Afternoon trading saw a collective downturn across Asia-Pacific markets. The Korea Composite Stock Price Index briefly turned negative in the afternoon session after surging more than 2% earlier in the day, eventually closing with a modest gain of 0.15%.

In the A-share market, all three major indices retreated in the final stretch of trading, with the ChiNext Index having previously climbed over 2%. By the closing bell, the Shanghai Composite Index edged up 0.06%, the Shenzhen Component Index dipped 0.05%, and the ChiNext Index inched forward 0.01%, while more than 3,000 stocks across the market finished in negative territory. Combined turnover on the Shanghai and Shenzhen exchanges reached 2.14 trillion yuan, expanding by 104 billion yuan compared with the prior trading day.

Among sector movers, AI application plays saw Nanwei Software secure a second consecutive limit-up, while Zhidu Co, Xinju Network, and Zhongxin Saike all hit their daily upside caps. Computing chip concepts demonstrated notable strength, with Xuanji Information locking in a limit-up alongside Hesheng New Materials and Zongyi Co. The pharmaceutical sector remained active, as Nanhua Biological posted a fourth straight limit-up, while Harbin Pharmaceutical, Xinhua Pharmaceutical, and Huaren Pharmaceutical each delivered back-to-back limit-ups.

Overseas institutions currently hold diverging views on the so-called AI investment bubble theory. Blackstone President Jon Gray pushed back against the bearish narrative in a recent closed-door address, arguing that large language models and AI enterprises are witnessing exponential revenue growth alongside a measurable leap in macroeconomic productivity. He characterized the current AI wave as akin to the dawn of the Industrial Revolution in 1870, with demand far outstripping supply, and identified the true investment theme as resolving real-world hard shortages: pouring substantial capital into chips, data centers, and power infrastructure that face formidable funding and capacity barriers.

Addressing concerns that capital expenditure amounts to little more than circular financing, Gray cited a series of striking figures to demonstrate that returns are genuine. First, large model revenues have exploded: companies such as Anthropic and OpenAI have grown revenue a hundredfold in a short span, with combined annualized revenue reaching an extraordinary $105 billion as of July 2025. Second, enterprise monetization is accelerating: 14 AI-related firms in Blackstone's portfolio have seen annualized revenue surge 21 times from $25 million in September of last year to $525 million. Third, macroeconomic productivity has jumped: average annual US productivity growth of 1.5% over the past decade has climbed to 2.6% over the last two and a half years, while EBITDA margins for S&P 500 constituents and Blackstone portfolio companies have expanded by 500 to 700 basis points over four years.

By contrast, independent research firm Redburn Atlantic has issued sell ratings on CoreWeave and Nebius, two core players in the AI computing infrastructure space. The firm's central thesis holds that while equity markets continue to price in AI growth optimistically, credit markets have already begun pricing risks that equity investors are largely overlooking. The core question revolves around earnings quality: after GPU rental income covers equipment depreciation, operating expenses, and financing costs, how much return ultimately accrues to shareholders. Notably, CoreWeave generated second-quarter revenue of $2.575 billion while incurring net interest expenses of $640 million over the same period, making financing costs a pivotal variable for profitability. The research firm contends that the new cloud competitive landscape in the AI inference era will increasingly reward order quality, prepayment arrangements, equipment utilization rates, and capital returns after accounting for hardware refresh cycles and financing expenses.

Meanwhile, prominent short seller Michael Burry estimates that leading technology giants collectively hold off-balance-sheet commitments exceeding $3 trillion. Burry calculates that Amazon, Meta, Alphabet, Microsoft, and Oracle together hold nearly $1.2 trillion in lease commitments that have not yet commenced, more than $1.5 trillion in purchase commitments, plus additional exposure through special purpose entities and guarantees, bringing the total above $3 trillion. His core concern centers on the structural mismatch between locked-in long-term lease and procurement expenditures on one hand, and the rapid pace of change in equipment economic value, technology requirements, and future demand on the other, leaving companies vulnerable to capital recovery challenges and hardware obsolescence.