🤝 Anthropic strikes SpaceX data center deal as it plows ahead on AI coding ↗
Anthropic landed a major compute deal with SpaceX, gaining access to Colossus 1 in Memphis. That gives Claude much more capacity, especially for Claude Code users who have been running into demand walls.
The stranger bit? SpaceX says Anthropic is also interested in orbital AI computing capacity. It sounds sci-fi, maybe a little toaster-in-the-clouds, but the compute race is clearly getting more serpentine.
🧠 Anthropic says it taught AI agents how to “dream” ↗
Anthropic introduced a feature called “dreaming” for AI agents. The idea is that agents review their past work between sessions, notice patterns, and improve how they behave next time.
It is launching as a research preview, so not everyone gets access yet. Still, the direction is clear - Anthropic wants agents that feel less like chat windows and more like semi-autonomous co-workers. Helpful, eerie, maybe both.
🔎 Google’s AI search summaries will now quote Reddit ↗
Google is adding more firsthand perspectives to AI Mode and AI Overviews, including Reddit, forums, and social posts. In practice, it is folding the old “add Reddit to your search” hack into AI search itself.
It is also adding more context around links, such as creator names and community names. Nice in theory, although letting AI summarize internet forums still feels like asking a raccoon to organize a pantry.
⚡ Microsoft’s AI data center push is colliding with its clean power goals ↗
Microsoft is reportedly debating whether its aggressive AI data center buildout is making one of its clean energy targets harder to meet. The pressure point is hourly clean energy matching, which is stricter than annual matching.
That is the awkward AI infrastructure story in miniature: everyone wants more models, more GPUs, more capacity - and then the electricity bill walks in wearing steel boots.
💾 AI boom pushes Samsung to $1T ↗
Samsung crossed a $1 trillion valuation as AI chip demand keeps lighting up memory markets. High-bandwidth memory is the main character here, because modern AI systems are ravenous for it.
The company’s shares jumped hard, and the broader message is simple: the AI boom is not just rewarding model labs. It is lifting the whole chip supply chain, unevenly but intensely.
🦾 French startup unveils AI model for robots and human-like hand ↗
Genesis AI unveiled GENE-26.5, a model built to control different types of robots, plus a dexterous robotic hand designed to mimic human movement. The demos included delicate tasks like chopping tomatoes and solving a Rubik’s Cube.
The startup is aiming at industrial work in automotive, electronics, pharma, and logistics. It is also collecting on-the-job worker data through sensor gloves, which is clever - and a bit socially piquant.
🧵 Corning partners with Nvidia to expand US fiber optic output for AI growth ↗
Corning and Nvidia are teaming up to expand U.S. production of optical connectivity products for AI data centers. Chips get the spotlight, but fiber is the quiet plumbing moving data between all those processors.
Corning also raised its long-term sales targets as AI infrastructure demand keeps spilling beyond GPUs. The AI supply chain is turning into a many-headed garden hose, basically.
FAQ
Why are AI companies making more data center deals?
AI companies need massive compute capacity to train models, serve users, and support agent-style products. Deals like Anthropic’s access to SpaceX’s Colossus 1 show just how central infrastructure has become. As demand rises, especially from coding tools like Claude Code, expanded data center capacity can help ease usage bottlenecks and keep services running more smoothly.
What does Anthropic’s “dreaming” feature mean for AI agents?
Anthropic’s “dreaming” feature is described as a way for AI agents to review past work between sessions. The goal is for agents to spot patterns, learn from earlier behavior, and perform better the next time around. Since it is launching as a research preview, it is still early, but it points toward AI agents becoming more persistent, context-aware, and work-oriented.
Why is Google adding Reddit and forum content to AI search?
Google is trying to make AI search summaries include more firsthand perspectives from places like Reddit, forums, and social posts. This mirrors how many people already search by adding “Reddit” to queries when they want lived experience or community answers. The challenge is making those summaries helpful without flattening sprawling internet discussions into oversimplified answers.
How does AI infrastructure affect clean energy goals?
AI infrastructure can require huge amounts of electricity, especially as companies build more data centers for advanced models and cloud services. Microsoft’s reported tension around hourly clean energy matching shows why this is difficult. Matching clean energy hour by hour is stricter than annual matching, so rapid AI growth can make sustainability targets harder to satisfy.
Why is high-bandwidth memory important for the AI boom?
High-bandwidth memory matters because modern AI systems need to move and process large amounts of data at very high speed. Samsung’s rise past a $1 trillion valuation reflects how demand for AI chips and memory can benefit suppliers beyond the model labs themselves. In many AI pipelines, memory performance can be just as critical as raw compute power.
How could AI models change robotics and industrial work?
AI models for robotics aim to help machines perform different physical tasks with greater flexibility. Genesis AI’s GENE-26.5 model and human-like robotic hand are examples of this direction, with demos involving delicate movements like chopping tomatoes and solving a Rubik’s Cube. Industrial uses could include automotive, electronics, pharma, and logistics, especially where repeatable manual work is common.