The wireless industry is notorious for snake-oil salesmen. When you actually claim that your technology violates Shannon channel capacity, it doesn't inspire much confidence that you understand information theory.
From the description (this is just a guess), it appears that DIDO offloads channel characterization to a centralized server to reduce the computational requirements of the clients (similar to AGPS). MIMO techniques (like DIDO) do not violate Shannon capacity. They improve Eb/N0 by increasing gain through beam-forming and physical antenna diversity. In addition, DIDO appears to be using some form of asynchronous spread spectrum (even though they deny using CDMA, there are many other flavors), which allows all clients to transmit simultaneously by using orthogonal codes. Spread spectrum techniques also improve gain at the expense of bandwidth.
While they may have a good implementation, I don't see anything earth shattering here.