BREAKING: $435 million has been liquidated from the cryptocurrency market in the past 12 hours.

According to South Korean media reports, Samsung Electronics and SK Hynix have recently rebounded, with their respective single-stock leveraged ETFs also surging sharply. However, investors who purchased near the June 25 high still face significant losses. As of August 14, Samsung Electronics was down 23.43% from its June 25 closing price, while SK Hynix fell 43.61% over the same period. Data shows that the seven Samsung Electronics single-stock leveraged ETFs have posted an average loss of 52.25% since June 25, and the seven SK Hynix leveraged ETFs have averaged a 76.51% loss. If investors had invested 1 million won each on that date, their holdings are now worth roughly 478,000 won and 235,000 won respectively. Based on current net asset values, the two types of ETFs would need to rise by approximately 109.4% and 325.7% respectively to break even. Since these leveraged ETFs track twice the daily return of their underlying assets, the required underlying asset gain to recover principal cannot be simply calculated as twice the ETF’s required increase, and repeated price fluctuations will also lead to volatility drag. Assuming the underlying assets rise by the same percentage every trading day for the next 20 sessions with no declines, Samsung Electronics would need to climb around 45.2% from current levels to 398,600 won, and SK Hynix would need to rise roughly 109.1% to 3,439,000 won, to erase the average losses of the related leveraged ETFs. Currently, the 12-month forward price-to-earnings (P/E) ratios of Samsung Electronics and SK Hynix have fallen to 4.5x and 3.7x respectively. Kim Dong-won, head of research at KB Securities, stated that storage demand is expected to strengthen further in the coming years, and he projects both companies could see a revaluation of their stock prices starting from the third quarter.

The Kobeissi Letter reported that retail investors collectively purchased approximately $27 billion worth of NVIDIA (NVDA) stock over the past year, ranking first among the "Magnificent Seven" U.S. tech giants. Since October 2025, retail buying of NVIDIA has surged more than fourfold. During the same period, Tesla (TSLA) ranked second with around $15 billion in retail purchases, while Microsoft (MSFT) saw roughly $9 billion in retail buying. By contrast, Apple (AAPL) recorded approximately $5 billion in retail selling over the past year, making it the only company among the "Magnificent Seven" to see net retail selling. For retail investors, NVIDIA remains the primary AI trading asset.

Newly emerging US stock star analyst KawzInvests (@KawzInvests) published an article stating that Alibaba (BABA) remains an undervalued asset in the market. He noted that Alibaba Cloud’s external revenue rose 40% year-over-year last quarter, while AI product revenue has posted double-digit growth for 11 consecutive quarters. Additionally, Alibaba holds approximately 5% of ChangXin Memory Technologies (CXMT)’s shares; its initial investment in this stake was around RMB 7.6 billion, and the holding is now valued at over $200 billion.

Nvidia has disclosed its latest 13F report as of June 30, showing holdings of 8 U.S. stocks, with a total quarter-end portfolio value of approximately $63.44 billion, a roughly 245% increase from $18.37 billion at the end of the first quarter. Specifically, Intel remains Nvidia’s largest holding, with 215 million shares valued at around $29.99 billion, accounting for about 47.3% of the 13F portfolio. SpaceX appears for the first time in Nvidia’s 13F report, holding 123 million shares worth approximately $20.98 billion, making it the second-largest position with a 33.1% share. The two combined make up roughly 80.3% of Nvidia’s reported portfolio. Other holdings include 47.21 million CoreWeave shares worth $4.7 billion, 7.79 million Coherent shares at $3.07 billion, 166 million Nokia shares valued at $2.21 billion, 4.82 million Synopsys shares worth $2.15 billion, 1.19 million Nebius shares at $330 million, and 830,000 Generate Biomedicines shares valued at $14.06 million. In terms of position adjustments, aside from SpaceX’s debut in the 13F, the share counts of the remaining seven stocks remained unchanged from the end of the first quarter. Thus, the sharp growth in Nvidia’s 13F portfolio value stems mainly from the first-time inclusion of the SpaceX position and the rise in market value of existing holdings, and does not indicate that Nvidia made a net purchase of approximately $45 billion in stocks on the secondary market during the second quarter. Notably, Nvidia’s SpaceX stake originated from its earlier investment in xAI, rather than direct secondary market purchases after SpaceX’s IPO. A 13F filing only reflects reportable securities holdings as of quarter-end, and does not disclose specific acquisition timelines, investment costs, or transactions executed after the quarter end.

ElizaOS founder Shaw published an article accusing daos.fun founder baoskee of using insider information to trade during the renaming and migration of the ai16z project, claiming he profited roughly $6.6 million from selling ai16z tokens. Shaw noted that in June 2025, the project team had promised to launch a Snapshot vote to let the community decide on a name change, as venture capital firm a16z had earlier demanded the project rebrand to avoid trademark disputes. Shaw alleged that when the voting mechanism failed to launch as promised, baoskee—aware of the renaming pressure and subsequent migration plans—sold all ai16z holdings in daos.fun’s execution wallet and continued offloading tokens ahead of the project’s move to ElizaOS, driving the token’s price down. Shaw argued baoskee held information unavailable to other token holders. In response, baoskee published a post denying the claims, stating the Snapshot vote had already been launched, and that daos.fun had additionally added and locked over $1 million worth of ai16z liquidity at its own cost. He also leveled accusations against Shaw, citing serious issues in project operations, token migration, and the use of development funds. The two sides’ accounts remain sharply conflicting, with Shaw saying he will attach relevant on-chain records from Solscan at the end of his article.

Daniel Roberts, co-founder and co-CEO of IREN, wrote in a post that each new AI data center will further boost, rather than reduce, demand for AI. He argues that comparing the current AI infrastructure cycle to past tech booms overlooks two key factors: stronger AI itself requires more computing power, and cheaper AI will drive higher usage. Roberts notes that every leap in AI capability creates new use cases that did not exist a year ago, meaning demand can grow at the pace of software development, while supply cannot match that rate. Expansion of computing power supply depends on access to electricity, transmission lines, concrete, steel, and large numbers of construction workers, with its growth constrained by real-world infrastructure limits. He added that in past tech boom cycles, technological progress often helped supply catch up with demand, but this time, technology itself is driving exponential growth in AI demand, and real-world construction speeds cannot keep pace fast enough.