The market is sorting itself into tiers: those building the infrastructure others depend on, those selling access to it, and everyone else scrambling to justify why they need a seat at the table. The fractures that emerged last year are hardening into structural divides.
Nvidia and Intel are coalescing around agentic AI as the next revenue engine, with Nvidia's DGX Rubin NVL8 running on Intel Xeon 6 processors and SK Hynix's chairman declaring a wafer shortage that will persist through 2030. That scarcity is real and priced in. Meanwhile, OpenAI is expanding its government footprint through an AWS partnership for classified and unclassified work while the Pentagon simultaneously develops alternatives to Anthropic and discusses secure environments for companies to train models on classified military data. The government is hedging, which means no single vendor owns the relationship. Anthropic, for its part, is in active legal conflict with the Justice Department over warfighting systems, with the government arguing it lawfully penalized the company for trying to limit how Claude could be used militarily. That's not a regulatory disagreement. That's the government asserting it will decide how AI gets deployed regardless of vendor preference.
Enterprise adoption is splintering by capability and control. Mistral is betting that enterprises want to train custom models from scratch on proprietary data rather than fine-tune or rely on retrieval. Google is expanding Personal Intelligence to all US users, letting its AI tap Gmail and Google Photos for more tailored responses. Microsoft unified its consumer and commercial Copilot systems under Jacob Andreou while narrowing Mustafa Suleyman's role to superintelligence and frontier models. These aren't cosmetic reorganizations. They're admissions that the integrated AI assistant play is fragmenting into specialized layers. Meanwhile, Yann LeCun left Meta, raised $1.03 billion in four months at a $3.5 billion valuation for AMI Labs in Paris, explicitly betting against large language models. That's not noise. That's one of AI's founding figures placing capital on a different architecture entirely, with backing from Nvidia, Samsung, and Bezos Expeditions.
Sloane Duvall