China gives DeepSeek conditional OK for Nvidia H200 chips

China has conditionally approved its leading AI startup DeepSeek to buy Nvidia’s H200 AI chips, with regulatory requirements still being finalised. The decision would add DeepSeek to a growing list of Chinese firms seeking access to the H200, one of Nvidia’s most powerful data-centre chips.

The reported approval follows earlier developments in which ByteDance, Alibaba and Tencent were allowed to purchase more than 400,000 H200 chips in total, suggesting Beijing is moving from broad caution to selective, case-by-case permissions. Separate coverage has described the approvals as a shift after weeks of uncertainty over whether China would allow imports, even as US export licensing was moving forward.

Nvidia’s CEO Jensen Huang, speaking in Taipei, said the company had not received confirmation of DeepSeek’s clearance and indicated the licensing process is still being finalised, underscoring the uncertainty for suppliers and buyers. China’s industry and commerce ministries have been involved in approvals, with conditions reportedly shaped by the state planner, the National Development and Reform Commission.

The H200 has become a high-stakes flashpoint in US-China tech ties because access to top-tier chips directly affects AI capability and competitiveness. US political scrutiny is also rising: a senior US lawmaker has alleged Nvidia provided technical support that helped DeepSeek develop advanced models later used by China’s military, according to a letter published by the House Select Committee on China; Nvidia has pushed back against such claims in subsequent reporting.

DeepSeek is also preparing a next-generation model, V4, expected in mid-February, according to reporting that cited people familiar with the matter, which makes access to high-end compute especially consequential for timelines and performance.

Why does it matter?

If China’s conditional approvals translate into real shipments, they could ease a key bottleneck for Chinese AI development while extending Nvidia’s footprint in a market constrained by geopolitics. At the same time, the episode highlights how AI hardware is now regulated not only by Washington’s export controls but also by Beijing’s import approvals, with companies caught between shifting policy priorities. With US lawmakers pressing for tighter safeguards and China balancing demand against strategic risk, the way these H200 decisions are implemented may signal whether the two sides are entering a more permissive phase for certain ‘frontier’ chips or simply refining a narrower, more controlled pipeline.

China moves toward data centres in orbit

China is planning to develop large-scale space-based data centres over the next five years as part of a broader push to support AI development. The China Aerospace Science and Technology Corporation (CASC) has announced plans to build gigawatt-class digital infrastructure in orbit, according to Chinese state broadcaster CCTV.

Under CASC’s five-year development plan, the space data centres are expected to combine cloud, edge and terminal technologies, allowing computing power, data storage and communication capacity to operate as an integrated system. The aim is to create high-performance infrastructure capable of supporting advanced AI workloads beyond Earth.

The initiative follows a recent CASC policy proposal calling for solar-powered, gigawatt-scale space-based hubs to supply energy for AI processing. The proposal aligns with China’s upcoming 15th Five-Year Plan, which is set to place AI at the centre of national development priorities.

China has already taken early steps in this direction. In May 2025, Zhejiang Lab launched 12 low Earth orbit satellites to form the first phase of its ‘Three-Body Computing Constellation.’ The research institute plans to eventually deploy around 2,800 satellites, targeting a total computing power of 1,000 peta operations per second.

Interest in space-based data centres is growing globally. European aerospace firm Thales Alenia Space has been studying its feasibility since 2023, while companies such as SpaceX, Blue Origin, and several startups in the US and the UAE are exploring similar concepts at varying stages of development and ambition.

Supporters argue that space data centres could reduce environmental impacts on Earth, benefit from constant solar energy and simplify cooling. However, experts warn that operating in space brings its own challenges, including exposure to radiation, solar flares and space debris, as well as higher costs and greater difficulty when repairs are needed.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

China’s AI battle heats up with Moonshot AI’s Kimi 2.5 upgrade

Moonshot AI has released an upgraded version of its Kimi model, intensifying competition among China’s leading AI developers. The launch comes ahead of an anticipated major release from rival DeepSeek.

The new model, K2.5, can process text, images, and video within a single prompt. Moonshot said the system improves performance across benchmarks and strengthens coding capabilities.

Several AI firms in China have rolled out upgrades in recent weeks. Companies including Alibaba, Zhipu, and MiniMax have introduced new models focused on reasoning, images, and robotics.

Investment has surged alongside the technology push. Moonshot recently raised $500 million, while rivals secured more than $1 billion through listings and funding rounds.

Analysts say the rapid pace highlights mounting pressure across the Chinese AI sector. Smaller developers are struggling to keep up with rising costs and the pace of innovation.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Meteorological authorities in China embrace AI for next-gen climate risk prediction

At a national meteorological work conference, the China Meteorological Administration said it will pilot an ‘imminent warning’ system and apply AI technologies to enhance detailed forecasts for extreme weather, including typhoons and heavy rain.

The initiative is part of a broader effort in 2026 to build new meteorological service systems, such as for agriculture, and improve disaster preparedness and climate risk management across the country.

Officials highlighted progress over the past year, including improved flood-season forecasting and reduced typhoon track-prediction errors. Strengthened interagency coordination and the development of new prediction products aim to support earlier warnings and better resource allocation for extreme climate events.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Zurich researchers link AI with spirituality studies

Researchers at the University of Zurich have received a Postdoc Team Award for SpiritRAG, an AI system designed to analyse religion and spirituality in United Nations documents. The interdisciplinary project brings together expertise from Zurich across computer science, linguistics, education and spiritual care.

SpiritRAG connects large language models with more than 7,500 UN texts, allowing users in Zurich and beyond to ask context sensitive questions grounded in original sources. The system addresses challenges where meaning varies across cultures, history and political settings.

The Zurich based team presented SpiritRAG at EMNLP 2025 in Suzhou, China, and later at the AI+X Summit in Zurich. Interest from organisations outside Zurich highlights demand for transparent AI tools supporting research and policy analysis.

Designed as open source infrastructure, SpiritRAG allows deployment with different datasets while using limited resources. Researchers in Zurich say the approach supports responsible AI use in complex domains where accuracy and context remain critical.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

China gains ground in global AI race

US companies are increasingly adopting Chinese AI models as part of their core technology stacks, raising questions about global leadership in AI. In the US, Pinterest has confirmed it is using Chinese-developed models to improve recommendations and shopping features.

In the US, executives point to open-source Chinese models such as DeepSeek and tools from Alibaba as faster, cheaper and easier to customise. US firms say these models can outperform proprietary alternatives at a fraction of the cost.

Adoption extends beyond Pinterest in the US, with Airbnb also relying on Chinese AI to power customer service tools. Data from Hugging Face shows Chinese models frequently rank among the most downloaded worldwide, including across US developers.

Researchers at Stanford University have found Chinese AI capabilities now match or exceed global peers. In the US, firms such as OpenAI and Meta remain focused on proprietary systems, leaving China to dominate open-source AI development.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

TikTok restructures operations for US market

TikTok has finalised a deal allowing the app to continue operating in America by separating its US business from its global operations. The agreement follows years of political pressure in the US over national security concerns.

Under the arrangement, a new entity will manage TikTok’s US operations, with user data and algorithms handled inside the US. The recommendation algorithm has been licensed and will now be trained only on US user data to meet American regulatory requirements.

Ownership of TikTok’s US business is shared among American and international investors, while China-based ByteDance retains a minority stake. Oracle will oversee data security and cloud infrastructure for users in the US.

Analysts say the changes could alter how the app functions for the roughly 200 million users in the US. Questions remain over whether a US-trained algorithm will perform as effectively as the global version.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

EU cyber rules target global tech dependence

The European Union has proposed new cybersecurity rules aimed at reducing reliance on high-risk technology suppliers, particularly from China. In the European Union, policymakers argue existing voluntary measures failed to curb dependence on vendors such as Huawei and ZTE.

The proposal would introduce binding obligations for telecom operators across the European Union to phase out Chinese equipment. At the same time, officials have warned that reliance on US cloud and satellite services also poses security risks for Europe.

Despite increased funding and expanded certification plans, divisions remain within the European Union. Countries including Germany and France support stricter sovereignty rules, while others favour continued partnerships with US technology firms.

Analysts say the lack of consensus in the European Union could weaken the impact of the reforms. Without clear enforcement and investment in European alternatives, Europe may struggle to reduce dependence on both China and the US.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

Amodei warns US AI chip exports to China risk national security

Anthropic chief executive Dario Amodei has criticised the US decision to allow the export of advanced AI chips to China, warning it could undermine national security. Speaking at the World Economic Forum 2026 in Davos, he questioned whether selling US-made hardware abroad strengthens American influence.

Amodei compared the policy to ‘selling nuclear weapons to North Korea‘, arguing that exporting cutting-edge chips risks narrowing the technological gap between the United States and China. He said Washington currently holds a multi-year lead in advanced chipmaking and AI infrastructure.

Sending powerful hardware overseas could accelerate China’s progress faster than expected, Amodei told Bloomberg. He warned that AI development may soon concentrate unprecedented intelligence within data centres controlled by individual states.

Amodei said AI should not be treated like older technologies such as telecoms equipment. While spreading US technology abroad may have made sense in the past, he argued AI carries far greater strategic consequences.

The debate follows recent rule changes allowing some advanced chips, including Nvidia’s H200 and AMD’s MI325X, to be sold to China. The US administration later announced plans for a 25% tariff on AI chip exports, adding uncertainty for US semiconductor firms.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Forced labour data opened to the public

Exiger has launched a free online tool designed to help organisations identify links to forced labour in global supply chains. The platform, called forcedlabor.ai, was unveiled during the annual meeting of the World Economic Forum in Davos.

The tool allows users to search suppliers and companies to assess potential exposure to state-sponsored forced labour, with an initial focus on risks linked to China. Exiger says the database draws on billions of records and is powered by proprietary AI to support compliance and ethical sourcing.

US lawmakers and human rights groups have welcomed the initiative, arguing that companies face growing legal and reputational risks if their supply chains rely on forced labour. The platform highlights risks linked to US import restrictions and enforcement actions.

Exiger says making the data freely available aims to level the playing field for smaller firms with limited compliance budgets. The company argues that greater transparency can help reduce modern slavery across industries, from retail to agriculture.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot