For most of the past two years, the story of enterprise AI had a clear protagonist. Anthropic’s Claude models earned a reputation among developers and corporate buyers as the serious, dependable choice, and the company rode that reputation into some of the largest deals in the sector. New data suggests that story is getting more complicated. OpenAI, according to figures reported by TechCrunch, is gaining on Anthropic among business users, and the shift says something uncomfortable about how loyal enterprise customers actually are.
The headline number is a direction, not a coronation. OpenAI is closing a gap, not erasing it. But the more revealing detail is why the gap is moving at all. Businesses, it turns out, are perfectly willing to switch allegiances every time one lab ships a better model. That is not the behavior of a captured market. It is the behavior of buyers keeping the receipt.
The Switch Nobody Wanted to Admit
Enterprise software is supposed to be sticky. That is the entire investment thesis behind selling to companies rather than consumers. Once a business wires a vendor into its workflows, retrains its staff, and builds internal tooling around a product, the cost of leaving is meant to be high enough that customers stay put even when a rival looks tempting. Salesforce built an empire on that friction. So did Oracle, SAP, and a dozen others.
AI models are behaving differently. When a lab releases a new flagship that tops the benchmarks or handles a particular task better, business users appear ready to route their traffic elsewhere with far less hesitation than the old software playbook would predict. Part of that is technical. Swapping one model’s API for another is closer to changing a setting than ripping out a platform. Part of it is competitive pressure. If your competitor’s support bot or coding assistant got smarter overnight because they switched providers, sitting still starts to look like a choice you have to defend.
That fluidity is exactly what the new data captures. OpenAI gaining on Anthropic is not necessarily a verdict on which company builds the better model this quarter. It is evidence that the quarter is the unit of measurement, and that leads can reverse.
Why Investors Should Feel a Chill
Both companies have raised money at valuations that assume enormous, durable revenue. Those valuations rest on an unspoken premise: that once businesses commit to a model provider, they stay, and their spending compounds year over year like any other enterprise subscription. The volatility now showing up in the data pokes a hole in that premise.
If customers flop back and forth with each model release, then market share becomes a snapshot rather than a moat. A commanding lead can shrink between funding rounds. Revenue that looks recurring might really be a series of short-term bets that get re-placed every time the leaderboard shuffles. For investors underwriting OpenAI and Anthropic alike, that is a reason to scrutinize how much of the reported enterprise spending is genuinely locked in versus how much is up for grabs the moment a competitor ships something sharper.
There is a version of this that cuts the other way, and it is worth being honest about it. The same churn that threatens Anthropic’s lead is what let OpenAI gain in the first place. Low switching costs are a weapon for whoever ships the best model next. The trouble is that no lab can count on always being that company, which means neither can bank on keeping the customers a good model wins.
What the Volatility Actually Means
The comforting narrative for the AI labs was that enterprise adoption would harden into something permanent, the way corporate software usually does. The data complicates that. What is emerging instead looks less like a settled duopoly and more like a running contest where the scoreboard resets with every major release, and where being ahead in August guarantees very little about November.
That dynamic favors customers, who get to extract better models and better pricing by staying mobile. It pressures the labs into a relentless release cadence, because standing still hands rivals an opening. And it complicates the tidy financial models that treat AI revenue as annuity-like income streaming in from captive accounts.
The interesting question now is whether either company can build the kind of lock-in that consumer switching costs never provided, through deeper integrations, proprietary tooling, or data advantages that make leaving genuinely painful. Until one of them manages that, enterprise leadership in AI may keep changing hands, and the next data drop could just as easily show Anthropic pulling back in front. Watch the release calendar, because right now that is where the market share is being decided.
For more coverage of enterprise AI competition, visit Mylistingo.
Source: Original Article
The Netherlands, for internationals
One email a week: the week's news for internationals in the Netherlands, new practical guides and what changed in the cities. No spam, unsubscribe any time.
