Financials near profitability, autonomous driving far from trust

After doing anti-fraud work for a long time, I don't think Momenta's recent big drop is because capital suddenly turned its back. It looks more like the market starting to reprice long-tail missed detections. The interim report on August 31 was pretty solid: H1 revenue of 1.602 billion yuan, up about 76% YoY, and adjusted net loss narrowed to 14.097 million yuan—just one step away from breakeven. Yet on September 1, the stock plunged 10.85%, wiping out about 6.6 billion HKD in market cap.

The reason isn't complicated. Owners reported that when using advanced autonomous driving on national roads, the IM LS6 failed to react promptly to small metal obstacles visible to the naked eye. This is like an anomaly detection model: the metrics look great on paper, rule coverage can be explained well, but a single out-of-distribution sample can shatter trust. In autonomous driving, false negatives are far more fatal than false positives. False positives just require human takeover; false negatives make people doubt the entire system.

So what Momenta needs to prove next is whether perception anomalies, takeover procedures, and accident reviews can leave auditable records, and whether responsibility boundaries and corrective actions can be traced after accidents. Financial reports can narrow losses, but one missed detection drags valuation back into the deep waters of R&D. That's it.