10 gig requires a remarkable 160 MHz+ but it makes a great headline. With this 8x8 chipset, realistic speed will head to a gigabit and higher, in a room with just the right walls for the MIMO to work. Update Sept 15. Quantenna writes me they expect production quantities in 3-4 months.
Speeds through a wall go down severely with any WiFi, but throughput even through walls should be excellent. MU-MIMO - dedicating the 8-12 streams 1, 2 or 4 to an individual device with beamforming - should give each of the devices excellent throughput. Quantenna has distributed samples but production quantities are in the future.
If you have no neighbors and completely uncontested WiFi frequencies, you will sometimes be able to measure a nominal 9 or 10 gigabit WiFi from Quantenna and Freescale's new chips. You'll need an ideal location, with just enough interference to bounce the 12 separate signals but not enough to significantly slow anything down. 160 MHz is about the entire bandwidth of the combined AT&T and Verizon LTE networks, far more than will be available to most of us in most places. So the many MU-MIMO connections will have more practical impact for most of us. In addition, they are squeezing a heck of a lot of data in, going to a 1024 QAM constellation, needing extraordinary analog circuitry and minimal interference.
Twenty years ago, Arogyaswami Paulraj discovered using more antennas could multiply the throughput of a wireless network. That day it rained on the Stanford campus and he moved his multi-transmitter experiments indoors. Because the transmitters would be close together, he expected poor results. To his surprise, the separate signals came in clearly. He had invented MIMO.