On July 21, the Financial Times reported, citing sources involved in the discussions, that Chinese regulators are considering tightening export controls on AI and semiconductor technologies. The Ministry of Commerce has held discussions with AI companies such as Alibaba, ByteDance, and Zhipu to restrict the transfer of key data used for model training overseas and to limit foreign users from downloading model weights. However, China will still allow overseas customers to access these models and services. The Ministry of Commerce has also sought feedback on potential restrictions that would prohibit overseas chipmakers like Qualcomm and TSMC from manufacturing advanced semiconductors based on designs developed by companies such as Huawei, Alibaba, and ByteDance. According to sources, most proposals are still under discussion, and regulators are evaluating industry feedback before making a final decision.
Welcome Back
Click “Sign in” to agree to Cointime’s Terms of Service and acknowledge that Cointime’s Privacy Policy applies to you.
Join CoinTime
Already have an account?
Click “page.Sign up” to agree to Cointime’s <a class="underline" href="#term-of-service">Terms of Service</a> and acknowledge that Cointime’s a class="underline" href="#privacy-policy">Privacy Policy</a> applies to you.
Sign in with email
Sign up with email
Your email
Check your inbox
Click the link we sent to to sign in.
Click the link we sent to to sign up.
China Reportedly Considers Tightening Export Controls on AI Models and Chips
-
Wechat scan to share
Recommended for you
-
Beijing to Establish Token Factories in the Second Half of the Year
On July 21, the Beijing Municipal Bureau of Economy and Information Technology reported that in the first half of the year, the city's digital economy grew by 7.8%, with the core industries of the digital economy increasing by 9.8%, significantly contributing to the city's GDP growth. In the second half of the year, Beijing will promote comprehensive empowerment through artificial intelligence, advancing applications such as intelligent translation and simultaneous interpretation in the cultural and tourism consumption sectors. The Bureau will focus on the annual key tasks for building a benchmark city for the digital economy, implementing the 'Artificial Intelligence +' action plan, and fully unleashing new momentum for the intelligent economy. Policies for the development of the Token economy will be formulated, focusing on key aspects such as Token production, distribution, and application, with plans to establish Token factories and distribution platforms, promoting innovative applications in key areas such as industry, education, and cultural tourism. An 'Innovate for the Future' OPC special roadshow event will be held to stimulate the creative energy of super individuals in AI applications. High-quality training courses on AIGC aimed at OPC will be developed to unlock the value of AIGC technology. Leveraging the Open Source Chip Research Institute and the Beijing Tongminghu Information Technology Application Innovation Center, a 'RISC-V + AI OS' open-source ecosystem will be created, building a full-stack autonomous technology system from chip instruction sets to operating systems to intelligent applications. (Xinjingbao) -
Changxin Technology: Online Investors Abandon Subscription of 6,586,227 Shares
On July 21, Changxin Technology (688825.SH) announced the results of its initial public offering on the STAR Market. The issue price was 8.66 yuan per share, with an initial offering of 6,688,088,608 shares, accounting for approximately 10% of the total post-issuance share capital. The final strategic placement was 1,667,064,720 shares, and the final online issuance lot-winning rate was approximately 0.47141739%. Online investors subscribed and paid for 3,844,517,273 shares and abandoned 6,586,227 shares; offline investors subscribed and paid for 2,173,101,821 shares and abandoned 31,567 shares. The joint lead underwriters underwrote 6,617,794 shares, with an underwriting amount of 57.3101 million yuan. Before the exercise of the over-allotment option, the issuance expenses were 281 million yuan. -
Insiders: China and US to Hold AI Talks in September on Frontier Model Regulation
According to Reuters citing five informed sources, China and the United States plan to hold AI-related talks in September to discuss regulatory risks posed by increasingly powerful frontier AI models. Sources said the US side is expected to be led by Treasury Secretary Scott Bessent, and both sides are still finalizing the participants, agenda, and meeting location. (CNBC) -
US President Trump May Impose New Tariffs on Dozens of Countries Within This Week
July 21, according to the Financial Times: US President Trump may impose new tariffs on dozens of countries within this week. -
Montage Technology: DDR5 RCD Chip Shipments Surge; Rising CPU Demand to Expand Memory Interconnect Chip Market
On July 21, Montage Technology (688008.SH) released an investor relations activity record announcement. The company benefited from the AI industry trend in the first half of the year, with strong industry demand, increased DDR5 penetration rate and continuous sub-generation iterations. The company's DDR5 RCD chip shipments increased significantly, with the proportion of third and fourth generation RCD chip shipments further rising. Revenue from new interconnect chip products such as MRCD/MDB, PCIe Retimer, CKD, and CXL MXC chips also climbed significantly. The company plans to complete the engineering development of the third-generation MRCD/MDB chip this year. The industry expects MRDIMM to start from large-scale application in the next two to three years and gradually enter the ramp-up penetration phase. Increasing CPU demand will drive up memory module usage, thereby expanding the memory interconnect chip market. In addition, the growing importance of CPUs in AI system architecture will synchronously drive demand for other CPU-related interconnect chips, including PCIe and CXL interconnect chips. The CXL 3.1 MXC chip has been sent to key customers for sample testing, and its penetration rate in data centers is expected to increase rapidly in the coming years. The PCIe 6.x/CXL 3.x Retimer chip is actively cooperating with customers for testing and validation, with the industry expecting it to enter the large-scale application stage by 2027. The company plans to complete the tape-out of the Ethernet PHY Retimer chip engineering samples this year. -
National Supercomputing Internet Launches Kimi K3 API Service
July 21 news, the National Supercomputing Internet recently launched the Kimi K3 model API call service. Enterprises and developers can quickly call the large model capabilities without tedious environment configuration. Currently, the API is compatible with OpenAI and Anthropic interface specifications, supporting rapid and low-cost construction of diverse AI applications. In addition, the Supercomputing Internet also launched the '100,000 Blocks' co-creation plan, opening a one-stop MaaS service. (Interface News) -
BlackRock, Hillhouse, and Temasek Reportedly to Become Cornerstone Investors for Zhongji Innolight's Hong Kong IPO
On July 21, according to Bloomberg, people familiar with the matter said that BlackRock, Hillhouse, and Temasek will become cornerstone investors for Zhongji Innolight's potential $8 billion Hong Kong listing plan. These cornerstone investors are expected to subscribe to nearly half of the issuance size, in exchange for guaranteed allocation quotas to hold the stock for a fixed period (usually six months). Zhongji Innolight plans to start accepting investor subscriptions as early as Wednesday, which could become the largest equity offering in Hong Kong in seven years. -
Spot Silver Surges 5.00% Intraday, Now at $59.23/oz
Spot silver surged 5.00% intraday, currently trading at $59.23 per ounce. (Gold Ten) -
S&P: Revises Samsung Electronics Outlook to Positive
S&P: Revises Samsung Electronics Outlook to Positive, Affirms Ratings at 'AA-/A-1+' -
Kimi Head Addresses Three Controversies: SOTA Models Should Have Their Own Pricing Power
According to 21 Finance, on July 21, Kimi's enterprise business head Huang Zhenxin responded to market concerns. Huang clearly stated: 'Open-source models and Chinese large models should not be labeled as low-cost. We have developed a SOTA-level model and can justify reasonable commercial pricing.' Addressing external speculation that 'K3 is a distilled small model,' Huang clarified that the core of K3's performance leap comes from underlying original architectural innovation, not from distilling or replicating existing models. On the commercialization front, Huang revealed that Kimi currently does not offer private deployment services. Its business model benchmarks mature overseas AI leaders, focusing on refining fundamental model capabilities. API business accounts for the highest proportion of the company's revenue; specific revenue breakdowns for private deployment and industry solutions have not been disclosed. Meanwhile, he outlined the trend in model parameter scale: large model parameter sizes will continue to expand upward and will not stop at the current 2.8 trillion. Addressing the open-source vs. closed-source debate, Huang believes the two are not opposing competitive relationships. Enterprise clients have differentiated needs: those requiring private deployment and fine-tuning prefer open-source, while those seeking stable managed services choose closed-source. Open-source or closed-source itself is not inherently tied to model token quality. Kimi has adhered to an open-source route since K2, and this approach will not change in the future. Core underlying technologies and papers are publicly disclosed, maintaining an open attitude toward the industry. Regarding Musk's claim that Grok 4.5 might surpass K3, Huang responded: 'We'll wait and see.' When asked about the difference between Kimi's open-source model and other domestic large models, Huang explained: The global open-source industry has standardized licensing agreements like MIT; Kimi adopts internationally accepted open-source licenses. The misconception that Kimi cannot be privately deployed stems from the current massive parameter size of K3 and high deployment thresholds, which is unrelated to the open-source licensing strategy itself.
Daily Must-Read
-
ENI Takes the Stage at WebX Japan and Builds a Bridge from Traditional Capital to the Future Economy
-
GWDC 2026 KOREA: The New Asian Paradigm of Web3 & AI Sets to Debut in Seoul on Sept 29-30, Inviting Global Builders to the Ultimate Developer Summit
-
From User Journey to Consumption Loop: Understanding UniKey’s Product Logic and the Three Core Use Cases of KEY
-
From AI Router to Agent Economy Network: How UniKey Builds the Next Generation of AI Infrastructure
-
SuperStrike: As AI Takes Over Financial Decision-Making, A New Era of Wealth Creation Begins
-
The largest IPO in history! Panoramic analysis of SpaceX's IPO: trillion dollar valuation, track change, and Musk's trillion dollar net worth
All Comments