The Information Machine

2026-05-18

Anthropic overtakes OpenAI in U.S. enterprise market share on the same day both labs make competing infrastructure moves, while documented AI offensive cyber capability reaches new operational thresholds and data on AI-linked layoffs undermines the workforce-replacement thesis.

What

The enterprise AI market flipped: Anthropic now leads U.S. business AI adoption at 34.4% vs. OpenAI's 32.3%, driven by Claude Code reaching $2.5B in annualized revenue and Anthropic's total annualized revenue hitting $30B against OpenAI's roughly $24B [1]. Both labs responded on the same day with infrastructure bets — Anthropic acquired Stainless, the SDK and MCP tooling company [2], while OpenAI announced a hybrid/on-premises Codex partnership with Dell backed by a new $4B Deployment Company with 19 enterprise partners [3], signaling that enterprise lock-in is now the central strategic contest. On the security front, the evidence for AI's autonomous offensive reach widened: a criminal actor used AI to discover and exploit a zero-day 2FA vulnerability via a hardcoded trust assumption that traditional tools would likely have missed, and the UK AISI documents that frontier models' autonomous cyber 'time horizon' has doubled in months, with one model completing a 32-step simulated corporate network attack 60% of the time [4]. A historical parallel to the pre-Stuxnet fast16.sys virus — which silently corrupted floating-point results rather than causing obvious failures — frames how AI-enabled attacks may increasingly favor subtle, hard-to-detect degradation over overt disruption [5]. Meanwhile, empirical data on AI-driven workforce cuts cuts against the dominant corporate narrative: 56% of S&P 500 companies announcing AI-linked layoffs saw stock declines averaging 25%, and a Gartner survey of 350 large-company executives found no return advantage for companies that cut headcount versus those that did not [6].

Why it matters

The competitive inversion — with Anthropic now leading OpenAI in enterprise adoption and both companies racing to own the infrastructure layer on the same day — means enterprise AI infrastructure lock-in is becoming the defining strategic battle faster than most observers anticipated. At the same time, the accumulation of evidence that AI can autonomously execute sophisticated offensive cyber operations, now accessible to criminal actors not just nation-states, means the security calculus for enterprise AI deployment is shifting beneath the governance frameworks meant to manage it.

Open questions

  • Anthropic acquired Stainless (SDK/MCP tooling) [2] and OpenAI partnered with Dell for on-premises Codex [3] on the same day — which enterprise lock-in strategy proves more durable: owning the developer toolchain or owning the on-premises deployment layer?

  • The UK AISI documents frontier models completing 32-step corporate network attacks 60% of the time, with the autonomous cyber 'time horizon' doubling in months [4] — at what demonstrated capability threshold does voluntary AI safety governance become structurally inadequate, and who has standing to force a harder regulatory response?

  • A criminal actor used AI to find a zero-day 2FA vulnerability via a hardcoded trust assumption [4], suggesting AI-assisted zero-day discovery is now accessible below nation-state level — how quickly can enterprise security tooling adapt when the attacker's marginal cost of vulnerability research is falling faster than the defender's?

  • Gartner data shows AI-linked layoffs produce no better returns than not cutting headcount, while the highest-ROI companies use AI for worker amplification rather than replacement [6] — will enterprise buyers apply this evidence when evaluating vendor efficiency claims, or will the 'AI-washing' dynamic Sam Altman acknowledged continue to dominate corporate narratives?

Thread movements (2)

  • enterprise-ai-coding-battle — Anthropic surpassed OpenAI in U.S. business AI adoption for the first time (34.4% vs. 32.3%), with Claude Code at $2.5B annualized revenue and total Anthropic revenue of $30B vs. OpenAI's ~$24B [1]; on the same day, Anthropic acquired SDK/MCP tooling company Stainless [2] while OpenAI announced a hybrid/on-premises Codex deal with Dell and launched a $4B Deployment Company [3].
  • ai-offensive-cyber — A criminal actor used AI to find and exploit a zero-day 2FA vulnerability via a hardcoded trust assumption, and the UK AISI reports frontier models' autonomous cyber 'time horizon' has doubled in months with one model completing a 32-step corporate network attack 60% of the time [4]; a Stuxnet-era historical parallel (fast16.sys silently corrupting floating-point results) frames how AI-enabled attacks may prioritize undetectable degradation over overt disruption [5].

Notable items (1)

  • 😼 AI layoffs are tanking stocks, not saving them
    The Neuron
    Concrete data undercuts the AI-as-workforce-replacement thesis: 56% of S&P 500 companies announcing AI-linked layoffs saw stock declines averaging 25%, a Gartner survey of 350 executives found no ROI advantage for companies that cut headcount for AI reasons, and the highest-returning companies used AI for worker amplification — not replacement [6].