I suspect that the lack of ability to form nulls in the beam is as big or even a bigger limitation than the reduction in gain when going from a big array to a phone.

The SNR in Shannon’s Law has a log in front of it, but spectrum reuse is more or less linear. If there are five visible satellites and I can null out four of them, then I can receive from and transmit to the fifth without substantial interference. (I’m not saying this is easy! Contemplate how many WiFi generations have had MIMO and how limited it still is.)

So I believe that it’s comparatively straightforward to demonstrate a shiny new direct-to-cell system with a single phone on a stage, but achieving usefully large aggregate bandwidth in a dense area will be more challenging.

FWIW the problem with Iridium, historically anyway, was that available bandwidth was very low, so they had to charge a silly amount for usage of that bandwidth, so very few people used it. Iridium used low-ish frequencies, with narrow bandwidth, and (I think) no MIMO whatsoever, not even polarization diversity.