Sources: Iran hopes to convey its negotiation terms to the US side via Pakistan.

Citrini analyst Jukan published an analysis stating that OpenAI’s "Jalape?o" AI chip may not scale to the level of NVIDIA’s Rubin platform. Jukan argued that if OpenAI’s HBM4 high-bandwidth memory relies entirely on Samsung for supply, its capacity scaling will hit a certain ceiling. To further expand the deployment of "Jalape?o", OpenAI may need to lower requirements for HBM4 specifications such as transfer speed, so that it can adopt HBM4 products from other suppliers simultaneously, thereby easing supply chain constraints. Earlier today, BlockBeats reported that OpenAI unveiled the latest test results of its self-developed AI inference chip Jalape?o (literally: jalape?o pepper), which beat NVIDIA’s GB200 and GB300 superchips in the InferenceX benchmark. On models including GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T, the chip delivers 1.5 to 1.9 times the AI workload per watt of competing products, with end-to-end latency reduced by 1.7 to 3.6 times.

Crypto analyst CW stated in a post that Bitcoin will need to meet three conditions to kick off a full-scale rally: Hyperliquid whales turning bullish, Bitfinex whales completing their BTC long position accumulation, and the elimination of negative Korea Kimchi Premium and Coinbase Premium. So far, Bitfinex whales have finished building their BTC long positions, while both the negative Kimchi Premium and negative Coinbase Premium have disappeared—meaning two of the three conditions are already satisfied. The market’s core focus moving forward will be whether Hyperliquid whales shift to a bullish stance; once this signal emerges, it could act as a catalyst for Bitcoin to start a new upward trend. It’s worth noting that the above assessment primarily reflects the analyst’s personal views, and related premiums and whale position changes are generally only used as reference indicators for market sentiment and capital flows.

Alibaba recently raised roughly $10.2 billion via a discounted placement of about 710 million new shares, earmarking the funds to step up investment in AI chips, computing power infrastructure, and large language model development. The move diluted existing shareholders’ stakes by approximately 3.6% and put pressure on the company’s stock price. However, amid market skepticism over the share dilution and AI-related capital expenditure burdens, Alibaba insiders have opted to boost their holdings against the trend. Reports note that after Chairman Joseph Tsai and CEO Daniel Zhang bought company shares, founder Jack Ma spent more than HK$600 million (around $77 million) to increase his Alibaba stake, signaling the management’s confidence in its AI strategy. Alibaba previously committed to investing 380 billion yuan over the next three years to build AI and cloud computing infrastructure. While large-scale capital expenditure is weighing on profits and free cash flow in the short term, demand for its AI cloud and computing services remains robust. The core focus for the market going forward is whether the massive AI investment will translate into sustained revenue growth, profit improvements, and free cash flow returns.

According to Coinglass data, the global crypto market recorded $566 million in liquidations over the past 24 hours, including $235 million in long-position liquidations and $331 million in short-position liquidations. A total of 83,754 traders were liquidated worldwide, with the largest single liquidation order occurring on Bitget’s BTCUSDT_UMCBL contract, valued at $103 million.

Decentralized AI inference network DGrid has officially launched its native token DGAI. Data from CoinGecko shows DGAI rose nearly 93% on its first day of listing, trading at around $0.73, with a market cap of roughly $110 million and a 24-hour trading volume exceeding $165 million. DGrid processes AI inference requests for users and developers via distributed nodes, and DGAI is used for network payments, node staking, and rewards. According to its litepaper, the network employs a "Proof of Quality" mechanism to evaluate AI inference results: node operators must stake DGAI and provide computing power, while poor performance or violations will result in corresponding penalties. Additionally, DGrid stated that its DClaw Box hardware—used to run the personal AI agent DClaw—sold over 140 units in just a few hours, priced at 1,580 USDT each, generating total sales of more than 221,000 USDT.

Beating AI’s news flash, citing The Wall Street Journal (WSJ), reports that AI company Anthropic will tell investors in its upcoming IPO filings that its total addressable market (TAM) exceeds $30 trillion, higher than SpaceX’s previously disclosed $28.5 trillion. Anthropic’s TAM calculation is primarily based on the full range of work AI models will be capable of completing in the future. The WSJ report notes that Anthropic’s second-quarter revenue has risen to $11.6 billion. For context, FactSet data shows that the combined revenue of 191 tech firms in the S&P 1500 index totaled roughly $2.4 trillion last year. Sources familiar with the matter also disclosed that Anthropic may seek to raise up to $100 billion via its IPO, targeting a valuation of approximately $2 trillion. The plans are still under discussion, so specific figures could change. The company is expected to release its IPO financial documents in the coming weeks, with a potential listing as early as September or early October this year.