On 4 May 2026, OpenAI and Anthropic announced, on the same day, that they were entering enterprise AI adoption — one forming a deployment company valued in the tens of billions, one a joint venture with large funds totalling US$1.5B. Two model companies said the same thing with real money: enterprise demand for adoption has outgrown what any single delivery model can carry.
What's notable isn't the amount — it's the method. The delivery model they chose is the forward-deployed engineer (FDE): engineers embedded directly in the client's workflow — not selling licences, not running courses, but getting things working on the ground. That is the model companies admitting: for AI to actually enter an enterprise, models and tools aren't enough — you need someone on site.
But that model has a boundary. An FDE costs upwards of US$300K a year — against typical Taiwanese engagement values, the economics don't balance. An FDE needs to go deep into the client's workflow, and Taiwanese process documents are in Chinese, with the real decision chain invisible on the org chart — a flown-in engineer can't read it and can't hold it. Taiwan's mid-sized enterprises are highly dispersed; there's no PE-fund portfolio of two thousand companies to package at once.
So outside the US, model companies run on the other leg: certifying local partners. The methodology and certification come from the vendor; the people on the ground who know the local industry do the delivery. Taiwan is on that track.
Whoever does AI adoption for Taiwanese enterprises won't be an engineer shipped from Silicon Valley — it'll be a local team that's vendor-certified and knows your industry. That gives companies two questions to ask directly: where does your certification come from? Where does your industry experience come from? Not many can answer both — yet.
(Draft. The published version will add source links and fact-checked figures; the interpretation here is Confluence's view.)