Huang Renxun has left Beijing, and NVIDIA has not disclosed whether Huang Renxun met with the DeepSeek founder.

According to official announcements, Binance Wallet has integrated Robinhood Chain. Users can directly access and trade tokens on Robinhood Chain via the Binance Wallet App and browser extension, delivering a more convenient multi-chain operation experience.

DigiTimes reports that fueled by surging AI demand and structural production capacity bottlenecks, the price of next-generation HBM4 may jump from $2 per kilobit to $4–5 or higher in the second half of 2026. This is partly due to the extreme complexity of HBM4 manufacturing: its production cycle lasts four to six months, and initial yields are notably low. Additionally, HBM production consumes approximately three times the wafer capacity of standard DDR5 DRAM, severely restricting the total memory volume manufacturers can output at existing facilities. (Jinshi)

Bitcoin’s BIP-110 proposal is approaching its early August deadline, yet miner support for the measure remains below 1%, signaling significant resistance to the initiative. Officially titled “Temporary Soft Fork for Reducing Data”, BIP-110’s core controversy centers on restricting non-financial data on the Bitcoin blockchain. The proposal aims to cap OP_RETURN data capacity within a year, ban most arbitrary data exceeding 256 bytes from being written to the chain, and limit certain script formats primarily used for data storage. Supporters argue the plan would refocus the Bitcoin network on its payment function and reduce node operational burdens; opponents counter that it would escalate policy disputes over block space usage into consensus rule changes, effectively determining which transactions qualify as “acceptable”. Strategy founder Michael Saylor and Blockstream co-founder Adam Back have both publicly opposed BIP-110. Saylor remarked, “There are 110 things more dangerous than junk data”, adding that the proposal would “turn the junk data debate into a consensus change, invalidating some currently valid transactions that pay fees”. Back stated that if supporters cannot accept the status quo, they may choose to fork, but “Bitcoin will not join”. Data shows BIP-110 uses a user-activated soft fork mechanism with a 55% miner signaling threshold, though miner signaling rates have never topped roughly 1% to date, with the current cycle sitting at 0 and no major mining pools backing the measure. The share of nodes running BIP-110 software also remains in the single digits, primarily from Bitcoin Knots users. The proposal’s current signaling cycle will end around block height 959,615, with a voluntary lock-in period expected in early August and activation targeted for around September. If broad support is still absent by then, the initiative could result in a minority of nodes forming a separate chain.

Crypto analyst Ali Charts noted in a post that Bitcoin is currently at a critical juncture for its next major move, with the $64,700 level serving as the key threshold to watch. If Bitcoin fails to break above the channel’s top at $64,700, it could trigger a deeper pullback. Should sellers hold that level, Bitcoin may dip to the channel’s mid-support at $63,000 and could further test the channel’s bottom around $61,500. However, Ali Charts added that a Bitcoin hourly close above $64,700 would invalidate the bearish outlook. After breaking through this resistance, Bitcoin could next target $66,400, with its second target set at $68,000.

The Kobeissi Letter stated that the U.S. economy is growing increasingly dependent on AI investment. AI investment now accounts for over 25% of U.S. GDP growth contribution, marking an all-time high. Related investments cover software, IT equipment, R&D, data center construction, and other areas. In other words, for every $4 of U.S. economic growth, more than $1 is driven by AI investment. At the same time, AI-related spending has risen to approximately 8% of U.S. GDP, also a historical record. By contrast, during the 2000 dot-com bubble, spending on IT equipment, software, and R&D peaked at around 6.5% of GDP. Today, U.S. economic growth is increasingly centered on AI.

Samsung Electronics plans to bring forward the mass production timeline of the first wafer fab at its Yongin semiconductor cluster in South Korea by two years, targeting to start operations in October 2029. Separately, SK Hynix is considering building new production facilities in the U.S., as AI demand drives South Korean semiconductor firms to accelerate expansion plans. Samsung originally planned to construct a total of six wafer fabs at the Yongin National Industrial Park, with the first unit scheduled to launch operations in 2031. Now, the South Korean government and Samsung have reached a consensus to advance the timeline to October 2029. The government will speed up the construction of key infrastructure including land development, power supply, and water resources to support the project’s early progress. For SK Hynix, following its successful listing on the NASDAQ on July 10, the company is weighing plans to build a new production base in the U.S. According to reports, SK Hynix has been studying the establishment of a semiconductor production hub in the U.S.—a region with a concentrated AI industry ecosystem—since the start of this year. Global storage chip leaders including Samsung Electronics and SK Hynix are ramping up production capacity to ease the chip supply crunch spurred by the AI sector. The South Korean government is also pushing for the development of new semiconductor industrial parks, and plans to accelerate project delivery by simplifying environmental assessments and optimizing approval processes. Industry insiders noted that while market concerns linger over a peak in the semiconductor cycle, current demand for chip production capacity remains robust. To compete for industrial dominance in the AI era, expanding semiconductor manufacturing capacity has become a key priority.