pow

At the ISC High Performance Computing Conference held in Hamburg, Germany, the latest edition of the global supercomputer TOP500 and energy efficiency Green500 rankings was officially released. The data shows that NVIDIA technology has powered more than 400 of the world's 500 fastest supercomputers, accounting for 81%, and nearly 90% of the newly ranked systems.In terms of hardware distribution, the list includes 376 systems that utilize NVIDIA networking technology, 238 systems accelerated by its GPUs, and another 26 systems that use Grace CPUs. In the Green500 ranking, which measures green energy efficiency, the top 8 positions are all driven by NVIDIA GPUs, with the KAIROS system based on the Grace Hopper superchip taking the top spot. Additionally, supercomputing systems based on the new Blackwell architecture (B200 and GB200) have also made their debut on this ranking.

According to a report by The Information, the confidential IPO registration draft submitted by OpenAI shows that as of the end of March 2026, OpenAI's balance sheet exhibits "light asset" characteristics, with zero debt on the books and capital expenditures of only $46 million in the first quarter. However, in reality, the company has placed substantial infrastructure expenditures off the books, with future procurement commitments in chips, energy, and data centers reaching up to $665 billion. Financial data indicates that OpenAI's actual net loss in the first quarter was approximately $8.5 billion, with revenue costs amounting to $3.5 billion.Additionally, OpenAI demonstrates a very high characteristic of related-party funding cycles. In the first quarter, 72% of its revenue costs and 45% of total expenditures flowed to related parties (expected to be primarily Microsoft), and it directly used $488 million in equity to settle part of its computing power bills. In the data center joint venture project within its consolidated financial statements, nearly $5 billion in book losses is accounted for as belonging to external partners. The documents also reveal that its main competitor, Anthropic, is similarly engaging in large-scale off-balance-sheet expansion, including $4.5 billion in data center service commitments and $35 billion in chip leasing orders.

According to Jinshi reports, SpaceX has signed a computing power agreement with the American AI startup Reflection, with a transaction amount of up to 6.3 billion dollars.

The MiniMax M 3.0 free computing power event launched by the B.AI platform on June 18 is now in the countdown phase and will end in 3 days. It is reported that new users who log in to the platform for the first time through Binance Wallet or Bitget Wallet can receive a reward of 1 million free Credits.In addition, during the event, the platform offers a recharge cashback mechanism. Users can receive a 1:1 equivalent Credits bonus for recharging on the BNB Chain, and a 1:0.5 ratio bonus for recharging on other public chains, with a maximum cumulative bonus limit of 100 USD equivalent per user. According to the information, the MiniMax M 3.0 model previously ranked first in the Artificial Analysis open-source model leaderboard.

ENS DAO has initiated a new Temp Check ("temperature check" or "preliminary review") governance proposal aimed at expanding the authority of the ENS Foundation to cover areas such as operational management, funding projects, and long-term financial management. According to the proposal, ENS token holders will retain ultimate control over the protocol and the power to remove board members.

Green energy infrastructure startup TAR announced the completion of a $27 million seed round financing to develop modular "plug-and-play" power systems for data centers, aimed at addressing the power and deployment bottlenecks faced by data centers in the AI era.According to reports, the solution combines solar energy, wind energy, battery storage, and natural gas backup units to achieve nearly round-the-clock (24/7) local power supply capability, reducing reliance on the public grid and thus bypassing issues such as grid access queuing, approval delays, and power price fluctuations. TAR's co-founder stated that the core idea is to significantly compress the deployment cycle of energy systems through factory prefabrication, pre-assembly, and pre-testing, enabling data centers to achieve "rapid go-live" capability.In pilot projects, the system can provide approximately 10 MW of stable power supply and plans to deploy over 200 MW of normal load capacity by 2027. The company noted that its first customer is an undisclosed "neocloud" service provider, aiming to provide a faster energy deployment path for AI computing infrastructure.In terms of the economic model, TAR stated that its solution does not aim for costs below those of traditional grids but prioritizes solving the "speed issue." Its off-grid energy system can be deployed in about three months, avoiding the time costs associated with grid access and land restrictions. As the demand for AI computing continues to grow, power supply has been identified by multiple studies as one of the main bottlenecks for data center expansion. Industry analysis suggests that the "off-grid energy + modular data center" model is becoming a new direction in the competition for AI infrastructure.

According to a report by The Information, Broadcom announced its support for the Anthropic computing power project participated by Apollo and Blackstone.On the surface, this deal appears to be a financing arrangement for AI computing infrastructure, but industry insiders believe that its core lies in indirectly driving the long-term demand growth for Broadcom's custom AI chips by promoting large-scale AI computing deployments.

The white-haired stock god stated that NVIDIA and Google are advancing the early application of 800V direct current (DC) power architecture, with related small batch shipments expected to start in the third quarter of 2026. It is reported that this technology will first be applied to NVIDIA's Vera Rubin platform and Google's next-generation AI data centers.Due to the acceleration of this process, the power semiconductor and data center infrastructure sectors are expected to receive significant boosts. Currently, the supply chain companies regarded as core beneficiaries include Delta Electronics (2308), Vertiv (VRT), Songchuan Precision (7788), as well as Schneider Electric, Eaton, and Siemens.

According to Cyvers Alerts monitoring, Token of Power was hacked, resulting in a loss of approximately $1.58 million. An address funded through Tornado Cash executed a malicious transaction, transferring funds from the Balancer V1 TOP/WETH liquidity pool.The stolen assets were subsequently transferred back to Tornado Cash for mixing.

According to The Information, informed sources reveal that Goldman Sachs and JPMorgan are exploring trading methods based on the cost of computing power, including futures contracts related to GPU leasing prices.As one of the scarcest resources in the AI boom, related futures are expected to be listed by exchanges later this year. Industry insiders say this initiative demonstrates that hundreds of billions of dollars flowing into data centers and the chip sector are reshaping the financial market landscape. For banks providing financing for AI infrastructure construction, such innovative tools may become new means of managing risk.

According to analyst Darkfost's data, Bitcoin has retraced and fallen below the 4% percentile of the Power Law model, entering an extremely undervalued range, a valuation level that has historically been present only 4% of the time. Darkfost emphasizes that this is a suitable time for long-term position building, not a short-term price prediction.

According to Jinshi reports, company documents show that Google (GOOG.O) has agreed to pay SpaceX $920 million per month for computing power as part of a cloud service agreement that will last until mid-2029.SpaceX stated in documents submitted on Friday that Google will pay SpaceX a monthly fee from October until June 2029, with the computing power provided including approximately 110,000 NVIDIA GPUs, CPUs, memory, and other related components. The related computing power will gradually increase before September, and the costs will decrease accordingly.The documents indicate that if SpaceX fails to deliver the agreed number of GPUs by September 30, 2026, Google has the right to terminate the contract and will enjoy a one-month grace period. Additionally, according to the agreement disclosed on Friday, either party has the right to terminate the agreement with 90 days' notice.

According to Coinglass data, the Coinbase Bitcoin Premium Index has been in negative premium for 18 consecutive days, currently reported at -0.1347%, as purchasing power in the U.S. market continues to weaken.The Coinbase Bitcoin Premium Index is used to measure the difference between the Bitcoin price on Coinbase and the global market average price. A negative premium typically reflects significant selling pressure in the U.S. market, a decrease in investor risk appetite, increased market risk aversion, or capital outflows.

According to FinanceFeeds, Tether-supported large agricultural company Adecoagro plans to launch a Bitcoin mining operation powered by electricity generated from sugarcane processing in Brazil, with an initial capacity of 10 megawatts, approximately 1,280 mining machines, expected to start around July 1, 2026. Adecoagro's business encompasses sugar, ethanol, rice, dairy products, and renewable energy, and Tether previously acquired a majority stake in the company.The mining project will utilize electricity generated from sugarcane residue (bagasse), and the company has over 230 megawatts of renewable energy generation capacity in South America. Brazil is one of the world's largest sugarcane producers, with a mature ethanol and bioenergy industry. If the pilot is successful, other agricultural and renewable energy producers may view mining as a flexible demand source for surplus electricity.

According to CoinDesk, Bittensor co-founder Ala Shaabana stated at the Proof of Talk summit in Paris that Bitcoin's computing power exceeds that of the top 100 supercomputers in the world combined by 600,000 times, demonstrating the scale of distributed computing.Shaabana pointed out that true computing power no longer belongs to isolated corporate data centers, but to open, global networks. Shaabana believes that through programmatic goal setting, open networks attract talent and computing power far more efficiently than traditional enterprises. The long-term bullish logic is no longer primarily based on technology, but driven by debt, liquidity, and a decline in trust in traditional sovereign systems.