Plumbing the
information age

On July 11, Apple filed suit in Northern California federal court, alleging that OpenAI leadership directed efforts to recruit Apple's silicon and systems engineering teams as part of a coordinated campaign to extract confidential information about unreleased products. The 400+ figure is significant—that represents an entire institutional division's knowledge base walking out the door over roughly two years, not casual attrition. The complaint emerges weeks before OpenAI's confidential IPO filing with Goldman Sachs and Morgan Stanley, targeting a September 2026 debut at roughly $730 billion valuation. For the broader industry: if Apple wins early procedural motions or forces broad discovery into OpenAI's recruiting practices, every frontier lab that has been aggressively hiring Big Tech engineers will face immediate pressure to slow down and document its recruiting far more carefully. This is the first major lawsuit of the AI talent-war era. The timing is particularly damaging for OpenAI given Anthropic's reported revenue advantage ($47 billion annualized vs. OpenAI's projected $25-33 billion for 2026)—public filings will expose these dynamics directly.

Fortune's reporting reveals a fundamental market shift: Anthropic is now the higher-revenue frontier lab despite operating a smaller model family and launching later than OpenAI. Claude Code alone went from $1 billion in annualized revenue at the end of 2025 to more than $2.5 billion by February 2026, and the company leads on developer market share and agentic coding reliability. This gap directly challenges the "scale at all costs" narrative that has dominated AI investment strategy. The revenue disparity matters immediately: OpenAI's S-1 filing will be audited, side-by-side with Anthropic's financials, during the most scrutinized IPO moment in tech history. It signals that enterprise willingness to pay depends less on raw model capability and more on reliability, governance, and domain-specific performance. For procurement teams: this validates the bet on Anthropic's Claude for enterprise workloads and raises questions about what OpenAI must demonstrate to justify its $730 billion IPO valuation against an incumbent that monetizes better per user.

This development marks a critical inflection point: even at hyperscaler scale, compute capacity is the binding constraint. Meta requested additional access to Google's Gemini models and was denied due to insufficient capacity at Google itself—a first public signal that the infrastructure arms race has created genuine scarcity. The consequences ripple internally: Meta delayed some AI projects and explicitly instructed teams to optimize for token efficiency, a classic sign of cap-driven rationing. For enterprise buyers: this validates the urgency around inference optimization and multi-model strategies rather than single-vendor dependency. For infrastructure strategy: this confirms that the next 18 months will see continued vendor lock-in pressures and potential capacity price escalation. Meta's response is equally significant—the company is accelerating its Alberta data center build (1 gigawatt), bringing custom chips into production (Iris, September 2026), and exploring external compute sales to offset internal constraints.

Perplexity's $500 million raise at $14 billion valuation signals investor confidence in the AI search category, but more importantly, demonstrates market readiness for a non-Google search experience. The company has executed disciplined product strategy—focusing on research and fact-finding use cases where citation and reasoning matter more than coverage breadth. For practitioners: Perplexity's traction validates the underlying market thesis that semantic search + reasoning is a viable wedge. For enterprise integration: the funding enables infrastructure expansion, which directly affects API latency and availability for customers embedding Perplexity. The valuation places Perplexity above many 2025-era "unicorn" AI startups, reflecting the market's repricing of AI search as a core infrastructure play rather than a niche tool.