cost

Fireworks AI released an evaluation report comparing the open-source model Kimi K3 and the closed-source model Fable 5 across approximately 1030 agent tasks (covering scenarios such as SWE, terminal operations, algorithms, multilingual coding, and legal matters). The results showed that the accuracy rates for the two models on the SWE benchmark were 92.4% and 92.6%, respectively, indicating similar overall performance, but each model has its strengths in specific areas: K3 leads in symbolic mathematics and development tools, while Fable excels in web and data visualization; in long-cycle terminal tasks, K3 independently tackled 11 tasks that Fable could not solve.The report pointed out that adopting a task-level routing strategy can dynamically allocate between the two, achieving an accuracy rate of 93%, surpassing either single model. Additionally, due to K3's significant cost advantage on the Fireworks platform, costs for long-cycle agent tasks can be up to 50 times lower than Fable. The report recommends using the open-source model as the default option and continuously learning the optimal match between tasks and models through routers to balance quality and cost.

According to a report by CryptoSlate, BitMine's Ethereum staking and validation business heavily relies on the third-party operator Ethereum Tower. According to the relevant management service agreement, Ethereum Tower holds an irrevocable interest of 2% in MAVAN, and the initial term of the agreement lasts for up to 10 years; if BitMine terminates the cooperation early under non-specific default circumstances, the other party can still choose to continue receiving revenue sharing for the remaining term or receive a one-time compensation according to a formula.Documents show that in the three months ending May 31, 2026, BitMine's staking and validation business revenue was $45.743 million, accounting for 98.3% of total revenue of $46.535 million. During the same period, the company held approximately 5.4169 million ETH, valued at about $10.856 billion, of which approximately 4.7187 million had been used for staking. The report suggests that this means BitMine's revenue performance is not only dependent on Ethereum staking returns but is also significantly constrained by third-party management relationships and potential termination costs.

Ethereum Improvement Proposal EIP-8222 aims to separate staking deposits from withdrawals through a STARK-based cryptographic mechanism and to re-anonymize validators to reduce the traceability of validator activities. Currently, about one-third of ETH is staked.Thibault Dubuis, Head of Staking and Decentralized Finance Products at Sygnum Bank, stated that the current deposit addresses, validators, and withdrawal credentials create visible links, making institutional holdings, entry times, and staking strategies nearly public. This proposal is still in the discussion phase and has not yet scheduled a deployment time. Its design may involve arrangements such as fixed deposit denominations and asset withdrawal waiting periods, and institutional users may need to bear higher execution costs, process delays, and compliance work.

CryptoQuant analyst Darkfost stated that the Bitcoin bear market has lasted for about 9 months, affecting both short-term and long-term holders. The cost basis of short-term holders has fallen below that of long-term holders and has been confirmed for 3 days, triggering a "bear market nearing its end" signal. This indicator excludes BTC held for more than 7 years when calculating the cost basis of long-term holders to more accurately reflect the economically active long-term positions.Darkfost emphasized that this signal does not mean the bear market is immediately over, nor does it indicate that the market bottom has been established; rather, it suggests that the market may be entering the final phase of the bear market, during which adopting a dollar-cost averaging strategy may be more reasonable. He stated that if the cost basis of short-term holders subsequently breaks back above the cost basis of long-term holders, it can be seen as a confirmation signal for the beginning of a bull market phase and can serve as a reference for ending dollar-cost averaging. Currently, the cost basis of short-term holders has fallen from $112,500 to $69,000.

According to a report by The Paper, at the 2026 World Artificial Intelligence Conference, renowned futurist and "father of the Silicon Valley spirit" Kevin Kelly stated in a media interview that if one day everyone starts to care about the cost of tokens, the existence of open-source models would be an advantage for Chinese AI.Kevin Kelly mentioned that the cost of token consumption is becoming increasingly important, and currently, it seems that people do not care about the cost. "But I believe that when we are constantly consuming such a large number of tokens, people will start to care (about the cost). If you can provide a cost that is only one-tenth of Anthropic's, then it will change the game."However, Kevin Kelly also reminded that the open-source model requires sufficient funding to maintain operations, as it is not as profitable as the closed-source model. "Building these large models requires massive funding."

According to Jinshi reports, the Federal Reserve's Beige Book shows that respondents generally expect the U.S. economy to continue expanding in the coming months, but several districts pointed out that there remains a high level of uncertainty regarding fuel cost prospects.

According to The Block, the Benchmark Report released by Borderless.xyz for Q2 2026 shows that the delivery quotes for stablecoin cross-border payments in each month of the second quarter were all below the interbank foreign exchange rate midpoint. Data covering 108 countries and 260 payment corridors indicates that the median "parity spread" for the quarter was -3.2 basis points, further widening to -5.9 basis points in June, marking the deepest negative value of the year.The report also shows that the typical delivery cost for a $10,000 cross-border payment in the second quarter was approximately $27, and it has remained around this level for five consecutive months. If companies only connect to a single service provider in the long term, they will pay an average of about $2,330 more compared to the optimal quote for every $1 million payment scale, known as the "routing tax."Regionally, the African market showed the most significant fluctuations, with the price point spread widening by 166 basis points to 512.8 basis points; Latin America compressed to 89.0 basis points, while Asia remained relatively stable at 6.1 basis points. In specific corridors, Malawi experienced a one-day repricing of 5.8% on April 9, with the typical spread jumping from about 296 basis points to 1975 basis points.

CryptoQuant analyst Darkfost stated that the current Bitcoin price has been below the short-term holder cost basis (STH Cost Basis) for more than 9 consecutive months. Historically, this kind of prolonged loss phase for short-term holders is often highly correlated with bear market cycles.Currently, the short-term holder cost basis for Bitcoin is approximately $70,700, which continues to act as a resistance level. The market trend in May has reflected this pressure, as BTC quickly encountered a pullback after testing this area near approximately $82,000.

According to KPMG's latest survey report involving 2,145 executives from 20 countries, as technology companies like Anthropic, OpenAI, and GitHub recently shifted some of their AI services from fixed subscription models to usage-based billing, businesses are facing challenges in cost forecasting and management during the scaling of AI deployment.The report indicates that 29% of corporate executives find it difficult to understand and control operational costs when scaling AI deployment, and one-third of executives believe that insufficient understanding of AI economics hinders the deployment of AI entities. Due to costs exceeding expected value, nearly half (about 49%) of corporate organizations have chosen to delay or readjust their AI deployment plans; meanwhile, low-cost, high-fidelity large models are accelerating their impact on corporate AI strategies.In addition, tech giants are increasing capital expenditures to build AI capacity. Amazon plans to spend about $200 billion on capital expenditures this year and is investing $1 billion in its AWS frontline engineering organization to assist customers in adopting AI entities; Microsoft's total capital expenditure is expected to reach $190 billion this year, with $2.5 billion allocated to the new entity Microsoft Frontier Company. KPMG emphasizes that, in addition to cost pressures, accountability in AI governance, employee engagement rules, and the prevention of system "hallucinations" remain core challenges faced by businesses today.

According to the Industrial and Commercial Times, due to the soaring prices of key components such as memory and SSDs, the cost pressure for configuring high-capacity storage in mainstream laptops has greatly increased. Cost-effective mainland Chinese memory and storage components are gradually penetrating the PC supply chain, with brands such as Lenovo, ASUS, MSI, Gigabyte, Acer, and even the American brand Apple accelerating the certification, introduction, or platform tuning of related products.The report pointed out that Lenovo has expanded its use of mainland components this year, and recently flagship laptop models equipped with Yangtze Memory Technologies (YMTC) SSDs have appeared on North American cross-border e-commerce platforms. The American brand Apple is also reported to be negotiating with the U.S. government to procure Changxin Memory Technologies (CXMT) memory to cope with the rising prices. In terms of Taiwanese board manufacturers, MSI recently announced that it is the first to complete the verification and tuning of Changxin Memory DDR5 chips on the AMD platform at DDR5-8000+, while Gigabyte has also adopted Changxin Memory chips in some motherboard models. ASUS and Acer have introduced memory modules from mainland manufacturers such as BIWIN through their own brand memory certification or OEM models.Industry analysts believe that although the short-term imbalance in supply and demand for storage has prompted non-mainland brands to accelerate related certifications, due to the limited production capacity of mainland manufacturers, brand Taiwanese manufacturers still emphasize that Korean original manufacturers with long-term contracts remain the main supply partners at present.

According to a report by Sing Tao Daily, the Hong Kong Securities and Futures Professionals Association stated that after a meeting with representatives from the regulatory authorities, including Deputy Secretary for Financial Services and the Treasury Chen Haolian and Executive Director of the Securities and Futures Commission's Intermediaries Division Ye Zhiheng, the Securities and Futures Commission has committed to advancing measures such as the separation of the "Virtual Asset Platform Practitioners Examination" (CVAP) exam and curriculum, optimizing review materials, and reducing assessment fees.The association will follow up on the relevant timeline and continue to communicate with the Financial Services and the Treasury Bureau and the Securities and Futures Commission on practical issues such as self-custody operation guidelines for private equity funds, the boundaries of technology services and regulated activities, and the VA Payment regulatory framework.

According to the official announcement, Bitget has supported the US stock rToken spot dollar-cost averaging feature, allowing users to regularly allocate stock assets through the spot dollar-cost averaging robot. This feature supports 24/7 trading, and users can execute dollar-cost averaging through on-exchange and off-exchange matching mechanisms, regardless of whether it is a trading day or a non-trading day in the US stock market. After upgrading the app to the latest version, users can enter "Trading Robot," select "Spot Dollar-Cost Averaging," and create a dollar-cost averaging plan by setting parameters such as the investment amount and cycle.It is reported that the rToken, identified by the letter r + stock code (for example, rNVDA for NVIDIA), is issued by Reality, a licensed RWA protocol under Bitget, and is directly connected to global liquidity pools such as NASDAQ and NYSE through collaboration with the compliant brokerage Alpaca.

CryptoQuant analyst Axel Adler Jr. stated that Strategy's preferred stock MSTR has fallen 78% from its peak, while Bitcoin has dropped 51% from its peak. The average cost basis for Strategy's 847,363 BTC holdings is $75,651, with a total cost of $64.1 billion. The current BTC price has fallen below this cost line for the first time since the bear market of 2022. The additional decline of MSTR relative to BTC has reached about 28 percentage points, approaching the upper end of the historical range, but has not yet touched the extreme of an 89% retracement from the 2022 low.Meanwhile, Strategy's purchasing strategy has clearly shifted to a defensive stance: the weekly BTC purchase volume has been cut by about two-thirds, with less than 11% of the $335.5 million raised through stock issuance used to buy BTC, and the remainder transferred to dollar reserves. At the end of May, Strategy also conducted its first net sell since 2022, selling 32 BTC to pay STRC dividends. Adler pointed out that the main risk currently lies in BTC remaining below the treasury cost line of $75,000, which would block the financing channel for ATM issuance by compressing the MSTR premium. However, nearly all of Strategy's debt is in convertible bonds, with no additional margin risk; the baseline scenario is the loss of marginal buyers rather than cascading liquidations. The real pressure point lies in the company's transition from selling stock to systematically selling BTC itself to pay preferred stock dividends and debt interest.

According to The Block, Bitcoin miner revenue continues to decline, with the 7-day moving average dropping to about $30,000,000 per day, down from over $50,000,000 last summer. The contribution from transaction fees has become negligible, at less than $250,000 per day. The current trading price of BTC is about $62,500, significantly lower than JPMorgan's estimated production cost of around $78,000, a gap that has persisted for five months, the longest in this cycle. Production costs have historically been seen as a soft price floor, and currently, about 20% of miners are operating at a loss.Network-level pressure is becoming evident. Over the past six months, the beta value of mining difficulty relative to BTC price has risen to 0.62, with high-cost miners switching off their machines based on price fluctuations rather than continuing to operate at a loss. In the second week of June, difficulty decreased by 10%, marking the second similar adjustment this year. Publicly listed mining companies have sold over 32,000 BTC in the first quarter to cover operating costs, rather than further reducing production capacity.

Bitcoin miner revenue has continued to decline over the past year, with the current 7-day moving average daily income at approximately $30 million, significantly lower than last summer's level of over $50 million. Among this, transaction fees have dropped to less than $250,000 per day, almost negligible compared to block subsidies.Meanwhile, the price of Bitcoin is around $62,500, below JPMorgan's estimated production cost of about $78,000. This state of being below production costs has persisted for five months, the longest duration in this cycle. Historically, production costs are often seen as a soft bottom area for Bitcoin prices. Currently, it is estimated that about 20% of miners are in a loss position at the current price, and the pressure is beginning to reflect at the network level.Over the past six months, the sensitivity of mining difficulty to Bitcoin prices has risen to 0.62, indicating that high-cost miners are increasingly inclined to turn off their mining machines based on price fluctuations rather than continue mining at a loss. In the second week of June, Bitcoin mining difficulty decreased by 10%, marking the second occurrence of a similar magnitude adjustment this year. A comparable adjustment also occurred in the previous quarter, with both instances happening during periods when prices remained below production costs, indicating that pressure on the miner side is deepening.

According to analyst Darkfost, the realized price of short-term holders (STH) of Bitcoin is currently about $74,800, while BTC has been running below this cost line for eight consecutive months, with short-term holders still generally under pressure from losses. Currently, the average unrealized loss for STH is about 14.4%, a significant improvement from over 34% in February of this year.It is noteworthy that the cost price for STH has decreased from $95,700 to $74,800, indicating that some short-term holders have continued to accumulate during the decline, thereby lowering the overall holding cost. In this cycle, the unrealized profit for STH peaked at 47% in March 2024, while when BTC reached an all-time high in October 2025, this metric was only 11%, showing that the market's profit level is significantly weaker than in previous stages. The STH cost price of $74,800 remains a key resistance level. If BTC can regain this level, it may indicate a positive shift in market structure; however, before that, this area is more likely to become a selling pressure zone for short-term holders looking to break even.

According to Cointelegraph, former Bank for International Settlements President Agustín Carstens stated at the Point Zero Forum that stablecoins can promote financial innovation and inclusivity while reducing costs.He pointed out that efforts should be made to create conditions for the coexistence of fiat currency and stablecoins. If we hope to establish a global system where stablecoins interact with global currencies, this must be a global cooperative effort. Improved regulation and a fair competitive environment for issuers can help stablecoins thrive dramatically.

According to Rest of World, in order to significantly reduce operating costs, American developers and startups are accelerating the adoption of Chinese AI models such as DeepSeek, Minimax, and Kimi. Due to their high cost-effectiveness in handling common tasks like code writing, the market share of Chinese AI models is rapidly increasing. Data shows that in May, the Token usage share of DeepSeek on the service provider Vercel platform surged from less than 1% to 17%. Additionally, it is reported that Microsoft is also exploring the use of DeepSeek or other open-source models as a low-cost alternative for Copilot.However, despite the popularity of Chinese AI models in the market due to their price advantage, many American companies and developers choose to access these models indirectly through local servers or American cloud service providers, constrained by domestic political scrutiny and data security concerns. This has led to Chinese AI companies facing high usage rates while still struggling to effectively convert traffic into substantial direct revenue and long-term enterprise-level customer trust in the U.S. market.

CryptoQuant analyst Darkfost stated that after the recent rebound in Bitcoin prices, miner selling pressure has returned, as evidenced by a significant increase in miner inflows to Binance. Darkfost pointed out that miner hash rate has decreased by about 28% since the end of October last year, along with a 20% difficulty adjustment. Currently, the production cost is about $76,000 while the Bitcoin trading price is about $65,000, resulting in most miners operating at a loss and being forced to sell coins.