
Tech • IA • Crypto
Intel shares jumped nearly 20% on reports of a preliminary manufacturing agreement with Apple. The move follows prolonged negotiations and signals renewed confidence in U.S.-based fabrication. Intel’s integrated design-and-manufacturing model could regain strategic relevance as firms diversify away from Taiwan. The development also reflects intensifying geopolitical pressure to localize advanced chip production.
Washington converted about $9 billion in grants into equity, securing roughly a 10% stake in Intel at around $21 per share. With the stock trading above $120, the position represents a substantial paper gain. Officials helped convene partners including Apple, Nvidia, and SpaceX to stabilize the ecosystem. The intervention underscores industrial policy aimed at semiconductor sovereignty.
A tight cluster of firms—the so-called “AI Big 10”—now accounts for roughly 40% of total market value. Leaders such as Nvidia, AMD, and Broadcom are capturing outsized gains from AI demand. This concentration contrasts with slower growth across the broader economy. The divergence highlights how compute and data advantages are translating into market power.
Anthropic secured access to SpaceX’s Colossus 1 data center, exceeding 300 megawatts and 220,000 NVIDIA GPUs. The capacity is available almost immediately, removing a key bottleneck. The deal directly addresses prior limits on throughput rather than model capability. It positions Anthropic to compete more aggressively on scale with OpenAI and others.
Following the infrastructure boost, Anthropic expanded usage across products. Claude Code rate limits doubled and peak-hour caps were lifted for key tiers. Claude Opus API throughput jumped from hundreds of thousands to millions of tokens per minute. The changes suggest demand had been constrained primarily by compute scarcity.
Elon Musk, despite leading rival xAI, is now providing infrastructure via SpaceX to Anthropic. The partnership reflects shared incentives to monetize idle compute and counterbalance OpenAI. It also illustrates how infrastructure alliances can cut across model competition. The AI race is increasingly defined by access to power and chips rather than ideology.
ChatGPT 5.5 introduces a new architecture using Cerebras processors instead of NVIDIA H100/H200 GPUs. Generation speeds rise from about 65 to up to 1000 tokens per second, enabling near real-time interaction. The system employs post-inference task overlap to run steps concurrently. It marks a shift toward continuous, high-throughput reasoning.
Under EU Digital Services Act pressure, Meta outlined protections for minors on Instagram and related platforms. Users under 13 are banned, while ages 13–17 default to restricted “teen accounts.” Controls include private profiles, limited DMs, and a night mode from 10 p.m. to 7 a.m.. Age estimation relies on automated analysis of behavior and content signals.