It is hard to scale. The taste and texture of ice cream depends a lot on things like fat content, air content, and ice crystal size. Every time you scale up you need to use different equipment to make your ice cream. Getting the settings right on the new equipment to match the product you had on the old equipment is difficult at best. The article points this out when it talks about their attempt to switch to a continuous process.

This seems like a startup opportunity for a group with the chemistry / food science / mechanical engineering chops. Build machinery that can make consistent ice cream at scale in a turn key operation. Sounds incredibly hard and capital intensive.

Not quite so simple.

It's not like mega-ice cream companies like Breyers don't already have an army of chemists, food scientists and mechanical engineers on staff to solve these problems - and yet they still have a tough time producing real ice cream at everyday prices.

The silicon valley wiz-bang startup crowd often forgets... sometimes an industry's incumbents are not lazy and incompetent.

There is machinery that can make consistent ice cream at scale. The problem is that it is very different from the machinery you would use at smaller scales so matching the results of the two processes is very hard.

It's the ice cream with mix-ins that is specifically hard to scale (it doesn't flow well at all). This was Ample Hills' main thing, and they never got the continuous production method to work.

Haagen-Dazs does well at scale, but obviously is quite expensive.