Particle.news

AI Surge Meets Friction as Oracle Slips Data-Center Timeline and Space Compute Gains Traction

Shortages, high costs, ethical strains steer investment toward standards, robotics, and off‑planet experiments.

Overview

  • Oracle has pushed several OpenAIStargate’ data centers from 2027 to 2028, with sources citing skilled‑labor gaps and materials shortfalls, though the multi‑million‑accelerator buildout scale remains unchanged.
  • China’s CAICT projects the country’s 2025 core AI industry above ¥1.2 trillion and reports a new national standard for high‑performance training and inference as governments elevate embodied intelligence with fresh policies and pilots.
  • OpenAI’s GPT‑5.2 is reported to outperform human experts on roughly 70% of work tasks with longer context handling and a lower hallucination rate, yet reviewers flag steep pricing and weaker image granularity versus rivals.
  • A Data Worker’s Inquiry account relayed by Futurism describes Kenyan contractors role‑playing as chatbots for about $0.05 per message for an Australian firm, highlighting low pay, performance pressure, and emotional toll.
  • Space‑based compute proposals are gathering publicity, with Google targeting a 2027 prototype satellite and a startup claiming an in‑orbit LLM, while critics question radiation resilience and economics despite cheaper launches and free solar power.