The artificial intelligence market just fractured. Meta's release of Llama 3.1—a 405-billion parameter open-source model matching OpenAI and Anthropic's benchmark results—fundamentally changes the economics of AI deployment. While OpenAI charges per API token and Claude charges by subscription, Llama 3.1 costs only computational power. This shift will likely redirect investment away from proprietary models toward companies with in-house compute capacity or those willing to run open infrastructure. Meta isn't naive about the margins. The company is embedding Llama directly into Instagram, WhatsApp, and Facebook as Meta AI—a distribution advantage OpenAI can only dream of. Three billion daily active users now have access to an AI assistant without downloading a separate app. The assistant also ships with agentic capabilities: it searches the web, calls APIs, and chains actions together without programmer intervention. This isn't just chat; it's orchestration. The company loses money on each inference, but gains engagement hours and higher ad revenue per user. Training Llama 3.1 cost hundreds of millions in GPU rental alone—sustainable only because Meta's core business generates $39 billion in quarterly revenue with 73% profit growth. That buffer lets Zuckerberg bet on AI without investor pressure. The next move is Llama 4, which Meta claims will require ten times more compute. This is an arms race now, and NVIDIA is the only real winner.
Meanwhile, the U.S. Department of Justice declared Google a monopolist. The search engine holds 90% market share—up from 80% in just over a decade—backed by exclusionary contracts worth billions annually. Apple receives roughly $20 billion per year to make Google the default search engine in Safari, with an additional 36% revenue share on ads flowing through iOS. Microsoft faced similar antitrust action in 2000; Google now faces potential divestitures (YouTube is vulnerable) or forced breakup of search revenue-sharing deals. The immediate beneficiaries are Meta and OpenAI, both exploring search products. The deeper question: can any competitor gain traction when Apple's installed base and Google's cash pile can simply outbid new entrants? The ruling is sound policy but execution matters enormously. Forced divestiture of YouTube would reshape media—but regulators rarely follow through completely.
The broader tech volatility this week stemmed from Japan's first interest rate rise in 17 years, triggering a cascade of automated selling across global equities. The Nasdaq fell 15%, the "Magnificent Seven" fell 20%, and the VIX spiked to pandemic-era levels. Weak U.S. jobs data (114,000 new positions, versus 179,000 prior month) spooked markets into fearing a hard landing rather than the "soft landing" consensus. The unemployment rate hit 4.3%—the highest since 2020, though still far below pandemic or 2008 crisis levels. Rate cuts from the Federal Reserve are now priced in at 90.5% probability by September. The market's fear isn't irrational: employment usually lags economic deterioration, meaning the damage might already be baking in.