IBM, Google and Rigetti are working on approximate gate model systems.
Rigetti has a 19 qubit chip.
The time to scale up and develop new chips with more qubits is a matter of how effectively the larger number of qubits can be tested.
There are multiple competitors with the non-error correcting gate model systems. I would predict that at least one and possible more of the competitors will try to create and test and publicly announce a 100 qubit system in 2018. A 200+ qubit prototype could be built by the end of 2018.
I think the multiple competitors in quantum computing will force a more aggressive development schedule.
DWave Systems has a 2000 qubit quantum annealing system and is working on a 5000 qubit system. The 5000 qubit was projected to be installed at a client site in 2019. This would mean that the 5000 qubit system is being tested now at Dwave. Dwave will probably publish speed tests and research papers about the 5000 qubit system in 2018. Although Dwave may choose to only use about 4000 or the 5000 total qubits.