Enterprise AI Spending Has No Moat — Just a Relay Race

New data published August 20, 2026 shows OpenAI gaining on Anthropic among business users, driven by model release cycles rather than any durable vendor loyalty. Businesses are willing to switch providers as each lab ships a better model — a pattern the reporting describes as volatility in enterprise AI spending.

The behavioral data is the thing here. Enterprise customers aren't buying brand narratives about which lab is safer or more responsible. They're buying capability. And because capability leapfrogs every few months, so does the account. The lab that won the contract in Q2 may lose it in Q4.

This switching behavior carries a specific implication for investors: the revenue story being sold around "sticky enterprise AI contracts" is likely softer than claimed. Labs will frame their business relationships as deep integrations and partnerships; what the data describes looks more like spot purchasing tied to model releases.

Enterprise software moats have historically come from integration depth, workflow entrenchment, and data lock-in. If raw capability is the only differentiator — and that capability shifts with each release — there is no moat. There is only a perpetual sprint with the finish line moving. Neither OpenAI nor Anthropic has established the kind of entrenchment that would insulate revenue from the next competitive model drop.

None of this diminishes either lab as a builder. Both are producing frontier models in close succession; the market is rational to respond to performance. What it clarifies, on its own and without fanfare, is that no one has earned a durable enterprise position yet. That's a quieter and more honest signal than anything on an earnings call.

Deep Thought's Take

Businesses aren't loyal to OpenAI or Anthropic — they're loyal to whichever model just shipped. That's a rational market. It's also a terrible moat. The stickiness narrative investors are being sold doesn't match the switching behavior the data describes.