There's a hard upper limit. 1 kW/meter^2 (mis-typed as 1W above). And you've got diminishing marginal returns of more efficient cells. The truth is that the land requirements are likely to go down by a factor of 2-4 maximum. Other cost factors (transmission, but especially storage) will dwarf these.
Once you've provisioned land for solar, the key costs are in replacing the panels every 20-40 years or so. Wind, stones, hail, and simple degredation will mandate this. Physical support infrastructure is likely more robust.
Land use for solar isn't dedicated-purpose for most applications. Even solar thermal can be used for grazing or other uses. PV can simply go on top of existing structures. You're not talking land acquisition so much as site acquisition.
Solar and wind are likely to be overprovisioned where possible. In the sense that you'll provide more capacity than is strictly needed to meet electrical demand. In part because you're not going to get 100% duty cycles ("capacity factor", which is how power installations are rated, are typically ~20 - 40%), and because you don't have an accelerator pedal, only a brake. Solar and wind aren't dispatchable, only sheddable. With overcapacity you've got the option of converting "excess" into other usable and storable forms of energy (hydrogen, methane, battery storage, electricity-to-fuels, etc.), or for intermittent but high-energy needs.
We've got cost reductions built in to PV production for the foreseable future. Again, efficiency improvements need to be considered in terms of total cost per Watt / kWh delivered*.
>We've got cost reductions built in to PV production for the foreseable future.
In which way? Production capacity-related economy of scale? Current production capacity is already close (well, at least half of) what can be sustainably loaded worldwide. Maybe, another 2x growth is possible because factory builders don't care much if their business is sustainble long-term, they will work until worldwide demand is filled. But i will be very surprised to see over 300 GW/y production capacity worldwide anywhere in the coming 10 years. Apart from economy of scale, there is little way to reduce price.
Imagine a 10 MW power plant built out of panels with 10% efficiency. If you replace them with 20% efficient panels, you increase output 2x without measurably increasing soft costs - like land provision, fixtures on which the panels stand, installation work, insurance, etc. Only part that will increase will be inverters (not a particulary big item) and thicker cables connecting it to the grid. So, soft costs per watt will fall by 40-45%.
So, given that the wholesale price of solar panels now stand at 40 cent per watt, and even in cheapest countries the complete plant is $2 per watt, increasing efficiencies 2x means building a 2x more powerful plant for about $1.4 per watt (soft costs fall from $1.6 to $1.0 - i.e. they slightly increase but power output doubles) and hard costs stay the same. Big deal.
And almost all of the progress in solar power costs we can have now lies in this field - increase of panel efficiency, not decreasing their costs. Even if panels were free, costs won't fall much...
Of course this article bears no relation to that - these are high concentration cells, and cost to build collectors is already much higher than just buying enough of flat panels... better show me a non-concentrated cell with at least 35% efficiency.