AI Sector Daily Digest — June 29, 2026

June 29 scan

1. South Korea put $576 billion behind an AI-chip plan

• South Korea unveiled an industrial strategy centered on semiconductors and AI, with President Lee Jae Myung presenting more than $576 billion in planned investment. 1
• Samsung, SK Hynix and suppliers are expected to invest 800 trillion won in four new fabs in the southwest, plus another 81 trillion won for a chip-packaging cluster. 1
• The plan ties high-bandwidth memory, DRAM, physical AI and data centers into one national capacity push; Reuters also notes power, water, logistics and skilled labor as execution risks. 1

2. CXMT reportedly won a $2.94 billion Tencent memory deal

• Reuters reported that China's CXMT signed a long-term supply agreement with Tencent worth more than 20 billion yuan, or about $2.94 billion. 2
• The reported deal covers DRAM for Tencent servers and cloud systems, placing domestic memory supply closer to China's AI-compute stack. 2
• The practical read: China's AI labs still need advanced accelerators, but memory supply is becoming its own strategic layer in the compute race. 2

3. Firmus and Nvidia mapped 170,000 GPUs to Indonesia

• Australia's Firmus Technologies said it signed a strategic partnership with Nvidia to provide AI firms with more cost-effective access to compute. 3
• Firmus said the deal will deliver 170,000 GPUs from the first quarter of 2027 to early 2028, with the infrastructure located in Batam, Indonesia. 3
• The company expects up to $30 billion in revenue over six years based on customer commitments, positioning the project as a cloud alternative for smaller AI-native firms. 3

4. WIRobotics opened the ALLEX humanoid simulation model

• WIRobotics launched a technology-disclosure roadmap for its Physical AI ecosystem and released the simulation model for its ALLEX humanoid robot. 4
• The ALLEX model is available in MJCF for MuJoCo, USD for Isaac Sim and URDF for ROS, so researchers can test control, learning and synthetic-data workflows before hardware access. 4
• The company says the physical research platform is planned for later this year; the useful signal is that humanoid robotics is moving toward open simulation assets, not only closed demos. 4

5. UNDP published a playbook for AI in elections

• UNDP released From Promise to Practice: AI in Electoral Administration, a policy document on how election bodies can use AI while protecting democratic safeguards. 5
• The document frames AI as potentially useful for inclusion, integrity and efficiency, but warns that poor deployment can affect human rights, political participation and public trust. 5
• Its baseline recommendation is not to ban or rush adoption, but to pair any election-use case with data governance, transparency, accountability and broad stakeholder engagement. 5