7*24

Hot topic

Market

Activities

On June 6, according to a report by Hong Kong Radio, a spokesperson for the Hong Kong Monetary Authority (HKMA) stated that Hong Kong's regulatory agencies have maintained close and regular communication with their mainland counterparts. In response to a circular from the Hong Kong Securities and Futures Commission (SFC) requiring brokers to take additional measures for the opening and management of accounts for mainland investors, the HKMA has also required banks to adopt high business standards similar to those of the SFC. The spokesperson noted that the banking sector has implemented the latest regulatory requirements outlined by the HKMA to ensure that account opening processes are compliant and orderly. The HKMA emphasized that Hong Kong banks have always maintained rigorous account opening procedures, ensuring that business operations are compliant and orderly, which is a constant regulatory requirement. Mainland residents can allocate assets and invest in eligible financial products through various legal cross-border investment channels, including the Cross-Border Wealth Management Connect, or invest through mainland brokers via the Shanghai-Hong Kong Stock Connect and Shenzhen-Hong Kong Stock Connect.

[Hong Kong Monetary Authority: Banking Sector Has Implemented Latest Regulatory Requirements to Ensure Compliance and Orderly Account Opening Processes]On June 6, according to a report by Hong Kong Radio, a spokesperson for the Hong Kong Monetary Authority (HKMA) stated that Hong Kong's regulatory agencies have maintained close and regular communication with their mainland counterparts. In response to a circular from the Hong Kong Securities and Futures Commission (SFC) requiring brokers to take additional measures for the opening and management of accounts for mainland investors, the HKMA has also required banks to adopt high business standards similar to those of the SFC. The spokesperson noted that the banking sector has implemented the latest regulatory requirements outlined by the HKMA to ensure that account opening processes are compliant and orderly. The HKMA emphasized that Hong Kong banks have always maintained rigorous account opening procedures, ensuring that business operations are compliant and orderly, which is a constant regulatory requirement. Mainland residents can allocate assets and invest in eligible financial products through various legal cross-border investment channels, including the Cross-Border Wealth Management Connect, or invest through mainland brokers via the Shanghai-Hong Kong Stock Connect and Shenzhen-Hong Kong Stock Connect.

Hong Kong Monetary Authority: Banking Sector Has Implemented Latest Regulatory Requirements to Ensure Compliance and Orderly Account Opening Processes

According to monitoring by Dongcha Beating, a joint team consisting of Shenzhen Hetao Institute, Harbin Institute of Technology (Shenzhen), Shenzhen Big Data Research Institute, and Huawei's related teams, in collaboration with the Deep Intelligence City AI computing power platform, has announced the successful post-training of the 16 trillion parameter large model DeepSeek-V4-Pro on a domestic AI computing power platform. This marks the first time a third-party organization has completed full parameter post-training of a model at the scale of 16 trillion parameters on a domestic computing platform. Unlike pre-training, which starts from scratch, the post-training phase (mainly including supervised fine-tuning SFT and reinforcement learning RL) focuses on teaching the model to follow instructions and perform specific tasks through high-quality instructions and alignment with human preferences. However, for a 16 trillion parameter MoE architecture model, full parameter post-training still imposes stringent requirements on the underlying hardware's memory capacity, inter-card communication bandwidth (such as the all-to-all communication triggered by MoE routing), and the stability of large-scale clusters. The joint team leveraged a Huawei Ascend 910C computing cluster with over a thousand chips, successfully overcoming communication bottlenecks by optimizing distributed load-bearing and load-balancing strategies. Throughout the training process, which lasted over 1500 steps, the system did not experience a single interruption, achieving a model compute utilization (MFU) of over 30%, with a 14% improvement in key operator efficiency, and all metrics meeting industrial-grade operational standards. Industry analysts point out that the successful training of trillion-scale models on the Huawei Ascend 910C cluster confirms the technical feasibility of domestic AI chips in handling ultra-large-scale model deep training tasks. Previously, the core pre-training of large models heavily relied on NVIDIA GPU clusters, with domestic computing power mainly undertaking inference or small parameter fine-tuning tasks. The success of this joint effort marks a significant transition in the domestic computing power ecosystem from 'only supporting inference' to 'supporting full parameter training of ultra-large parameter models.'

[Milestone in Domestic Computing Power: Ascend 910C Successfully Trains 16 Trillion Parameter Model DeepSeek]According to monitoring by Dongcha Beating, a joint team consisting of Shenzhen Hetao Institute, Harbin Institute of Technology (Shenzhen), Shenzhen Big Data Research Institute, and Huawei's related teams, in collaboration with the Deep Intelligence City AI computing power platform, has announced the successful post-training of the 16 trillion parameter large model DeepSeek-V4-Pro on a domestic AI computing power platform. This marks the first time a third-party organization has completed full parameter post-training of a model at the scale of 16 trillion parameters on a domestic computing platform. Unlike pre-training, which starts from scratch, the post-training phase (mainly including supervised fine-tuning SFT and reinforcement learning RL) focuses on teaching the model to follow instructions and perform specific tasks through high-quality instructions and alignment with human preferences. However, for a 16 trillion parameter MoE architecture model, full parameter post-training still imposes stringent requirements on the underlying hardware's memory capacity, inter-card communication bandwidth (such as the all-to-all communication triggered by MoE routing), and the stability of large-scale clusters. The joint team leveraged a Huawei Ascend 910C computing cluster with over a thousand chips, successfully overcoming communication bottlenecks by optimizing distributed load-bearing and load-balancing strategies. Throughout the training process, which lasted over 1500 steps, the system did not experience a single interruption, achieving a model compute utilization (MFU) of over 30%, with a 14% improvement in key operator efficiency, and all metrics meeting industrial-grade operational standards. Industry analysts point out that the successful training of trillion-scale models on the Huawei Ascend 910C cluster confirms the technical feasibility of domestic AI chips in handling ultra-large-scale model deep training tasks. Previously, the core pre-training of large models heavily relied on NVIDIA GPU clusters, with domestic computing power mainly undertaking inference or small parameter fine-tuning tasks. The success of this joint effort marks a significant transition in the domestic computing power ecosystem from 'only supporting inference' to 'supporting full parameter training of ultra-large parameter models.'

Milestone in Domestic Computing Power: Ascend 910C Successfully Trains 16 Trillion Parameter Model DeepSeek

HTX data monitoring shows that AAVE has fallen below 60 USDT, currently reported at 59.84 USDT, with a 24H decline of 13.27%.

【AAVE drops below 60 USDT】HTX data monitoring shows that AAVE has fallen below 60 USDT, currently reported at 59.84 USDT, with a 24H decline of 13.27%.

AAVE drops below 60 USDT

On June 6, the US Central Command issued a statement indicating that on June 5 Eastern Time (June 6 Beijing Time), US forces intercepted multiple ballistic missiles and drones launched by Iran towards the Strait of Hormuz and neighboring Gulf countries. The statement reported that Iran fired 7 ballistic missiles at Kuwait and Bahrain. Initial assessments show that 6 of the 7 missiles launched by Iran were intercepted, while the 7th deviated from its intended target. There have been no reports of US military casualties, and Iran's claim of 'hitting the US Fifth Fleet headquarters in Bahrain' is 'false.' The statement also noted that hours earlier, US forces shot down 4 one-way attack drones launched by Iran, which were heading towards the Strait of Hormuz—'these attack drones posed a direct threat to regional maritime traffic.' Subsequently, US forces targeted Iranian coastal surveillance radar stations located in Goruk and Qeshm Island to prevent further maritime attacks. The statement emphasized that US Central Command forces 'remain highly vigilant and prepared to respond to Iran's unwarranted aggression and to exercise self-defense.' (CCTV International News)

[US Military Claims Iran Fired 7 Missiles at Kuwait and Bahrain]On June 6, the US Central Command issued a statement indicating that on June 5 Eastern Time (June 6 Beijing Time), US forces intercepted multiple ballistic missiles and drones launched by Iran towards the Strait of Hormuz and neighboring Gulf countries. The statement reported that Iran fired 7 ballistic missiles at Kuwait and Bahrain. Initial assessments show that 6 of the 7 missiles launched by Iran were intercepted, while the 7th deviated from its intended target. There have been no reports of US military casualties, and Iran's claim of 'hitting the US Fifth Fleet headquarters in Bahrain' is 'false.' The statement also noted that hours earlier, US forces shot down 4 one-way attack drones launched by Iran, which were heading towards the Strait of Hormuz—'these attack drones posed a direct threat to regional maritime traffic.' Subsequently, US forces targeted Iranian coastal surveillance radar stations located in Goruk and Qeshm Island to prevent further maritime attacks. The statement emphasized that US Central Command forces 'remain highly vigilant and prepared to respond to Iran's unwarranted aggression and to exercise self-defense.' (CCTV International News)

US Military Claims Iran Fired 7 Missiles at Kuwait and Bahrain

On June 6, on-chain analyst @ai_9684xtpa reported that an entity linked to a16z has accumulated 90,091 HYPE in the last 10 hours, valued at $5.238 million. The total amount accumulated in 2026 has reached 6.996 million HYPE, with a total value of $327 million and an average price of $46.85.

[a16z-Linked Entity Accumulates 90,091 HYPE Worth $5.238 Million in Last 10 Hours]On June 6, on-chain analyst @ai_9684xtpa reported that an entity linked to a16z has accumulated 90,091 HYPE in the last 10 hours, valued at $5.238 million. The total amount accumulated in 2026 has reached 6.996 million HYPE, with a total value of $327 million and an average price of $46.85.

a16z-Linked Entity Accumulates 90,091 HYPE Worth $5.238 Million in Last 10 Hours

According to monitoring by Beating, Elon Musk's xAI has been engaged in a long-term cat-and-mouse game with Anthropic. To surpass its competitor, xAI launched a months-long distillation project, directly training its code model on the outputs of Claude. Even after Anthropic blocked official access channels, guerrilla-style data extraction continued. In January of this year, xAI co-founder Tony Wu informed employees that the official channel had been cut off. Subsequently, xAI engineers accessed the Claude interface through personal accounts, prompting Anthropic to ban associated accounts. After the accounts were blocked, xAI turned to the third-party encrypted intermediary platform Blackbox AI to indirectly call the interface, continuing to obtain data for benchmarking until mid-May 2026. In May, Musk admitted in court during a lawsuit with OpenAI that part of Grok's training used competitor models, claiming that cross-model training is a norm in the industry. This cat-and-mouse struggle evolved into a dramatic power leasing deal in May. Facing a computational capacity bottleneck, Anthropic decided to rent data center power from xAI at a price of $1.25 billion per month. Although Musk has repeatedly stated that the power rental is only a short-term measure, the influx of funds alleviated the loss pressure on xAI's department. While the power contract may not offset the massive losses incurred by xAI in the first quarter, it provided a key AI growth narrative for SpaceX's upcoming IPO sprint on Wall Street next week.

[xAI and Anthropic's Cat-and-Mouse Game: Distilling Claude While Becoming a Power Landlord]According to monitoring by Beating, Elon Musk's xAI has been engaged in a long-term cat-and-mouse game with Anthropic. To surpass its competitor, xAI launched a months-long distillation project, directly training its code model on the outputs of Claude. Even after Anthropic blocked official access channels, guerrilla-style data extraction continued. In January of this year, xAI co-founder Tony Wu informed employees that the official channel had been cut off. Subsequently, xAI engineers accessed the Claude interface through personal accounts, prompting Anthropic to ban associated accounts. After the accounts were blocked, xAI turned to the third-party encrypted intermediary platform Blackbox AI to indirectly call the interface, continuing to obtain data for benchmarking until mid-May 2026. In May, Musk admitted in court during a lawsuit with OpenAI that part of Grok's training used competitor models, claiming that cross-model training is a norm in the industry. This cat-and-mouse struggle evolved into a dramatic power leasing deal in May. Facing a computational capacity bottleneck, Anthropic decided to rent data center power from xAI at a price of $1.25 billion per month. Although Musk has repeatedly stated that the power rental is only a short-term measure, the influx of funds alleviated the loss pressure on xAI's department. While the power contract may not offset the massive losses incurred by xAI in the first quarter, it provided a key AI growth narrative for SpaceX's upcoming IPO sprint on Wall Street next week.

xAI and Anthropic's Cat-and-Mouse Game: Distilling Claude While Becoming a Power Landlord

On June 6, according to monitoring by Onchain Lens, the government of Bhutan transferred 738 BTC to a newly created wallet, valued at approximately 45 million USD.

[Bhutan Government Transfers 738 BTC to New Wallet]On June 6, according to monitoring by Onchain Lens, the government of Bhutan transferred 738 BTC to a newly created wallet, valued at approximately 45 million USD.

Bhutan Government Transfers 738 BTC to New Wallet

On June 6, Cathie Wood, CEO of ARK Invest, stated that the latest employment report was very strong. Non-farm payrolls increased by 172,000, while the market had expected an increase of 88,000; previous months' data was also revised upward by 93,000. Wage growth was approximately 0.3%. However, the market experienced a decline. In our view, the market is misreading this signal. It assumes that stronger-than-expected employment and economic growth will lead to accelerating inflation. But historical experience shows that this is often not the case. Currently, productivity growth is close to 3%, while unit labor costs are about 0.5%. These figures are not typical characteristics of inflationary booms but rather indicate healthy, productivity-driven growth, which tends to suppress inflation. Meanwhile, despite oil prices rising about 55% year-on-year when calculated using a three-month moving average, the yield curve continues to flatten. In past cycles, such a significant energy shock would typically lead to a steepening of the yield curve under a loose monetary policy by the Federal Reserve. However, this time, the bond market seems to be pricing in a more powerful force: the deflationary effects of technological innovation, particularly as artificial intelligence begins to enhance productivity across various sectors of the economy. If tensions with Iran ease and oil prices retreat, we believe inflation could even enter negative territory by the end of the year. In our view, the Federal Reserve's significant interest rate hikes in 2022, in response to inflation primarily driven by supply shocks, were a historic policy error. We do not believe that the next generation of monetary policymakers will be willing to repeat this mistake. Notably, gold peaked on the day Kevin Walsh was appointed. Inflation trading may have already ended. If our research is correct, the next phase of this economic cycle may manifest as: accelerating economic growth, declining inflation, lower interest rates, and a stronger dollar. This combination will create a highly favorable environment for innovation-driven stocks and technologies that propel the next leap in productivity.

[Cathie Wood: Next Phase of Economic Cycle to Feature Accelerating Growth, Declining Inflation, Lower Rates, and Stronger Dollar]On June 6, Cathie Wood, CEO of ARK Invest, stated that the latest employment report was very strong. Non-farm payrolls increased by 172,000, while the market had expected an increase of 88,000; previous months' data was also revised upward by 93,000. Wage growth was approximately 0.3%. However, the market experienced a decline. In our view, the market is misreading this signal. It assumes that stronger-than-expected employment and economic growth will lead to accelerating inflation. But historical experience shows that this is often not the case. Currently, productivity growth is close to 3%, while unit labor costs are about 0.5%. These figures are not typical characteristics of inflationary booms but rather indicate healthy, productivity-driven growth, which tends to suppress inflation. Meanwhile, despite oil prices rising about 55% year-on-year when calculated using a three-month moving average, the yield curve continues to flatten. In past cycles, such a significant energy shock would typically lead to a steepening of the yield curve under a loose monetary policy by the Federal Reserve. However, this time, the bond market seems to be pricing in a more powerful force: the deflationary effects of technological innovation, particularly as artificial intelligence begins to enhance productivity across various sectors of the economy. If tensions with Iran ease and oil prices retreat, we believe inflation could even enter negative territory by the end of the year. In our view, the Federal Reserve's significant interest rate hikes in 2022, in response to inflation primarily driven by supply shocks, were a historic policy error. We do not believe that the next generation of monetary policymakers will be willing to repeat this mistake. Notably, gold peaked on the day Kevin Walsh was appointed. Inflation trading may have already ended. If our research is correct, the next phase of this economic cycle may manifest as: accelerating economic growth, declining inflation, lower interest rates, and a stronger dollar. This combination will create a highly favorable environment for innovation-driven stocks and technologies that propel the next leap in productivity.

Cathie Wood: Next Phase of Economic Cycle to Feature Accelerating Growth, Declining Inflation, Lower Rates, and Stronger Dollar

On June 6, according to Iran's Tasnim News Agency, the public relations department of the Iranian Revolutionary Guard stated that enemy bases in the region were targeted by aerospace missiles due to the aggression of US forces against Sirik and Qeshm Island. Thick smoke was seen rising from the headquarters of the US Navy's Fifth Fleet located in Bahrain.

[Iranian Revolutionary Guard Strikes US Military Bases]On June 6, according to Iran's Tasnim News Agency, the public relations department of the Iranian Revolutionary Guard stated that enemy bases in the region were targeted by aerospace missiles due to the aggression of US forces against Sirik and Qeshm Island. Thick smoke was seen rising from the headquarters of the US Navy's Fifth Fleet located in Bahrain.

Iranian Revolutionary Guard Strikes US Military Bases

HTX data monitoring shows that HTX has surpassed 0.0000018 USDT, currently reported at 0.000001801 USDT, with a 24H increase of 1.87%.

【HTX breaks through 0.0000018 USDT】HTX data monitoring shows that HTX has surpassed 0.0000018 USDT, currently reported at 0.000001801 USDT, with a 24H increase of 1.87%.

HTX breaks through 0.0000018 USDT

Bitcoin Correction Pushes 580,000 BTC Into Loss Territory

One Article Deconstructs the Investment Methodology of 'Stock God Serenity'

Deconstructing the Investment Methodology of the 'Stock God Serenity' in One Article

Sharplink CEO: Ethereum's Future Is Playing Out Now

Kalshi and Coinbase Receive CFTC Approval, Ushering in the Most Regulation-Friendly Era for the Crypto Industry?

Weekly Token Unlocks: HYPE Unlocks $650M Worth of Tokens

A $300 Used Mining Rig Mined a $230,000 Bitcoin Block

Selling 32 BTC to Pay Dividends: Strategy Breaks the 'Never Sell' Vow

Over $1 Billion in Cryptocurrencies Forced to Liquidate, Bitcoin Plunges 6% Below $67,000

Bulls Approach the Ceiling

Article

NVIDIA CEO Jensen Huang Visits Korea, Meets with Faker: Korea as the Origin of Esports

HTX breaks through 0.0000018 USDT

Застереження：Наведені вище матеріали не представляють позицію компанії HTX.，HTX не дає жодних торгових рекомендацій.。

Схожі статті

Усі коментарі0НовіПопулярно

Популярні теми

Схожі статті