Youdao (DAO) Q2 2026 Earnings Call Transcript

Youdao (NYSE: DAO) reported Q2 total net revenues of RMB 1.5B (+3.5% YoY) alongside a record operating profit of RMB 111.5M (nearly 4x YoY), producing an 80.7% YoY surge in operating cash flow to RMB 334.2M. Learning services grew 20.9% YoY to RMB 795.6M with gross margin rising to 65.5%, while smart devices revenue fell 31.5% YoY to RMB 86.8M (gross margin 32.8% vs 41.5% last year) amid elevated memory costs. The company highlighted AI-driven momentum—AI subscription sales of ~RMB 100M (+20% YoY) and AI-enabled advertising growth (>50% YoY for the second quarter in a row)—and announced a new share repurchase of ~7.5M ADSs for ~$33.8M as of June 30, 2026. Management guided to recovering smart-device gross margin to >40% in 2H26 as architecture changes reduce memory reliance.

Analysis

DAO is transitioning from a low-multiple “story stock” into a self-funded operating asset, which matters because the market typically rerates China consumer-internet names only when cash generation becomes repeatable, not when revenue simply ticks up. The key second-order effect is that the company is quietly shrinking its dependence on low-quality traffic and hardware while using AI to widen moat-like features in learning and ad tech; that mix is more valuable than the headline growth rate implies.

The market is likely underappreciating how much of the profit step-up is coming from mix and discipline rather than cyclical demand. That makes the near-term setup good, but also fragile: if learning retention slips or ad spend softens, the operating leverage works in reverse because the business still has meaningful sales/R&D fixed costs. Smart devices look like an intentional de-emphasis, but the margin recovery thesis there is highly execution-sensitive given component inflation.

Over 1-3 months, the main catalyst is whether September product launches convert into visible monetization, not just feature demos. Over 6-18 months, the real question is whether AI subscriptions and AI-ad tooling can offset device decline enough to reaccelerate consolidated revenue above low-single-digit growth. A fail case would be any hint that “AI” is improving engagement but not pricing power; the thesis breaks if gross margin stalls while R&D keeps rising.

Contrarian view: consensus may be too focused on the optics of profitability and not enough on durability. The move is probably not overdone if the stock still trades like a distressed edtech name, but it becomes overdone quickly if investors extrapolate one quarter of margin expansion into a structural earnings story without proof that AI monetization is recurring.

Key Decisions for Investors

• Buy DAO on any post-earnings pullback, 4-8 week horizon, with a thesis that the stock should rerate toward profitable internet peers if upcoming September product launches translate into measurable subscription/ad monetization; falsify if next quarter shows revenue growth still stuck in low single digits and operating margin fails to hold.
• Use DAO as a tactical long vs. KWEB pair trade for 1-3 months: long idiosyncratic margin expansion and buyback support, short broad China internet beta where macro/news flow can swamp fundamentals; cover the pair if KWEB outperforms on policy stimulus or if DAO loses momentum after the event window.
• Avoid chasing calls immediately after the print; instead, wait for the next catalyst check in September and consider a small long position only if management shows early conversion from AI features to paid usage. Risk/reward is better on confirmation than on the headline quarter.
• Watch for a break in DAO gross margin back below the mid-40%s or learning-services retention below 75% as the key falsifier; if either occurs, reduce exposure because the market will likely de-rate the quality of earnings quickly.
• For more speculative accounts, use a limited-risk call spread on DAO into the September AI launch window only if implied volatility compresses; target upside from product announcements, but keep size modest because the monetization path is still unproven.