AINewsnow

OpenAI Courts the Grid and the Pentagon While Anthropic's $900B Bet Stuns

2026-11-03

OpenAI Courts the Grid and the Pentagon While Anthropic's $900B Bet Stuns

OpenAI dominated Monday's news with a grid-friendly power pledge and a national security hire, while Anthropic's IPO filing revealed staggering revenue and even more staggering spending commitments. Meanwhile, a supply-chain scare hit Claude Code and Marvell inked the AI era's latest mega-deal.

  1. OpenAI pledges to power down when the grid strains OpenAI launched the Grid Protectors Program, committing future US compute campuses to reduce electricity demand during grid strain. A pilot at its planned Effingham County, Georgia campus will free up to 1,000 megawatts.
    Why it matters: Data centers' surging power appetite is becoming a political and infrastructure bottleneck, so voluntary curtailment could shape how AI expansion coexists with fragile grids.
  2. OpenAI hires ex-Army Secretary Dan Driscoll for national security role Dan Driscoll, former US Secretary of the Army, joins OpenAI as Senior Counselor for National Security Modernization, a newly created role. He will advise on responsible AI for national security, cyber resilience, and government collaboration.
    Why it matters: The hire signals OpenAI is deepening its Washington posture as AI becomes central to defense and cyber policy.
  3. Backdoored Claude Code extension slipped into Cursor's marketplace Malicious versions 2.1.312 and 2.1.313 of the Claude Code VS Code extension carried a backdoor and sat in Cursor's extension marketplace for roughly eleven hours before removal. A security advisory followed the takedown.
    Why it matters: Developer toolchains are now prime supply-chain attack targets, and even trusted AI coding ecosystems can be compromised in hours.
  4. Anthropic's S-1 shows $17B quarter and a $900B bill Anthropic's updated S-1 filing ahead of its IPO shows quarterly revenue surged more than 1,100% to about $17 billion, though the company still posted a net loss. It also disclosed nearly $900 billion in spending commitments.
    Why it matters: The filing is the clearest look yet at the AI sector's economics: explosive growth paired with commitments that dwarf current revenue.
  5. Anthropic's $900B commitments are a bet on the boom continuing Anthropic has taken on nearly $900 billion in spending commitments, reflecting aggressive infrastructure and compute investments. The figure represents a massive wager that AI demand keeps climbing.
    Why it matters: If the boom cools, commitments of this scale could become the industry's biggest stress test since the dot-com buildout.
  6. Marvell to buy Lumentum for $115B in AI optics play Marvell Technology agreed to acquire optical components maker Lumentum in an all-stock deal valued at $115 billion including debt, announced November 2, 2026. The deal targets AI data-center connectivity.
    Why it matters: Optical interconnects are a critical bottleneck for AI clusters, and consolidation here shows how far chipmakers will go to own the data-center stack.
  7. DeepSeek open-sources proactive coding agent dsh-superagent DeepSeek's Harness team open-sourced dsh-superagent on October 31, a research prototype coding agent that takes requests via GitHub and optionally Feishu. It works in separate sessions and returns results proactively.
    Why it matters: Proactive, asynchronous agents point beyond chat-based coding assistants, and open-sourcing one lowers the bar for developers to experiment with the pattern.
  8. Musk admits Grok Bot limits cut over 'unsustainably high' costs Elon Musk confirmed Grok Bot's usage limits were intentionally throttled, citing serving costs that were 'unsustainably high relative to revenue.' The admission came after subscribers complained their weekly caps ran out early.
    Why it matters: Even well-funded AI products are hitting the wall on inference economics, a sign that consumer AI pricing may have to rise.
  9. 42AI exits stealth with $62.5M to model human behavior Tel Aviv-based 42AI emerged from stealth with $62.5 million in seed funding to build a behavioral world model that predicts how humans will act. Brokerages are its first customers.
    Why it matters: Predicting human behavior is a lucrative and sensitive frontier for world models, and finance is a natural first proving ground.
  10. October roundup: local LLMs get bigger and more capable A community roundup of October 2026 highlights local-capable LLMs like Ornith-1.5-122B for million-token agent tasks on a single GPU and Raon-1-mini-preview for Korean agentic AI on an RTX 5090. Google's monthly AI announcements also featured prominently.
    Why it matters: Frontier-adjacent capabilities running on consumer hardware signal how quickly the local AI ecosystem is closing the gap with cloud models.