Dolphin Research

Trip.com 2Q26 First Take: Results were broadly in line with market expectations. More importantly, regulatory issues have been finalized, removing the biggest uncertainty.

As guided earlier, total revenue growth was just above 5%, a clear slowdown vs. the prior 10–20% range. Domestic hotel and travel monetization was cut following regulatory penalties. Air ticketing demand softened on geopolitical factors in the Middle East and Japan and higher oil prices.

At first glance, headline OP turned negative, driven by an approx. RMB 5.2bn regulatory fine this quarter. Ex the one-off, adj. profit was about RMB 4.4bn, down 6.5% YoY and broadly in line with the Street. This highlights near-term profit pressure for Trip.com.

Dolphin Research believes that with regulation now settled after this print, the harsh de-rating phase is over. The next stage should be more measured, as management intentionally lowers monetization and the hotel/travel cycle stays soft, implying a period of negative profit growth. After a base-building phase, we may not see an immediate snap-back, but the stock should return to a gradual long-term uptrend.$Trip.com(TCOM.US) $TRIP.COM-S(09961.HK)

Oracle 1QF27 First Take: results remained in a 'go big' mode. OCI revenue growth accelerated again, with YoY growth above 120% this quarter.With OCI's mix rising, its lift to total revenue became more pronounced, reaching 30% this quarter. Capex hit a fresh record at 28.5bn (+73% QoQ), far ahead of Street expectations, pushing net cash outflow further underwater.

The core driver: 0.85GW of compute came online this quarter, slightly below the prior 1GW guide but still a clear acceleration vs. last FY's 1.2GW deployed for the full year.We also infer that recent broad-based increases in cloud rental pricing contributed. Bottom line: AI cloud growth was strong and beat expectations.

However, prior weaknesses persisted. Non-OCI legacy businesses remained sluggish with decelerating growth, and GPM fell sharply and missed estimates given the higher OCI mix and rising depreciation.Overall, the good is quite good, and the bad remains bad. The market is focused on the outperforming AI cloud.

On guidance, management expects revenue growth to further accelerate next quarter and modestly beat the Street. Growth is expected at 68% vs. 61.5% this quarter.Driven primarily by cloud (OCI). However, profit guidance is slightly below consensus. $Oracle(ORCL.US) $ORCL 1X Short ETF(ORCS.US) $ORCL 2X Long ETF(ORCX.US)

$Meta Platforms(META.US) Zuckerberg’s AI personal assistant vision, discussed for nearly three years, has finally launched. Dolphin Research shares takeaways vs. peers:

1) Overall OK: Meta Muse is essentially a mainstream personal assistant for everyday multi-tasking, akin to a consumer-oriented Openclaw. It isn’t radically novel, but testing shows it’s fast and smooth thanks to ample compute and a model designed for multi-task workflows, with a very low usage barrier. With Meta’s 3bn social graph, it can scale virally.

Muse is accessible via App, PC client, Web, and Meta AI glasses, spanning multiple devices. Its entry-point coverage rivals China’s Doubao and Qianwen AI assistants.

2) Privacy-first: Muse addresses Western users’ privacy concerns. It is shipped as a standalone app rather than embedded in a super-app, creating a sense of separation from social data even if personalized signals underpin training. On the backend, Muse runs each user on a dedicated cloud VM, isolating interaction data.

Sensitive information such as account passwords and card numbers is neither accessible nor stored. This is more compliant than the Instinct standalone AI assistant that trended months ago.

Despite the privacy stance, Muse can still remember key habits and preferences to proactively suggest actions. That builds trust while leaving room for future monetization.

3) Monetization: Muse is live for users aged 18+, with tiered pricing purely by token consumption and no feature gating. It also takes commissions on merchant GMV and leverages Meta’s core strength in ads.

Early user feedback is positive, but we remain cautious on near-term direct C-end monetization. ChatGPT serves more complex productivity use cases, yet with ~1bn WAU and high stickiness, paid penetration is only ~5%.

We therefore do not expect subscriptions to contribute meaningfully near term, and most users will likely stay on the free tier (~100mn tokens/week). Even assuming 80% penetration of 3bn users, 2% pay rate, and ARPPU of $20/mo, annualized revenue would add ~$11.5bn, just ~5% incremental vs. Meta’s current ~$250bn top line.

More meaningful are fortifying the ads moat (defense) and gaining commission upside (offense). The former can cushion revenue growth concerns amid heavy AI investment and high base, plus TikTok/OpenAI competition, driving sentiment-led rerating from an 18x P/E (~$1.3tn, on tax-after OP of ~$71bn in 2026 at a ~20% effective tax rate) toward a 25x mid-point (~$1.8tn).

The latter is subject to risk appetite and can be priced by sentiment, which historically peaked above 28x. Ecosystem integrations are underway with platforms like Duffel, OpenTable, and Shopify.

That said, we do not expect an immediate move above the valuation mid-point. Near-term fundamentals remain a drag, with profit pressure visible this year, and relief likely starting next year via cloud compute leasing plus Muse monetization.

4) Competitive impact: If Muse continues to leverage social distribution, the biggest immediate impact is on OpenAI’s ToC logic, where free users would also push OpenAI to monetize via ads and commissions. Second, it weighs on Google’s active search conversion value; unlike our prior doubts on OpenAI’s ad monetization, Meta’s ad infrastructure is far more mature, making budget testing shifts easier to attract.

$Alphabet(GOOGL.US) $TENCENT(00700.HK)

Yesterday, $Bilibili(BILI.US) and $TENCENT(00700.HK) announced major moves — BILI issued a convertible bond and launched a buyback, while Tencent fully exited its equity stake and subscribed to part of the CB. Dolphin Research walks through the mechanics first, then the impact.

1) Execution: viewed from BILI and Tencent

(1) BILI: net raise of $400mn

BILI is issuing $700mn of CBs in total, with Tencent taking $200mn and the remaining $500mn placed in the market. Concurrently, it conducted an equity placement of ~7mn shares worth $100mn, and will offset the placement via a matching $100mn repurchase and cancellation.

In addition, for the 40mn shares Tencent is selling (~$600mn), BILI will purchase $200mn worth.

Net-net, BILI raises $700mn and spends $100mn (placement hedge) plus $200mn (buying part of Tencent’s block), resulting in a net financing of $400mn.

(2) Tencent: fully exits equity, retains debt plus option-like upside via potential conversion

Before the deal, Tencent held 40.01mn BILI shares, a 9.6% stake (our earlier read on a 6% cut in May was incorrect). Post-transaction, Tencent exits its entire equity position, accepting a ~5% placement discount as the cost.

By subscribing to $200mn of CBs, it effectively keeps a debt position equal to roughly one-third of its prior equity value, with upside optionality via potential conversion.

2) Impact on BILI: near-term choppy, long-term constructive as the Tencent overhang is removed

(1) Tencent’s sale plus concurrent placement will still cause near-term volatility despite the buyback hedge

BILI’s CB carries a zero coupon and a 35% conversion premium vs. the reference price. In today’s high-rate environment, that does not appeal to traditional long-only investors. Hence the concurrent equity placement to help CB buyers establish initial short positions (delta placement).The logic is as follows: typical buyers are arbitrage funds, who go long the CB and short the underlying to hedge single-direction risk. They are not taking a directional view; they seek to monetize volatility.

Borrowable float in BILI is limited and borrow costs are high (China ADR short demand is elevated and dual listings fragment the borrow pool). If multiple CB buyers source borrow simultaneously in the open market, it would push borrow costs even higher.Therefore, the underwriters aggregate borrow supply via a delta placement to lower the borrow cost for CB buyers.

To mitigate the initial short interest’s impact on price moves, BILI executed a matching $100mn repurchase. While this helps offset shorting pressure in aggregate, the process can still introduce near-term swings.Meanwhile, of the Tencent block, BILI is taking $200mn, but another ~$388–400mn (~5%) still needs to be placed by the banks.

(2) Longer term: low-cost funding and removal of non-fundamental pressure

BILI is raising $700mn, and the $300mn of buybacks do not consume the existing authorization. Overall, net proceeds are $400mn.

The CB is zero-coupon with a high conversion premium (US ADS reference around $20, HK at HK$155), more than 40% above yesterday’s close. Near to mid term, equity dilution from the CB alone should be limited.

All in, funding cost is low, and the USD proceeds can be deployed into AI investment or further buybacks.

In the Q2 update, we highlighted the valuation overhang from Tencent’s potential selldown. We had expected Tencent to wait for some recovery before trimming.

In practice, Tencent moved fast (to raise USD for compute and buybacks), choosing to sell at a low price and accept a 5% discount rather than wait for a valuation rebound.

To preserve the strategic relationship (BILI is an effective channel for Tencent’s games and AI applications), Tencent still subscribed to $200mn of CBs. For Tencent, that effectively means receiving ~$400mn of cash.

For BILI, the selldown overhang is removed. Versus Tencent drip-selling or dumping in the open market, this structure is more conducive to share price stability.

3) BILI valuation: back to fundamentals; inflection tied to the new game cycle starting in Q4

Friday’s close implies a $6.37bn mkt cap. On adj. net profit of 3bn/4bn in 2026/27, that is ~14x/11x PE; with a mid-cycle multiple and matched growth, there is room for multiple repair, implying $8bn+ on 18x this year or 15x next year.

Near-term catalysts hinge on Q4 timing for 'Three Kingdoms: Kingship Under Heaven'. 'Shining Lumi' goes live in mid-Sep, but as a lighter casual title its contribution to the new cycle may be limited.