SpaceX's First Post-IPO Earnings Heighten Caution Over AI Investments
SpaceX has reported its first quarterly earnings since going public, beating market expectations on revenue, but concerns have emerged that its massive investments in AI are weighing on profits. On Wall Street, cautious views are spreading even toward investment plans premised on early returns (CNBC), and the earnings report has once again put "AI spending sprees" on the agenda (Bloomberg Law/Tax). How the company demonstrates earnings visibility to investors in the wake of its IPO is likely to shape how it is judged going forward.
Monolith, Backer of MoonShot, Aims to Raise $500M for China AI
Monolith Management, a venture capital firm known for its investment in Chinese AI company MoonShot (月之暗面), is seeking to raise $500 million for a new China AI-focused fund, according to people familiar with the matter. The firm, which has staff in Hong Kong and mainland China, is said to be aiming to close the fund within the year, with the final amount expected to be adjusted within a range (Bloomberg). It suggests that funding to China's AI sector continues to flow unabated.
Infineon Forecasts Higher Revenue on Rising Demand for AI Power Chips
Germany's Infineon expects revenue of around €4.7 billion (approximately $5.4 billion) for the fourth quarter of its fiscal year ending in September, driven by growing demand for data center chips and the effects of price increases. That beats market expectations of around €4.6 billion (Bloomberg). The buildout of AI infrastructure is boosting demand for power management chips, with the ripple effects extending beyond cutting-edge compute chips to the surrounding semiconductors.
Official Rust Repository Introduces LLM Usage Policy
The Rust project has formulated a new policy on LLM usage for its official repository, rust-lang/rust (Rust Blog). As the use of generative AI for code writing and review expands, this marks a concrete step in the open-source community's move to codify how quality, copyright, and attribution should be handled. It is drawing attention in technical communities and could spread to other major projects.
80% of "Identifiable AI Spending" in the U.S. House Goes to ChatGPT
On the AI use front in the U.S. Congress, a report said that roughly 80% of identifiable AI-related spending in the U.S. House is directed at OpenAI's ChatGPT (CNBC). Other vendors such as Anthropic reportedly hold only a limited share, once again highlighting how usage across government is concentrated in a single service. How to handle procurement diversification and dependency risk is likely to remain a subject of debate.
Intern-S2-Mobius: A New Architecture Derived from Qwen3.5-35B Emerges
In the local LLM community, InternLM's "Intern-S2-Mobius" has been making waves. It is described as a derivative model based on Qwen3.5-35B with architectural changes aimed at improving throughput and reducing token consumption (Reddit). Technical details remain limited at this point, and performance validation is still ahead, so real-world benchmark reports are eagerly awaited.