South Korea's KOSPI index plunged 5% at the opening of trading.

According to The Wall Street Journal, OpenAI disclosed to investors that its second-quarter revenue rose 18% quarter-over-quarter, climbing from $5.7 billion in Q1 to $6.7 billion. While nearly $7 billion in quarterly revenue is an impressive scale for most startups, this growth rate has still left some shareholders disappointed amid OpenAI’s current valuation, financing plans, and IPO expectations. What has drawn even more market attention, however, is the widening losses. The report states that OpenAI’s Q2 operating loss expanded to $12.3 billion from $9.3 billion in Q1, outpacing revenue growth. Since this figure includes stock-based compensation expenses, the company is further from its profitability target, prompting investors to reassess its pre-IPO financial path. By contrast, Anthropic is emerging as a source of pressure for OpenAI. The WSJ reports that Anthropic’s Q2 revenue rose to $11.6 billion, surpassing OpenAI for the first time, and the firm posted a small operating profit. While Anthropic’s profitability metrics still need further verification from its IPO prospectus, the company has already achieved notable growth in segments such as enterprise AI, code generation, and Claude Code, sparking market discussion over whether the commercialization pace between leading AI labs has reversed.

Solana core developer and Anza CEO Brennan Watt announced in a post that Solana is rolling out its first slot time reduction upgrade on the mainnet, which is expected to lower block generation time from the current ~400 milliseconds to 350 milliseconds, with the adjustment set to take effect starting at Epoch 1020. Watt cautioned developers about transition issues linked to the upgrade, noting that some SDK constants—including DEFAULT_MS_PER_SLOT—have not yet been updated to match the new parameters. The official team will release a version containing the latest values after the feature is activated. The upgrade uses a delayed activation mechanism: the feature will enter a pending activation state at Epoch E, activate at Epoch E+1, and become fully effective at Epoch E+2. For applications that rely on SDK constants to align with the mainnet’s actual slot time, Watt advised adding adaptation logic during the transition period, such as setting function switching mechanisms based on Epoch slot boundaries, to prevent service anomalies caused by parameter changes. Long-term, Solana plans to migrate these network parameters on-chain to allow direct queries by clients. For now, the team needs to complete this upgrade, describing the process as similar to Solana’s early "tough but fast iteration" development phase. Watt also revealed that the team expects the vast majority of nodes to meet the "two-slot latency" target in most scenarios; related restrictions will be further relaxed in the upcoming Anza v4.3 update.

The latest Bank of America (BofA) Global Fund Manager Survey reveals that global investors’ risk appetite is surging rapidly. As US stocks edge closer to record highs, fund managers’ allocation to equities has climbed to a five-year peak, while cash holdings have fallen to 3.5% — a sign the market has largely recovered from earlier jitters over growth slowdown and an AI bubble. BofA strategist Michael Hartnett pointed out that a record 56% of surveyed fund managers expect no significant “landing-style” slowdown in the global economy. In other words, the market’s dominant positioning is betting on economic resilience, sustained corporate earnings expansion, and risk assets will continue to get liquidity support. Notably, the survey shows AI capital expenditures have not yet become a top concern for investors. Even as tech giants keep boosting budgets for data centers, GPUs, servers and power infrastructure, talk of “overheated AI spending” is on the rise — but the BofA survey finds fund managers are currently not overly worried about growth, interest rate hikes, AI capital expenditures or US political risks.

Multicoin Capital deposited another 172,710 $HYPE($10.15M) into #CoinbasePrime 7 hours ago.

According to monitoring by EmberCN, Unitree’s pre-market contract price on Hyperliquid is currently quoted at $90 (equivalent to approximately RMB 603), with a corresponding market capitalization of around RMB 276.4 billion. This marks a roughly 6.7x surge from its IPO offering price of RMB 150.8. At this level, the expected profit per IPO lot is approximately RMB 266,000. To date, addresses have placed orders on Hyperliquid at $90, planning to open a Unitree long position worth $5 million.

Bank of America analyst Vivek Arya says the market may currently be overestimating risks facing the artificial intelligence (AI) investment cycle, resulting in an undervaluation of NVIDIA’s stock, with the chipmaker potentially trading at a 34% to 50% discount to its fair value. Arya reaffirmed a "Buy" rating on NVIDIA in his latest research note and maintained a $350 price target, meaning the stock has substantial upside from current levels. Despite recent pressure on the broader AI stock sector, as investors remain concerned about slowing AI infrastructure investment, progress in enterprise AI commercialization, and high valuation pressures, BofA believes the market has overpriced these risks. Arya points out NVIDIA remains at the core of AI computing infrastructure, and its advantages in GPU products, software ecosystem, and partnerships with cloud computing vendors will enable the company to keep benefiting from rising demand for AI computing power. He notes current market worries about the sustainability of the AI cycle may be masking NVIDIA’s long-term growth potential. Bank of America has repeatedly expressed optimism about NVIDIA’s leading position in the AI data center space, arguing that as generative AI applications expand, demand for high-performance computing resources from enterprises and cloud service providers will continue to support the chipmaker’s future performance growth.