Intelligence Just Got Cheaper. Proving What You Do With It Just Got Harder.

On July 16, a Beijing-based lab called Moonshot released Kimi K3, a 2.8-trillion-parameter open-weight model priced at a fraction of frontier rates. The weights land July 27. After that, anyone with the hardware can run it, tune it, and own it outright.

Two Optura leaders read the moment from different angles.

Andy Fanning, Optura’s CEO, writes about the economics underneath the headline. Renting intelligence by the token made sense when AI was a novelty line item; it doesn’t anymore. Andy makes the case for owning the model outright. Hosted on your infrastructure, trained on your data. A model everyone else can also rent, he argues, can never be your competitive edge. Read Andy’s take.

Mike Hollis, Optura’s President, explains what it means for the executives responsible for patient and member outcomes. When intelligence gets cheaper and far more available, no one owns an advantage in the model itself. What’s left to compete on is who deploys it responsibly, and who can prove what it actually did. Read Mike’s take.

Both arrive at the same place from different directions. Cheaper, more available intelligence doesn’t hand anyone an advantage by itself. It raises the bar on who can prove what they built with it worked. That’s the whole premise behind ROAI™(Return on AI Investment). 

AI is Potential. ROAI is Performance.