
Tech • IA • Crypto
OpenAI introduced GPT-5.6 models—Soul, Terra, and Luna—with aggressive pricing from $1–$5 per million input tokens and improved efficiency. The company claims up to 54% better token usage in coding tasks, signaling a push toward cost leadership. The release emphasizes cybersecurity applications, including threat modeling and defensive testing. The launch intensifies pressure on rivals like Anthropic in enterprise AI.
ChatGPT Work consolidates chat, coding, file handling, and autonomous browsing into a single desktop environment. The platform introduces multi-agent task execution, enabling longer and more complex workflows. Early adoption has been uneven, with reports of bugs, quota resets, and feature inconsistencies. The move reflects a broader shift toward agent-based productivity systems.
Soul Ultra reportedly generated a proof for the cycle double cover conjecture, a decades-old graph theory problem. The result was produced in under an hour using parallel agent systems. Early expert reactions describe the proof as promising but unverified, underscoring the need for peer review. The episode highlights both the potential and uncertainty of AI-driven scientific discovery.
Major players including China, Google, Amazon, Nvidia, and SpaceX are advancing orbital computing infrastructure. China has already deployed 12 computing satellites, while Blue Origin seeks approval for over 50,000 units. The trend reflects growing urgency to expand compute capacity beyond Earth. It marks a new frontier in the infrastructure race underpinning AI growth.
AI data centers could consume 950 TWh annually by 2030, about 3% of global electricity. The issue is not just volume but local grid saturation, with single facilities matching 2 million households. Grid expansion can take 4–10 years, slowing deployment. These constraints are pushing companies to explore space-based alternatives despite high costs.
Orbital data centers face severe engineering challenges, particularly heat dissipation. A 1 MW system may require around 2,500 m² of radiators, scaling dramatically at higher capacities. Unlike Earth, space limits cooling to radiative methods only. Additional risks include cosmic radiation damage and hardware degradation, complicating long-term viability.
Scality highlights structural weaknesses in European tech sovereignty, despite serving governments in 35 countries and major banks. Public contracts often flow through integrators like Capgemini or Atos, masking reliance on non-European technology. Unlike the U.S. Small Business Act, Europe lacks coordinated industrial policy. This gap risks weakening regional control over critical infrastructure.
Enterprises are reassessing hyperscaler dependence as cost pressures rise. Scality claims its solutions are about 30% cheaper than negotiated cloud pricing. Large organizations are increasingly shifting workloads back to on-premise or hybrid models. The trend signals a broader recalibration of cloud economics and control.