation, and life sciences. At a roughly $80M annual run-rate, that one relationship dwarfs Ambarella's entire FY2026 revenue base of $390.7M.

All of Ambarella's growth-leading CV75, CV72, and CV3-AD685 parts are fabricated on Samsung Foundry's 5nm process — a relationship dating back to the CV3-AD685 automotive AI central domain controller announcement in February 2023. The strategic value of that pick is suddenly obvious: Samsung reported Q2 2026 revenue of KRW 171.5 trillion (a record, up 28% sequentially) and signaled foundry utilization running above 80% in Q1 2026, with a stated target of 100% utilization in H2 2026 as automotive-grade 5nm demand tightens.

The supply-chain contrast matters. TSMC booked Q2 2026 revenue of $40.2B with a 67.7% gross margin; Samsung's foundry business is rebounding off sub-70% utilization but is now constrained at the advanced nodes Ambarella needs. That gives Samsung leverage on pricing — but it also means Ambarella has secured capacity at the node its competitors increasingly want.

On July 31, 2026, the Financial Times reported — and Reuters confirmed — that NXP Semiconductors is in talks to acquire Ambarella at a valuation of 'nearly $3.3 billion.' Ambarella jumped ~19% on the news; NXP fell more than 3%. The rationale is a clean adjacency: NXP's automotive franchise (58% of mix, $1.94B in Q2 2026) plus Ambarella's low-power CVflow vision silicon equals a software-defined vehicle stack that can compete against Mobileye's EyeQ and Qualcomm's Snapdragon Ride on a single silicon-plus-perception roadmap.

The structural picture is unambiguous: FY2026 revenue of $390.7M was +37.2% YoY, edge-AI was ~80% of the mix and grew ~50% YoY, and the 6-month run-rate through July 2026 is $208.5M (+14.9% YoY) — a deceleration from the FY2026 sprint into a steadier FY2027 in the 10–15% growth band management has set.

But R&D is still 46.8% of Q2 revenue and SG&A another 18.5%. The non-GAAP gross margin held at 59.3% — inside the 59–62% target band but down 120 bps YoY as 5nm wafer costs and product mix moved. Free cash flow stays positive ($33.7M TTM) but the company burned $24.8M on a GAAP basis over the first half of FY2027, and the Q3 guide of 59–60% non-GAAP gross margin signals no near-term reprieve. Long-term guidance of FY2027 revenue growth of 10–15% is now a softer slope than FY2026's 37%.

Near-term, the catalysts are binary and dated. The first is the next NXP headline — a confirmed deal would clear the standalone-vs-premium debate; a deal dying would re-anchor on the Sept. 3 print. The second is the Q3 FY2027 print in early December, where the $115–124M guide implies 7–15% sequential growth, and management will need to defend the 59–60% gross-margin range against rising Samsung wafer costs. The third is the X7 AI accelerator, which is sampling now and targeting edge infrastructure design wins — a category with no equivalent on Mobileye's roadmap.

Over one to three years, the thesis is that the FY2032 SAM of $22.9B translates into a $1B-plus revenue company with Samsung as a captive 5nm/2nm partner, the Hanwha Vision deal alone worth ~$80M/year, and either an NXP exit or a standalone gross-margin expansion into the low 60s. The X7 and the 2nm CVAS semi-custom chip (first revenue FY2028) are the two product vectors that change the shape of the model — not just the slope. The 2nm CVAS is also why Samsung's 2nm roadmap matters: Ambarella is locking in next-node capacity before TSMC's automotive allocation tightens.