Could solar supply all our electricity?

Yes, it could if necessary, but it wouldn't be the cheapest way to do things.

It makes no sense to try to solve our supply problems with just one source. A pure solar US doesn't make much sense for a lot of reasons:

1) The US uses an average of 450GW of electricity. At 20% capacity factor, we'd need 2,250GW of solar to provide that much. 2,250GW of solar would produce a peak of production about 3x as high as noon-time consumption (very roughly 750GW). PV doesn't have built in storage, so that would be a big problem. Demand Side Management could help (see my earlier posts), but at some point there would be diminishing returns. For CSP, even if molten salt storage is pretty cheap that's still a lot of unnecessary infrastructure.

2) Winter is also a big problem. In particular, wind is better at night and during winter. A combination of wind and solar makes much more sense, as they're complementary. We can keep existing fossil fuel plants around for the 5% of the time that we have prolonged, widespread calm periods, or use bio-mass (wood provides 1% of US electricity currently, and is much more sensible as a use for biomass than ethanol).

3) Solar is more expensive than wind (currently at least $15 per average Watt vs $6/W for wind), so we need a higher proportion of wind in the mix. Both wind and nuclear will provide power during the day, and solar would only make sense for the peak component. I can't see a need for more than about 500GW of solar, which would give us a market share of very roughly 30% of KWH's.

Here's an extremely simplified preliminary model: let's assume today's 450GW consumption, plus 75 for PHEV/EV's for a total of 525GW. Current consumption is probably 250GW for 7PM to 7AM, and roughly 650GW for 7AM to 7PM. PHEV/EV charging might raise night time demand to 350GW, and daytime to 700GW. Wind and nuclear might provide the baseload of roughly 350GW and solar could provide the daytime an extra 350GW. That would give an average from solar of 175GW, or 33%.

What about net-metering - doesn't that allow consumers to provide all of their needs through solar power?

Yes, but most net-metering programs are limited by statute to a small % of KWH's, perhaps 1%. Those caps can get lifted, as they did in CA lately, but it won't make sense economically to raise them much above 10% or 15%.

Couldn't ice-storage A/C and electric vehicles time-shift demand?

Sure, but there is a cost to ice-storage A/C, both in terms of capital and efficiency, and night time charging is a bit more natural for PHEV/EV's.

What about a massive PV farm?

Something like that would be on the utility side, and would have to compete with wholesale prices, which are half those of retail. PV is best on the retail side.

It makes more sense for PV to be on I/C rooftops, which is where 80% of CA installations (by KW) are happening, because of economies of scale and flat roofs. New construction is best because of integration with the roof, but these days new I/C construction won't provide a lot of square feet. The ideal size for consumer-side installation is about 90% of the consumer's noon demand - that eliminates the need to deal with selling power back to the utility, and maximizes savings.

Residential is 80% of installations, but they're much, much smaller than I/C on average, and much less economic. Residential only make sense currently because of the non-economic value to the home-owner, which can be substantial. Residential new construction might make sense in the future if the industry achieves very large integrated roof-module manufacturing economies of scale (right now PV roof tiles are surprisingly expensive), and if installation becomes very efficient through integration with the construction process.

How much would the non-module cost of a large PV massive solar farm run us?
The Balance of System includes controls, inverters, mounting structures, and wiring. These are a very fast moving target. Their prices have fallen very quickly in the last few years, and some innovative approaches are in competition. I would hope that overall system prices would fall to $2/W in 10 years, which would give $.10/KWH. That's the average US price currently, so I would think that it would beat peak prices in most of the country at that time.

As I discuss in a previous post, I think that we're now at grid parity without subsidies in ideal locations. That means that PV will continue to grow very, very quickly due to demand. Heck, even now in the worst bank panic since 1929, PV is still growing, albeit slowly enough that supply can finally catch up with demand and prices can begin to fall as quickly as costs.

Is there a problem with solar for peaking in the evening when people come home from work? Won't we still need natural gas for evening peaking?

That's the traditional utility point of view, which is incentivized by a regulatory framework that's based on capital investment ROI. When faced with a peak demand, they think first of new generation, then of expensive central storage.

Demand Side Management is far better, faster, easier and cheaper. Charge based on time of day, and sell cheap timers that lower the thermostat on the A/C (as well as the fridge) at noon, instead of middle evening. Overall KWH consumption rises slightly (due to a larger differential between inside and outside temperatures), but this would be far more efficient than ice storage.
Also, PHEV/EV charging will avoid peak times - GM is working very closely with EPRI and a large array of utilities and DSM companies to make it work.

People will do simple things like reducing lighting in the evening, which will reduce lighting KWH as well as A/C. People will be creative.

Dynamic pricing could cut peak demand, certainly. The regulators just have to allow it. When's that gonna happen?

Actually, it's not only allowed, it's mandated by the energy bill of 2005 - all utilities have to be offering it now. I believe PG&E is aggressively rolling it out over the next several years. The stimulus bill throws some money at this as well.

The utilities aren't all that excited by it, because the capex ROI regulatory model is still in place for most utilities. That means that most are sticking at the pilot program point, where customers can have it on request. Here's an example: http://www.thewattspot.com/ .

Even with dynamic pricing some people are going to turn on the air conditioner.

Sure, but 1) I wasn't talking about less A/C, I was talking about earlier A/C (and refrigeration) and 2) you only have to shift part of demand to move the curve earlier, to where solar shines (pun intended).

"Reduce lighting in the evening: This really undermines the utility of the light bulb. "
Much lighting isn't needed, and that if people pay a little more attention to turning out the lights when it's most expensive, that will make a difference.

I'd also note that if we move to residential PV, the capex needed for energy will become much more transparent to residential customers, and they're likely to realize that there are much higher ROI opportunities than PV, like better lighting, appliances and A/C. A/C in particular could be much more efficient with relatively small marginal investments at routine replacement points.

Won't the big cost cut for installation will come when solar panels replace roof shingles?

That certainly makes sense, but solar shingles exist now, and for some reason they're quite pricey - I'm not sure why.

That would be most true of new construction (after builders perfect the integration), as residential retrofits require a lot of custom work: evaluation of site, angle, and insolation and custom design, sizing, and wiring (including controls and inverter). Installation of wiring is going to be somewhat involved in most multi-story buildings, involving significant pulling of cable, with every installation requiring solving new problems. We only have 500K units/year of residential new construction lately, which would only get us about 2GW per year. Even at more normal construction rates, new residential isn't enough.

Industrial/commercial flat roofs are really the best place for retrofits: you tend to have unobstructed insolation (low-rise I/C districts, and few trees), flat roofs, much simpler wiring (conduit and cable chases designed for easy access and additional wiring) and much larger installations, which give you economy of scale. It's a nice fit with chains, like Walmart, that have access to a lot of buildings of similar design.