The big money is flowing toward AI that gets embedded into existing workflows rather than standalone products. Google is pushing Gemini deeper into Workspace, Amazon is integrating health AI into its retail platform, and Zoom is layering agents into its suite to automate enterprise tasks. Separately, the market is learning that early adoption doesn't equal durability. RevenueCat's report shows AI-powered apps drive strong monetization at launch but fail to retain users over time. The pattern is clear: generalist AI features bolt-on to platforms with existing distribution win. Standalone AI applications struggle to stick.
The infrastructure layer is consolidating around a few proven players, and the incentives are moving toward vertical integration. Thinking Machines Lab just locked in a gigawatt of Nvidia compute through a multiyear deal that includes strategic investment from Nvidia itself. Legora reached 5.55 billion dollars on the back of legal AI, which works because it solves a specific pain in a high-margin industry. Meanwhile, open source licensing is becoming a battleground. The question of whether LLM-generated code can legally strip a GPL license is unresolved, and it matters. If models can rewrite licensed code into unlicensed variants, the economics of open source shift dramatically in favor of whoever controls the model.
Operational reality is catching up with deployment velocity. Amazon has mandated that senior engineers sign off on AI-assisted code changes after multiple outages linked to AI tooling. Claude Opus found 22 vulnerabilities in Firefox in two weeks, which is impressive for security research but also a reminder that AI agents can find problems humans miss, and miss problems humans catch. The Trump administration is preparing executive action against Anthropic even as Microsoft backs the company's legal challenge to the Pentagon's supply chain risk designation. The regulatory pressure is real, but so is the business incentive to deploy these tools. The tension between caution and speed is not resolving in favor of caution.
Sloane Duvall