META Muse Tops Charts, Lifts AI Agent Trade; Tencent Rallies on AI | Daily News Recap

Sep 22 | Dolphin Focus:

🐬 Stocks

1.$Meta Platforms(META.US)

Meta jumped over 11% in the prior session, the biggest one-day gain since Apr 2025. Its personal AI assistant Muse topped the US iPhone free chart and the Google Play free chart, reaching 902k+ downloads in six days. The breakout of consumer AI agent products boosted expectations for server CPU demand. With a true hit on the consumer side, AI apps are shifting from model races to user adoption, and we will track user retention, monetization, and actual compute hardware consumption next.

2.$TENCENT(00700.HK)

Tencent rallied as the market repriced the potential value of its consumer-grade agents on the Muse catalyst. The company also released its Hy Image 3.5 Preview image-generation model, upgrading both text-to-image and image-to-image, already integrated into its own product ecosystem. Social use cases plus multimodal models expand Tencent’s AI potential on the consumer side. However, no breakout product has emerged yet, and earnings contribution needs time to observe.

3.$NVIDIA(NVDA.US)

Per an SEC filing, Nvidia will invest an additional $1.5bn in SB Energy ahead of its IPO, bringing total investment to $3bn. SB Energy builds large-scale data centers and power infrastructure, with core projects serving OpenAI’s compute needs. Near term, this is more strategic positioning, as storage contributes limited profit in the short run. The key watch is the long-term energy stack supporting compute infrastructure.

4.$SpaceX(SPCX.US)

SpaceX AI launched Grok 4.7, claiming 2x inference speed vs. peers at half the price. It strengthens long-context handling and complex professional tasks, and the API is open to developers. Using speed and low price as weapons further intensifies price competition in foundation models. Note the claims are benchmark-based, while real task costs can rise with output length and retry counts.

5.$Intel(INTC.US)

Intel is in talks with AUO to co-develop advanced packaging for Micro LED substrates, targeting CPO co-packaged optics and high-density compute integration. This remains an industry rumor with no formal agreement announced, but shares rose premarket on the headlines. If executed, it would signal faster progress in advanced packaging and optoelectronic heterogeneous integration, bridging display panels and semiconductor tech. At this stage there are no shipped products or orders, so it is a longer-term industry narrative.

6.$Alibaba(BABA.US)

At the Cloud Summit, T-Head unveiled its in-house AI chip Zhenwu V900. The company also disclosed plans to develop 5–10 trillion parameter models and target Alibaba Cloud global DC capacity of 20GW+ by 2032. The chip is slated for mass production in Q1 2027. The DC scale and ultra-large models are mid-to-long-term goals rather than already-deployed realities.

7.$Alphabet(GOOGL.US)

Google unveiled a new laptop category, Googlebook, starting at $899 for entry models and produced with Acer, Dell, and Lenovo. It deeply integrates the full Gemini AI stack, bundles Google AI Pro membership, and promises up to 10 years of OS updates, with overseas sales starting in Oct. Google is formally entering the AI PC track, competing with MacBook and Windows AI PCs by embedding foundation model capabilities in hardware. As it relies on OEMs rather than in-house manufacturing, share gains will hinge on end-user experience, channels, and ecosystem fit.

8.$XIAOMI-W(01810.HK)

Xiaomi open-sourced its MiMo‑V2.6 all-modal model series. The Pro version scored near the top among open-source peers on third-party benchmarks, prioritizing RL while balancing inference efficiency and cost, and Xiaomi opened model weights and the training stack to developers. This meaningfully strengthens China’s open-source model cohort, mainly to attract developers and improve edge and cloud ecosystems. Under open-source, direct monetization is difficult, so the value lies in serving Xiaomi’s full-stack AI strategy, with limited near-term profit impact.