Saudi company Humain invests $3 billion in xAI company

According to South Korea’s Chosun Ilbo, South Korea’s DRAM export prices continue to surge, as AI-driven HBM production squeezes supplies of standard memory. Data shows that from the 1st to the 20th of this month, South Korea’s DRAM export unit price reached $92,183 per kilogram, up 401% year-on-year, and roughly 12.5 times higher than the low point in January 2023. Goldman Sachs forecasts that the DRAM supply gap will widen from 5.0% this year to 5.9% next year, noting that HBM for AI servers and high-capacity server DRAM are absorbing limited production capacity, further tightening supplies of standard DRAM, with shortages potentially lasting until 2028. TrendForce projects that by the end of next year, the wafer input for HBM production by Samsung, SK Hynix, and Micron will account for 30% of total DRAM wafer input, though HBM will only make up around 13% of actual DRAM bit supply. Memory manufacturers warn that tight DRAM and NAND supplies driven by AI demand could persist beyond 2027. SK Hynix stated that from the supply side, 2027 could become “the most severe shortage year in the memory industry’s history.”

According to Bitget market data, South Korea’s KOSPI index opened 0.53% higher. Samsung Electronics rose 1.36%, and SK Hynix gained 1.32%. Japan’s Nikkei 225 index opened down 342.29 points, a 0.52% decline.

Sources from Pakistan’s military and Iran’s security services told Russia’s Sputnik News that the United States and Iran have reached a consensus on ceasefire agreement terms, including freedom of navigation in the Strait of Hormuz. The sources stated: “A consensus has been reached on the ceasefire agreement, which includes a clause on freedom of navigation in the Strait of Hormuz.” The two sides are expected to make relevant announcements in the coming days. Per the Islamabad Memorandum, negotiations and a round of technical meetings will be initiated.

According to Nikkei News, Japan is preparing to build a blockchain-based payment infrastructure to enable instant settlement for stock and Japanese government bond transactions. The Financial Services Agency (FSA), Ministry of Finance, Bank of Japan, and relevant financial institutions will launch a research group this summer, with the development plan to be finalized as early as early 2027. The plan will cover blockchain design, division of responsibilities among institutions, and a future roadmap. If formally approved, the system is expected to be built within a few years, go operational as early as the early 2030s, and may further expand to cross-border remittance services. Currently, cash settlement for Japanese stock transactions takes two days after execution, while government bond transactions settle the next day. The real-time settlement system will eliminate the time gap between transactions and fund settlement, allowing investors to reinvest proceeds almost immediately after selling their assets.

According to BIT (bit.com) market data, US stocks closed higher: the Dow Jones Industrial Average rose 0.3%, the S&P 500 gained 0.32%, and the Nasdaq climbed 0.66%. Nvidia snapped its seven-day losing streak, surging over 2%. Sectors including storage, optical communications, and cryptocurrencies rallied. For the storage and semiconductor sectors: Intel (INTC) rose 0.25%; Advanced Micro Devices (AMD) jumped 4.91%; Micron Technology (MU) gained 2.48%; SanDisk (SNDK) fell 0.83%; Seagate Technology (STX) climbed 3.4%; Western Digital (WDC) rose 3.53%; SK Hynix ADR advanced 2.68%. The optical communications sector: Lumentum (LITE) surged over 6%; Applied Optoelectronics (AAOI) jumped more than 5%; Coherent (COHR) gained over 4%. Crypto-related stocks: Circle (CRCL) rose over 4%; Coinbase (COIN) climbed more than 4%; Strategy advanced over 3%.

Beating AI Insight News Brief: OpenAI has released the first batch of test data for its first custom inference chip Jalape?o, claiming its performance surpasses NVIDIA’s GB300. The chip has reached the Pareto optimal frontier across three publicly available external models: GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Compared to current best commercial systems, Jalape?o’s peak AI output per watt is 1.5 to 1.9 times higher, end-to-end latency is 1.7 to 3.6 times lower, and its advantage expands to 2.1 to 4.1 times in high-interaction agent scenarios. Jalape?o has a rated power of 700 watts, with measured continuous power not exceeding 550 watts. OpenAI emphasized that under agent workloads, real costs should be measured by "AI workload completed per unit of power consumption" rather than individual chip performance. Jalape?o took only 9 months from design to tape-out, with AI deeply involved in circuit optimization and verification. Leveraging Codex and GPT-Astra, the team optimized three open-weight models not originally planned for the project to high performance in just two months; AI-generated implementations of some modules were 1.5 to 1.8 times faster than human-written ones. OpenAI plans to deploy Jalape?o to its own computing infrastructure by the end of the year, while continuing to use external accelerators like NVIDIA at scale. This is the first generation in a multi-generation chip roadmap; Gen 2 is already in deep development, and Gen 3 is taking shape.