5G is high frequency millimeter wave.

Millimeter wave is an important part of 5G but at least for the next few years many antenna MIMO is likely to dominate. MIMO doesn't require new cell sites. Shamsunder of Verizon. 4G carrier aggregation will get us past the gigabit and few telcos see demand for more until well into the next decade. Update Sept: Confirmed. World's first 5G goes live at Softbank Japan. It's Massive MIMO, not millimeter wave.

 5G requires < 1 millisecond latency. 

Marcus Weldon of Nokia opened the 5G Summit with a stirring call for < 1 millisecond latency for high speed traffic control as well as future virtual and augmented reality. Huawei supports the same, but virtually all the telcos think driving to 1 millisecond is unnecessary and wasteful. It will be brutally expensive to bring speeds that low. Besides the short range, each 5G cell would need expensive intelligence to avoid router delays going up the network to a switch. A few hops through routers would slow things down too much, so traditional wireless architecture can't achieve that speed. (That range may not be as short as previously thought. Beamforming is working well.) September update: David Small of Verizon gave details of their Boston 5G plans. They will use controllers in the cloud (C-RANs,) perhaps only half a dozen for the entire system. A system like that can come under 10 milliseconds but almost certainly not 1 millisecond. 

 The delay in standards to 2019-2020 will prevent earlier 5G.

Tom Wheeler at the U.S. FCC has decided to eff the 3GPP-ITU procedure and authorize a 2018 deployment. Verizon almost certainly wants to move quickly. Their business model and stock price depend on being perceived as by far the best in the States but others have almost caught up. T-Mobile claims to run faster in NY than VZ. Verizon's McAdam wants the pr; he also hopes the FCC will let him monopolize the spectrum before anyone notices.

Millimeter waves always have short reach and require line of sight.

Beamforming is dramatically improving the performance of millimeter wave. Verizon demonstrated 1.8 gigabits 174 feet using 800 MHz of 28 GHz spectrum without good line of sight. NTT CTO Seizo Onoe calls "5G is always short range" myth. Consumer millimeter deployments will nonetheless require a very high number of base stations.

Fiber is always needed for backhaul

Ericsson provides convincing evidence that microwave wireless, now available up to 5 gigabits, has latency and other characteristics appropriate for 5G. For now, fiber is the primary choice.

 Virtual Reality requires 1 millisecond

Oculus Rift runs at 11 milliseconds.

 Wireless networks are often congested

Deutsche Telekom is promoting a DSL/LTE combo router that draws on the LTE signal for higher speeds. LTE downloads now reach hundreds of megabits peak speed, soon going to a gigabit. That dwarfs most DSL speeds.  Drawing on LTE bandwidth makes sense, because something like 95+% of wireless nodes are not congested most of the time. (No hard data available.) AT&T cut capes as they introduced the iPhone, yielding severe problems in media capitals New York and San Francisco. Since then, claims of major congestion have been given credence. Most of the problems have been remedied. Congested nodes are surprisingly few and even those have spare capacity most of the time. That makes free unlimited overnight for downloading attractive. Telco lobbyists find exaggerating the problem useful. Telco suppliers like to point to future demand. 

Densification is necessary to get increased capacity

Verizon and AT&T have doubled and tripled capacity while adding very few towers or small cells. Carrier aggregation and now MIMO have often been enough to deliver the needed capacity. Brett Feldman of Goldman Sachs and the CEO of Zayo have both confirmed the slow growth in the number of cell sites. Investors have lost billions betting on the predictions that telcos would need to add many sites, including one from the top of Verizon. It didn't happen. Verizon is again talking about small cells but I haven't seen them. 5G highband will require many cells because of the short reach. The high cost of cells and backhaul will be a major drag on hignband deployment. Some think the cost is so high 5G hignband will not make sense most places. We're all waiting on data.  Verizon in 2016 is installing small cells where they have good fiber access. The cost is down to about $20,000/per.

Spectrum is the main problem for capacity.

AT&T currently has 40 MHz of unused spectrum across the United States. More spectrum is a good thing because it reduces the cost of adding wireless capacity. But MIMO, densification, and many other technologies can also add capacity, often more cheaply. Marty Cooper claims, "There's never been a spectrum shortage and there never will be." He built the first cellphone. There have always been sensible choices for adding capacity, although it can take several years. 

There is such a thing as 5G wireless.

It's purely a marketing term. Six months ago, experts at the Brooklyn 5G conference considered Massive MIMO a crucial part of 5G. Last week, Ericsson announced they would be shipping Massive MIMO in 2017. They are calling it 4.5G and 4.9G, because everyone knows 5G is coming later. Things are constantly improving in wireless at a ferocious rate. Whether they are 5G, 4G, or something else has become arbitrary.

pCast will cover most of San Francisco in 2015, Boston and many sports stadiums in 2016.

Steve Perlman of pCell has a reality distortion field second only his mentor, Steve Jobs. In 2014, he convinced the NY Times and a dozen others his "revolutionary" pCell would bring on a utopia of almost unlimited bandwidth. What he actually showed was a demo rig of MU-MIMO, which was already the mainstream of research. MU-MIMO and Massive MIMO are now starting to deploy; pCell still hasn't been seen in the field.