The real story today isn't about AI's capabilities, it's about the collision between the infrastructure being built and the actual demand for what it produces. Hyperscalers are deploying $725 billion in AI compute this year while consumers actively reject the output, Wikipedia bans AI-generated content by a 44-2 vote, and Google's AI Overviews have tanked click-through rates by 58 percent. This is structural overcapacity meeting consumer resistance, and it's reshaping where money flows and who wins.
The demand problem manifests differently by market segment. Anthropic now serves more business customers than OpenAI according to Ramp expense data, 34.4 percent of participating businesses pay for Anthropic services versus 32.3 percent for OpenAI, a reversal that signals model quality and product fit matter more than first-mover advantage. Clio hit $500 million in annual recurring revenue as legal tech startups see massive adoption, while Anthropic courts small business owners and Notion turns its workspace into a hub for AI agents. The Fortune 500 was always the easier sale. The real expansion is downmarket into the 36 million small businesses that make up the U.S. economic backbone, and into vertical software where AI becomes embedded rather than bolted on. Meanwhile, enterprises discovering that 97 percent of organizations have active AI initiatives but only 5 percent say their data is ready reveals the gap between infrastructure spending and actual execution capability. The bottleneck isn't compute. It's data governance, and that's where leverage is shifting.
The courtroom drama around Sam Altman and the Musk v. Altman trial matters less than what it exposes about OpenAI's governance and incentive structure. Altman testified he is "an honest and trustworthy businessperson" while facing claims he's a "prolific liar," and the trial is forcing confrontation with facts about how OpenAI lost control of its own direction. SoftBank booked a $25 billion gain on its OpenAI stake in the fourth quarter, meaning the financial incentives now run through late-stage investors and their liquidity timelines, not through the original nonprofit mission. This matters because it explains why the company is racing to monetize every surface, WhatsApp now has Meta AI chats in incognito mode, Amazon launched an AI shopping assistant for the search bar powered by Alexa+, and Origin Lab is building a marketplace where AI labs buy licensed video game data from game companies. The infrastructure is real. The business model is still being written, and whoever figures out how to extract value from the compute without triggering consumer backlash wins the next phase.
Sloane Duvall