Musk outlines orbital AI compute blueprint; INTC soars on inclusion in leading AI chip foundry shortlist | Daily News Recap

0609 ｜Dolphin Research Focus:

🐬 Macro/Industry

1. The US DoD updated its Section 1260H list of Chinese military-linked firms, spanning internet, semis, NEV, and innovative drugs, with additions including BABA, BIDU, WuXi AppTec, BYD, and Nio, while Huawei, Tencent, and Hikvision were already on it. Listed firms issued statements denying defense operations and said they would appeal. The list does not directly impose financial sanctions, only restricts US Gov. procurement, so near-term operating impact is limited but it dampens market risk appetite.

2. WeChat formally opened an access channel for AI ecosystem developers, with Meituan and Trip.com announcing pilot compatibility, and Didi confirmed as a first-batch partner on Jun 9. Through WeChat AI agents, users can one-click invoke food delivery, ride-hailing, and flight/hotel services without switching apps. The platform offers both zero-code auto-adaptation and deep customization. The first pilot batch also includes JD, Tongcheng, KFC, and Dewu across multiple verticals. For Tencent, this fortifies WeChat’s billion-scale traffic as an AI entry point and sets a commercialization benchmark for agents, while Meituan, Didi, and Trip.com can lower CAC and improve conversion with social-scale traffic. Overall, it opens a new domestic consumer AI interaction model.

🐬 Single Names

1.$SpaceX(SPCX.US)

Ahead of a potential IPO, Elon Musk disclosed detailed plans for an orbital AI data center as a core growth engine, targeting 1 GW of annual space compute deployment by 2027, leveraging Starship for low-cost launches of mass AI satellites. The plan is paired with a Tesla in-house chip fab, and the company has filed with the FCC for authorization to launch millions of AI satellites. Space compute can sidestep grid power constraints and complement xAI’s rented compute as a second growth curve, materially enriching the IPO narrative. However, space infrastructure demands heavy capex and long payback, lifting near-term capex pressure while building a differentiated moat vs. ground-based compute over the mid to long term.

2.$XIAOMI-W(01810.HK)

Xiaomi’s MiMo team launched the MiMo-V2.5-Pro-UltraSpeed mode, achieving the world’s first trillion-parameter model inference speed above 1,000 tokens/s on a generic 8-GPU node, with peak at 1,200 tokens/s. No custom chips are needed, and a limited-time paid API is open. High-speed inference on low-cost, off-the-shelf hardware suits phone, auto, and enterprise AI deployments, strengthening Xiaomi’s all-endpoint AI ecosystem. This could support higher AI multiples for HK consumer electronics.

3.$CHINA LIT(00772.HK)

Yuewen announced it will spend approx. RMB 401 mn to acquire Tencent’s 28.22% stake in Yihua Kaitian; post-deal, Yuewen’s stake rises to 59.7%, becoming the largest shareholder. Yihua Kaitian holds leading domestic anime IP such as ‘Ling Cage’ and has mature industrialized animation capacity. The deal will drive deep synergy between upstream web fiction and midstream animation, filling Yuewen’s capacity gap in long-form animation. Consolidating equity from Tencent to the listed platform clarifies business boundaries, with a small near-term cash outlay and long-term full-chain operations to raise the monetization ceiling of domestic anime IP.

4.$Coinbase(COIN.US)

Coinbase became the official USDC treasury deployer for the Hyperliquid protocol, responsible for reserve wallet setup and asset routing, and increased HYPE staking to align long-term interests. Hyperliquid’s on-chain USDC stock is around $5 bn. This is a landmark tie-up where a compliant top-tier CEX deeply custodies a leading DEX’s treasury assets. Coinbase expands into on-chain treasury custody and strengthens institutional trust, while Hyperliquid brings in compliance to reduce regulatory risk and attract institutional liquidity, advancing institutional stablecoin custody.

5.$Intel(INTC.US)

Reports indicate Google has finalized outsourcing Marvell-designed next-gen TPU custom network chips to Intel’s 18A process, with mass production (Est.) by end-2027, and Nvidia is also testing the node, spurring a rally in Intel’s stock overnight. With AI demand surging and TSMC’s capacity saturated, Intel Foundry (IFS) is hitting an inflection point by entering top global designers’ supply chains. This helps break TSMC’s near-monopoly, reduces overseas foundry concentration risk for US tech giants, and reshapes the advanced wafer foundry landscape.

🐬 Top-Performing Sectors

A-share: Data processing & outsourcing, semi materials & equipment, electronic components.

HK: Electronic components, security & alarm services, silver.

US: Kitchenware, semi materials & equipment, electronic components.

🐬 Watch for Tomorrow

1. China May M2 supply YoY. China May CPI YoY.

2. US May CPI MoM (SA) and core CPI MoM (SA). US May CPI YoY (NSA) and core CPI YoY (NSA).

Risk disclosure and statement. Dolphin Research disclaimer and general disclosure