Two whales have collectively shorted 50,838 ETH, valued at approximately $98 million.

According to an official announcement from Bitget, the platform has partnered with structured product liquidity provider VETA Finance to upgrade its stock dual-currency investment products. The upgrade expands the product pool beyond existing popular underlying assets including NVIDIA (rNVDA), Google (rGOOGL), and AMD (rAMD), while adding more investment term options. Leveraging deeper liquidity support, the upgraded products offer more competitive annualized returns, and optimize product architecture and settlement mechanisms to deliver a better trading experience and more diverse revenue opportunities for users. Currently, Bitget’s stock dual-currency investment supports two strategies: "buy low" and "sell high". Users can subscribe using USDT or eligible US stock tokens, and based on preset target prices and investment terms, complete corresponding asset conversion and earn interest income per the maturity settlement price. In addition, Bitget plans to launch structured products such as accumulation and accumulation options (AQ) based on the rToken ecosystem with VETA Finance, expanding its product portfolio from standardized wealth management to customized solutions, and further enriching multi-asset wealth management options under the Universal Exchange (UEX) ecosystem.

Beating AI Express (Dongcha News) – Menlo Ventures investor Deedy Das has found that some AI startups simply rent computing resources, deploy open-source models including DeepSeek, Kimi, and GLM, and charge fees per token, quickly scaling monthly revenue to $1 million. A large number of such inference service providers have recently emerged, with no need to purchase GPUs or develop models in-house. There remains a sizable price gap in this business model. Per Deedy’s estimates, even if a B300 GPU has an hourly rental cost of up to $8, as long as its processing speed is sufficient and customer orders are robust, it can generate $10–$20 in token revenue per hour. Barriers to entry are also dropping rapidly: some small teams can deploy open-source models on rented GPUs using only AI programming tools. Platforms like OpenRouter and Vercel further help these service providers aggregate customer demand. Deedy notes that nearly dozens of similar companies enter the market every week. The scramble for GPU rentals is driving up computing power prices, while firms keep cutting fees to compete for clients.

According to monitoring by Lookonchain, crypto influencer Kimchi has re-added pump.fun to his profile three days after initially removing it. The price of Kimchi’s token surged by over 150% at one point. Coincidentally, just hours before Kimchi re-added pump.fun to his profile, trader ja2trader spent $28,900 to acquire 4.61 million $Kimchi; the trader has since booked a profit of $11,000.

3 days after removing @pumpfun from his bio, @kimchi1x has added it back. $Kimchi briefly surged over 150%. Coincidentally, trader ja2trader spent $28.9K buying 4.61M $Kimchi just hours before @kimchi1x added @pumpfun back to his bio, and is now up $11K.

According to BIT (Bit.com) market data, SK Hynix rose more than 1% in pre-market U.S. trading. The company plans to distribute a $12.1 billion special dividend, using proceeds from the sale of its Kioxia stake to return value to shareholders.

According to monitoring by @ai_9684xtpa, an ancient Ethereum ICO-era whale address 0x6C6…d2941 has cumulatively deposited 1,500 ETH to trading platforms recently, worth approximately $3.46 million. The address initially purchased 100,000 ETH at a cost of $0.31, and in February 2016 it is believed to have sold 50,000 ETH at an average price of $5.47, valued at around $273,000, with a return of approximately 1664% at that time. If the ETH deposited this time is sold, the return will reach roughly 744,017%.