Bitmine added 12,532 ETH to its holdings last week, bringing its total ETH position to 6.03 million, approaching its 5% supply target.

Beating AI News Flash: Broadcom CEO Hock Tan told attendees at Singapore’s DBS Asia Future Forward conference that capital is replacing chips, storage, and power supply as the biggest constraint on AI industry growth. He warned that many AI investors do not understand the technology they are betting on yet expect high returns, with their investment behavior resembling casino gambling. Tan cited OpenAI as an example, noting the firm’s annualized revenue hit $50 billion in September, but its steep model training costs still left it operating at a loss, adding that revenue growth does not equal a validated business model. He also projected that wafer shortages could improve between 2028 and 2029, while power supply is expected to gradually ease as new infrastructure comes online. Intel CEO Chen Liwu also noted that rising financing costs and uncertainty around investment returns are emerging as core challenges for the AI sector. While both executives agreed demand for computing power will stay robust over the next two to three years, Tan warned that if the rollout of advanced AI models falls short of expectations or if model capability improvements hit bottlenecks, demand for computing infrastructure could slow sharply.

Well-known crypto analyst Killa published a post stating that in the current cycle, the BTC supply held by short-term holders (STHs) has risen again to overheated levels after breaking away from the macro bottom, a pattern mirroring the previous cycle, which could serve as another signal that the market bottom has already formed. Killa noted that after the STH supply hit a similar level in 2023, BTC continued its sideways rally and remained above its long-term trend, and the market is now exhibiting a comparable trajectory. He reminded investors not to be misled by the Fear & Greed Index, adding that periods of market irrationality can last far longer than anticipated.

Mike Wilson, chief US equity strategist at Morgan Stanley, stated that the yield on the 10-year US Treasury has risen by around 120 basis points since February. Coupled with a stronger US dollar and high oil prices, financial conditions have tightened significantly, with the bond market pre-emptively fulfilling some of the tightening tasks that the Federal Reserve would have otherwise achieved through interest rate hikes. If this process is sufficient to curb economic demand, the Fed may not need to raise rates further in the future. Wilson noted that the US September CPI will be released this Wednesday. If inflation continues its moderate decline, it could ease upward pressure on long-term interest rates and reduce market expectations of additional rate hikes. He advised focusing on bond market volatility and the performance of financial stocks to determine whether financial conditions are stabilizing. Wilson believes the recent resilience of US stocks is mainly driven by large AI-related stocks. The S&P 500 market-cap weighted index has performed significantly better than its equal-weighted counterpart, reflecting that the pressure of high interest rates on most stocks has not yet faded. If bond market volatility declines and bank earnings reports are solid, the market rally could potentially spread from AI leaders to more sectors. He also pointed out that the current rise in US Treasury yields partly reflects capital demand from AI infrastructure investment, rather than purely recession risk.

Microsoft plans to gradually increase the adoption of its in-house developed AI chips in data centers to cut reliance on NVIDIA GPUs. The tech giant said it will primarily use its self-designed Maia chip architecture to handle AI model requests going forward, with plans to deploy multi-gigawatt-scale computing power across the Maia series. On Monday, Microsoft unveiled its third-generation AI accelerator chip, the Maia 300, which is slated for production in 2027. Each data center rack can hold 96 Maia 300 chips—double the capacity of the prior-generation Maia 200. Compared to GPUs, it delivers up to seven times the AI token processing capacity per rack, with a roughly 40% increase in tokens processed per megawatt. Microsoft noted the Maia 300 is expected to boost AI service performance and lower data center operational costs. Evercore ISI analysts project that by Q1 2027, Maia chips will make up around 10% of Microsoft’s AI chip spending. Still, NVIDIA will remain a key supplier for Microsoft’s AI infrastructure, the company clarified, adding that model training will continue to use NVIDIA and AMD chips. Unlike offerings such as Google’s TPU and Amazon’s Trainium, Microsoft’s in-house accelerator chips are not yet available for rental to third-party enterprises.

According to EmberCN’s monitoring, Strategy purchased 250 BTC last week at an average price of $86,194, totaling around $21.55 million. The firm currently holds 848,250 BTC, with a total holding value of approximately $70.532 billion, an average cost price of $75,444, unrealized gains of about $6.536 billion, and a return rate of 10.2%. Bitmine bought 12,532 ETH last week at an average price of roughly $2,601, worth around $32.6 million. It now holds 6,028,946 ETH, valued at approximately $153.19 billion, with an average cost of $3,334, an unrealized loss of about $47.81 billion, and a loss rate of 23.8%. Tom Lee previously stated at Singapore’s TOKEN2049 conference that Bitmine would halt its ETH purchases once its holdings reach 5% of the total ETH supply. Currently, Bitmine is about 77,442 ETH short of this target. Based on its recent buying pace, Bitmine could hit the goal in roughly one month.

Bank of America (BofA) adjusted price targets for three crypto trading platforms ahead of their third-quarter earnings releases. Analyst Craig Siegenthaler raised Robinhood’s price target from $140 to $156, retaining a "Buy" rating; lifted Coinbase’s target from $174 to $203, also keeping a "Buy" rating; and cut eToro’s target from $46 to $39, maintaining a "Neutral" rating. In terms of earnings forecasts, the market expects Robinhood to generate roughly $1.4 billion in Q3 revenue with earnings per share (EPS) of $0.64; the firm has beaten EPS estimates for 11 consecutive quarters. Coinbase is projected to post around $1.77 billion in revenue with a per-share loss of $0.20, after missing profit estimates for three straight quarters. eToro is forecast to have revenue of approximately $203 million and EPS of $0.40. The report notes that BofA’s upward adjustment to Coinbase’s price target was partly driven by expectations of stablecoin-related revenue growth, though the bank also cut its near-term trading volume forecast for the firm. For eToro, BofA lowered its profit forecast due to factors including expected declines in contributions from stock, commodity, and forex trading, as well as slowing account growth.