The core logic driving tech stocks is showing an anomaly: the stronger their fundamentals, the heavier the selling pressure, as the market reassesses AI capital expenditures.

U.S. Securities and Exchange Commission (SEC) has formally proposed the new "Regulation E-Delivery", significantly expanding the authority of entities including issuers, broker-dealers, and investment advisers to deliver information electronically. The rule aims to meet disclosure and delivery requirements under federal securities laws, and promote electronic delivery as a more mainstream, flexible option to replace traditional paper-based delivery. SEC Chair Paul Atkins noted that this is another step toward building a modern-era regulatory framework and a key pillar of his agenda since taking office. "In the age of artificial intelligence and blockchain technology, default paper delivery should be a historical relic, not the standard." Earlier, Atkins launched "Project Crypto" to modernize on-chain markets, and the electronic delivery rule aligns with his broader regulatory modernization agenda. The proposal will subsequently enter a public comment period.

According to EmberCN’s monitoring, the address that drained the BONK treasury via a governance attack transferred 1.186 trillion BONK tokens (valued at approximately $4.11 million) to Binance three hours ago. This address spent roughly $4.4 million to acquire enough BONK tokens 10 days ago to meet the governance vote approval threshold, then submitted a governance proposal that was forcibly passed, siphoning 4.426 trillion BONK from the BONK treasury, worth around $21.2 million. On the day of the incident, it had already transferred 40 billion BONK (about $190,000) to OKX, with the remaining tokens held on-chain. As of press time, approximately 3.2 trillion BONK (valued at roughly $10.98 million) still remains from the amount drained from the treasury. The incident has caused BONK’s price to drop by a cumulative 28% since the event.

According to Bitget market data, Japanese storage stock Kioxia hit its daily limit during intraday trading, currently down 15.55%, with its market capitalization halved from the June peak. The Nikkei 225 index extended its intraday decline to over 4%. Per BIT (bit.com) market data, US storage-related stocks SanDisk and Western Digital extended losses by over 4% in after-hours trading, while Micron Technology fell nearly 4%.

The Japanese domestic multimodal foundation model development project "Noetra" officially launched on July 16. Core enterprises include Sony Group, SoftBank, NEC, and Honda, with a total of 44 companies and groups contributing to the initiative, covering a wide range of industries such as manufacturing, finance, logistics, and communications. Engineers from institutions including the National Institute of Advanced Industrial Science and Technology (AIST) and Preferred Networks will also join the R&D. This project is part of the "Multimodal Foundation Model Development Project for AI Robots and Physical AI" promoted by Japan’s Ministry of Economy, Trade and Industry (METI), officially named the FRONTia Project by NVIDIA. Its goal is to build a Japanese domestic foundation model for physical AI scenarios like manufacturing sites and robots, rather than just a Japanese-language conversational AI. At the hardware level, Noetra will collaborate with NVIDIA to build a computing platform equipped with approximately 27,500 latest Rubin GPUs and 13,750 Vera CPUs, adopting NVIDIA Vera Rubin NVL72 racks and DSX platform architecture, with a designed power capacity of 140 megawatts. Construction is scheduled to start in April 2027 and operations to launch in June 2028, when it will become Japan’s largest AI computing infrastructure. The R&D roadmap is divided into three phases: starting from fiscal 2026, developing an inference foundation model centered on AI agents and natural language processing; achieving a full multimodal foundation model that seamlessly integrates text, images, video, and audio in fiscal 2028; and realizing real-world native AI that understands spatial and physical attributes in fiscal 2030, with final applications spanning manufacturing, logistics, healthcare, and communications.

Prediction market platform Predict.fun has announced that all its Up/Down markets are now live with a Maker order rebate mechanism. Users who complete trades via limit orders will receive a 25% rebate on Maker fees. Compared to Polymarket, Predict.fun’s latest Maker rebate offers two core, more straightforward advantages: a higher rebate rate, and real-time rebate disbursement immediately after trade execution, with no need to wait for market settlement. For high-frequency traders, professional market makers, and users who regularly use limit orders, real-time rebates enable faster capital release, reducing the ongoing erosion of trading profits by fees; the higher rebate rate also means that as trading frequency and volume grow, the actual cost savings will become more pronounced. This mechanism currently covers all of Predict.fun’s Up/Down markets. Users do not need to register or submit additional applications—rebates will be automatically credited to their accounts once eligible limit orders are filled.

According to market data from BIT (bit.com), Apple closed 1.76% higher in U.S. stock trading, hitting a record high of $333.26 per share, and rose an additional 0.47% in after-hours trading. On the news front, on July 15, Apple Intelligence completed its first domestic generative AI filing in China. Alibaba’s Qianwen AI will be integrated into Apple Intelligence as its AI capability, providing Chinese users of iOS, iPadOS, macOS, and visionOS with services including text and image understanding, content generation, and more—allowing users to experience these features directly without switching between apps. In addition to Alibaba’s Qianwen AI, Baidu’s AI capabilities will also be integrated into Apple Intelligence, marking the official implementation of Apple’s localized AI strategy in the Chinese market.