The current largest sample of quasars, the SDSS quasar list (Schneider et al. 2010) contains about 100000 objects. Given a typical optical depth in lensing galaxies towards z ~ 2 sources, this would imply 100-150 lensed quasars, of which at least half have already been found in the SQLS and other surveys. In principle, many more active galaxies are available in radio surveys - the FIRST survey, for example, contains about 106 objects with accurate position information - but their faintness in the radio makes systematic large surveys difficult, even with the current generation of upgraded radio arrays. Of such arrays, the only one which has the required combination of high sensitivity and sub-arcsecond resolution is LOFAR, a low-frequency radio interferometer array centred in the Netherlands (van Haarlem 2005) but high resolution surveys at great depth are probably a few years away.
In the optical, a combination of wide area and high resolution is likely to be achieved in the near future, by a number of telescopes. The first is GAIA, scheduled for 2013, a satellite primarily designed for astrometry and measurements of proper motions in Galactic stars. By virtue of area coverage and accurate measurements of point sources, however, it is also well suited to making a large census of about half a million quasars, to determine which such objects are extended and thus potentially lensed. This alone should increase the lensed quasar sample by a large factor (Surdej et al. 2002). Towards the end of the decade, two major advances are likely with the advent of the Large Synoptic Survey Telescope (LSST) and Euclid. Euclid is an ESA medium mission scheduled for launch in 2019 which will have close to all-sky coverage and 150-200 mas resolution at optical and NIR wavebands. It will provide imaging at slightly less angular resolution and sensitivity than the 2 square-degree COSMOS HST field, but over the whole sky. Its mission includes the detection of about 1000 quasar lenses, about an order of magnitude increase on the present sample (as well as hundreds of thousands of galaxy-galaxy lens systems). Still vaster samples will be provided by LSST (Abell et al. 2009), with the additional advantage of multiple observations of the same field, allowing quasars to be identified by variability (Kochanek et al. 2006) and probably yielding a sample of several thousand lensed quasars. On the same timescale, the Square Kilometre Array will provide similarly large samples, selected at radio wavelengths (Koopmans et al. 2004).
Large samples are good for two reasons. The first is that "more of the same" approaches can be tried with larger numbers, although they do rely on systematic biases being eliminated. If we assume that the mass slope and mass sheet model degeneracies can be controlled, so that the accuracy improves as the square root of the number of objects, then spectacular results can be obtained; Coe & Moustakas (2009) calculate that, in conjunction with Planck priors, an accuracy of ~ 3% can be obtained on w. This assumes that time delays will be measurable given the LSST cadence. If a smaller sample of the quasar lenses are measured, however, but with more intensive followup, then extrapolation from the work of Suyu et al. (2012) suggests that comparable results on post-H0 parameters can be achieved in conjunction with other (BAO/SNe) cosmological probes, since the lensing constraints are often orthogonal to others in parameter space. Linder (2011) calculates that the dark energy figure of merit is potentially improvable by a factor of 5 by including lensing information from future surveys.
The second advantage of large samples is that they are likely to contain a small number of high-value objects. One of the most prized type of lens systems is a quasar lens system with a second source at a different redshift, since such double source plane systems allow mass model degeneracies to be immediately broken. The problem has been investigated by Collett et al. 2012, who find that a small number of these rare objects give 15% accuracy in w very quickly. Many such sources would be currently difficult or impossible to follow up, due to faintness of some of the sources, but the era of 30-m class telescopes is around the corner, and such followup operations will become routine if not easy.