⛵ Anthropic Rejects Open Ban
Good Morning, Early Adopters!
The AI buildout now runs on Nvidia's balance sheet, and credit markets just started charging for it.
Amodei Denies Pushing Open-Weight Ban
TL;DR: After days of pressure from the open-source camp, Anthropic CEO Dario Amodei publicly rejected claims the company wants open-weight models banned. He says Anthropic only supports capability thresholds for frontier systems, not a blanket prohibition on releasing weights. Read more →
Nvidia Backstops $250B OpenAI Loan
TL;DR: Nvidia is in talks to guarantee up to $250 billion in debt so OpenAI can finance a planned 10GW data center campus in Ohio. The arrangement would deepen an already circular relationship between chip supplier and largest customer, and pushes AI infrastructure leverage into territory Wall Street has never underwritten before. Read more →
Nvidia, Microsoft Launch Open Secure AI Alliance
TL;DR: Nvidia, Microsoft, SpaceX and other industry leaders formed the Open Secure AI Alliance to audit and harden open-weight models and their supply chains. Notably absent from the founding roster: OpenAI, Google and Anthropic, the three labs currently arguing hardest for restrictions on open model releases. Read more →
Sutskever's SSI Partners With Nvidia
TL;DR: Safe Superintelligence will work with Nvidia to scale its research compute, the clearest infrastructure signal yet from a lab that has shipped no product. For a company valued on Ilya Sutskever's reputation alone, tying itself to Nvidia's roadmap is a concrete commitment. Read more →
Moonshot Open-Sources Kimi K3 Weights
TL;DR: Moonshot AI released Kimi K3 open weights alongside its serving infrastructure, letting teams self-host a frontier-class model without license fees. The release lands squarely on closed-lab pricing assumptions and reframes the open-versus-closed fight now consuming Silicon Valley. Read more →
Enigma Raises $71M For Robot Models
TL;DR: Enigma Ltd. closed $71 million to build foundation models for robots, targeting general-purpose manipulation across different hardware form factors. The round adds to a crowded embodied-AI funding cycle where investors are betting cross-embodiment models beat single-robot software stacks. Read more →
Microsoft Ships Cheap Cybersecurity Model
TL;DR: Microsoft launched MAI-Cyber-1-Flash, a low-cost model for security analysis, while admitting the hardest tasks still route to OpenAI models. The candor underlines how far Microsoft's in-house stack remains from replacing its partner on frontier workloads. Read more →
Amazon Plans Satellite Phone Network
TL;DR: Amazon asked the FCC to approve 5,105 satellites for a direct-to-device network launching in 2028, letting ordinary phones connect without ground towers. It puts Amazon Leo head-to-head with Starlink's direct-to-cell service and reopens the orbital spectrum fight. Read more →
Samsung And Broadcom Forge Chip Alliance
TL;DR: Samsung and Broadcom will collaborate across memory, foundry and advanced packaging, combining HBM supply with custom ASIC design. The pairing gives hyperscalers a credible path to build their own accelerators instead of queueing for Nvidia allocation. Read more →
AMD Splits Zen 7 Into Three Families
TL;DR: AMD will fracture Zen 7 into three distinct EPYC families for 2028 and begin pricing server CPUs by agent capacity rather than core count. It is the first major vendor to formalize per-agent economics in data center silicon. Read more →
Two Drones Escape FCC Foreign Ban
TL;DR: Two more foreign drone models won exemptions from the FCC's equipment ban, widening a loophole in Washington's push to strip Chinese hardware out of American drone fleets. Domestic manufacturers get a longer runway than the rules implied. Read more →
Data Center Opponent Arrested For Clapping
TL;DR: A resident opposing a proposed data center was arrested after clapping at a city council meeting, a small incident that captures how sharply local politics have turned against AI infrastructure buildouts and the permitting fights ahead. Read more →
Kimi K3 Runs On 80 Consumer GPUs
TL;DR: An infrastructure team got the full 2.8-trillion-parameter Kimi K3 serving at 20 tokens per second on 80 RTX 5090s, untuned, on day one. Frontier-class intelligence is now self-hostable on consumer silicon with zero licensing cost. Read more →
READ MORE
Let the Future Come to Your Inbox
Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.
- AI Secret uncovers what really matters in AI
- Bay Area Letters decodes tech and business shifts from Silicon Valley
- Robotics Herald tracks how robots move from labs into daily life
- Marketing Secret breaks down real growth and go-to-market playbooks
- The Hardwire explores hardware, consumer tech, and what’s coming next
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
