Bitcoin ETF Fund Flow Turns Positive, IBIT Sees Net Inflows of Over $3 Billion Year-to-Date

According to official announcements, Binance Wallet has launched a Pre-Access campaign powered by PancakeSwap, allowing users to gain indirect exposure to high-profile private companies via tokenization ahead of their potential IPOs. The service is accessible via Binance Wallet and operates on a fully on-chain, self-custody model, with its first project set to be announced soon. Users can increase their Pre-Access quotas by earning Alpha points, as well as through on-chain bStocks trading volume and holdings. All related assets are transacted on-chain in users’ own wallets, with full self-custody retained by the users. Binance Wallet noted that Pre-Access is just one part of its on-chain TradFi (Traditional Finance) initiative, adding that it will integrate products like Pre-Access and RWA (Real-World Assets) in the future to offer users a unified on-chain investment gateway. The platform stated that early investment opportunities, which were previously restricted by geography, wealth levels, and institutional status, are now gradually being opened up to retail users.

According to Farside's monitoring data, U.S. spot Bitcoin ETFs recorded a net inflow of $6.1 million this week, while U.S. spot Ethereum ETFs saw a net outflow of $140.6 million.

Beating AI News – A batch of internal documents has been unsealed in the copyright lawsuit between The New York Times and OpenAI, Microsoft. Microsoft Director of Applied Science Brent Hecht once described large-scale scraping of online content to train AI as "unprecedented, astonishing theft"—even potentially "the largest labor theft in human history." Former ChatGPT lead Nick Turley also stated that AI products pose an "existential threat" to publishers, essentially acting as content substitutes, with stronger models leading to higher substitution rates. Microsoft CEO Satya Nadella admitted in testimony that chatbots provide users with direct answers, reducing traffic to original content sources. Another internal Microsoft document dubs this model a "death spiral": AI siphons revenue from content creators, leading to a decline in high-quality content that ultimately harms the AI’s own training. Plaintiffs also cited internal Microsoft data showing that, compared to traditional Bing Search, Copilot drives 83% to 93% less click-through traffic to news sites like The New York Times. OpenAI and Microsoft continue to argue that using copyrighted content to train AI qualifies as fair use. Microsoft clarified that Hecht’s remarks were merely his personal views, not representative of the company’s stance, nor a legal conclusion on copyright issues.

According to official announcements, PancakeSwap has launched a Pre-Access portal, offering users a channel to gain exposure to relevant tokenized assets ahead of potential corporate listings. The initiative operates on a limited-time subscription model: PancakeSwap provides the event entry, a third-party service provider manages the underlying tokenized exposure, and users can access the relevant event page via Binance Wallet. Each event page will disclose key details including company and token information, subscription assets, issuance price, implied valuation, event timeline, distribution arrangements, and risk disclosures. Eligible users, after completing the qualification check and accepting the relevant terms, may subscribe via self-custody wallets, with assets remaining under their own control. Specific allocation rules, subscription limits, and eligibility requirements are subject to the official terms of each individual event. Following the event’s conclusion, participants will proceed to either a redemption or refund process per the specific rules. Eligible users will receive their corresponding tokenized exposure once the redemption window opens; users who fail to meet relevant conditions or are required to refund based on event outcomes will get a refund. PancakeSwap reminds users to carefully review each event’s redemption schedule and terms before participating.

UBS currently forecasts that global artificial intelligence (AI) capital expenditure will reach nearly $1 trillion in 2026, and climb further to around $1.4 trillion in 2027, with a sharp surge in memory costs being the primary driver of this growth. UBS estimates that memory-related spending will jump from $71 billion in 2025 to $367 billion this year, and hit $923 billion in 2027. The remaining AI-related costs are projected to total $631 billion in 2026, before falling to $525 billion in 2027. This means that the rise in memory costs will contribute roughly 60% of the total increase in AI capital expenditure this year; by 2027, as spending on other components declines, the growth in memory costs will even exceed the entire net increase in AI capital expenditure.

On-chain analyst Ai Yi (@ai_9684xtpa) reports that GSTOCK surged more than 170% intraday. A trader added to their position at the peak, spending a total of $261,000. The address first purchased GSTOCK two days ago, then increased their holdings again during a 20-minute sharp rally. Currently, the address holds 12.31 million GSTOCK at an average price of $0.02117 per token, with an unrealized profit of $51,000, ranking as the token’s fourth-largest holding address.