SpaceX's IPO Symbolizes the AI Infrastructure Investment Boom
SpaceX has gone public through an IPO, and investor appetite for AI infrastructure has reached unprecedented levels. According to Bloomberg, Goldman Sachs' Waldron described the IPO as follows: "SpaceX's record-breaking IPO demonstrates that investors are extremely eager to fund AI infrastructure."
Meanwhile, WIRED reports that although an unusually large allocation of shares in this IPO was reserved for retail investors, experts warn that "the odds of getting rich are slim." The key lies in the structure of AI demand behind the IPO. Training and operating AI models requires enormous computing infrastructure, and data center investment is proceeding at a never-before-seen pace. The success of SpaceX's IPO could become a tailwind for such infrastructure investment across the board.
Moonshot AI Releases Kimi-K2.7-Code as Competition Among Coding-Specialized Models Intensifies
China's Moonshot AI has released **Kimi-K2.7-Code**, a coding-specialized model, on Hugging Face. It is already a topic of discussion in the LocalLLaMA community, with its performance on real-world coding tasks drawing attention.
However, multiple users in the same community have also reported that "Kimi K2.6's behavior has recently changed." The suggestion that model behavior is inconsistent across versions should be recognized as an operational challenge.
Competition among coding-specialized models is heating up between major players such as OpenAI Codex, Anthropic Claude, and Google Gemini. A continuously growing set of open-weight options is a definite benefit for the developer community.
Huawei Releases openPangu 2.0, Set to Be Open-Sourced on June 30
Huawei has released the large language model **openPangu 2.0** and announced that it will be open-sourced on June 30.
With China-originated open models such as DeepSeek and Qwen already drawing global attention, Huawei's full-fledged entry further expands the options. Adoption in enterprise use cases is particularly anticipated, and the model may possess performance characteristics distinct from existing Chinese-made models. Once it is open-sourced at the end of June, its benchmark results will be closely watched.
A Zero-Cost Browser-Operation AI Agent Running on WASM
**browser-use-wasm**, which drew simultaneous attention on Hacker News and LocalLLaMA, is a browser-operation AI agent that runs on WebAssembly (WASM). It is available on GitHub as "pdufour/browser-use-wasm".
Conventional browser automation tools required server-side resources, but by running on WASM, everything can be completed on the client side alone. It can have AI click and type on web pages, and its design features an effectively zero cost.
The approach of providing an AI agent execution environment with no infrastructure cost is attractive to individual developers and small teams. Its practical range of uses is expected to be broad, including automated browser testing and data collection.
11 Model Families Ported to Apple's CoreAI Framework
Eleven model families have been ported to **CoreAI**, the on-device AI framework Apple announced at WWDC 2026. They are available on GitHub as "coreai-model-zoo", making it possible to try a variety of models in Apple's on-device environment.
Apple's AI strategy has been consistent in aiming to complete inference on the device rather than on the server. The rapid expansion of its model portfolio suggests that the CoreAI ecosystem is entering a practical stage. It looks likely to become a strong option for app developers who prioritize privacy.
Debate Over Government Stakes in AI Companies Heats Up
According to Semafor, policy debate is intensifying over whether governments should hold stakes in AI companies. As AI becomes directly tied to national security and economic competitiveness, opinions are split over the approach of government investment in private companies.
Proponents argue that "AI has a nature close to public infrastructure," while opponents warn that "government intervention hinders innovation." It is worth noting that the debate over AI regulation is shifting from "how to restrict it" to "how to invest in and engage with it."
---