A term that was much funnier during the previous administration, Caltech theoretical physicist John Preskill coined "quantum supremacy" to mean the point at which a quantum computer is sufficiently large and error-free that it can perform a task that cannot be calculated by the most powerful supercomputer. It's one of the milestones of the field that tends to get hyped up, and Google has been at the forefront for releasing blueprints to achieve it. Google maintains they will demonstrate quantum supremacy by the end of this year, although you can watch John Martinis, Head of Quantum Hardware at Google, push that timeline into 2018 here and here.
That we're even talking about this in the near term seems pretty exciting, until you read that Google plans to demonstrate quantum supremacy by randomly choosing qubit operations and showing that the scrambled output is more scrambled than could be made classically (technical details here). This quantum speckle problem is analogous to a laser pointer, that is, the resulting interference from something coherent going through something chaotic, as described by the man himself John Martinis here. Still, researchers think (and hope!) that a quantum computer will demonstrate a useful result that a supercomputer cannot in the next few years.
Last month, IBM researchers simulated the quantum speckle problem with a classical computer for both 49 and 56 qubits. Or, to be more specific, their "results confirm the expected Porter-Thomas distribution as the distribution of the outcome probabilities for universal random circuits," which is exactly what Google is planning to do with a 49-qubit quantum computer. But this result means they'll need more than 56 qubits to show quantum supremacy with the speckle problem. Or they'll have to find a quantum problem for their 49-qubit computer that is immune to the "slicing" technique used by the IBM researchers. Either way, I'll put good money on quantum supremacy not being demonstrated this year.
Then, IBM announced a 50-qubit prototype device and said that a 20-qubit device will be available to clients by the end of 2017 (IBM already offers access to a 5-qubit and 16-qubit device to everyone). Qubits in these devices are arranged in a surface code lattice and reported to have an average "coherence time" of 90 microseconds, quite high for superconducting qubits. From the press release, “Clients will have online access to the computing power of the first IBM Q systems by the end of 2017, with a series of planned upgrades during 2018.” As of now, not much else is known about them, but I'm sure we'll find out more soon. For devices that operate at a hundredth of a degree above absolute zero, the race for a functional quantum computer is heating up.
[Editor's note: content was modified after the original posting]