In the not-too-distant future the fair market value of all coal assets will be $0, or negative if governments force owners to safely decommission the plants and mines. This turning point will come when the total cost of energy from utility scale solar drops below the operating cost of coal power plants.

Ramez Naam has a fantastic post about this trend. This Exciting Tech post is written less well than most because really, you should just go read his post and Casey Handmer's rant about Australia neglecting this technology. The one tweak I would make to Naam's post is to mix some hockey sticks in with his asymptotes. Here's IRENA data on how much solar energy $1 could get you over the last ten years.

That looks a little scary if you own coal assets, doesn't it? The Al Dafra solar installation in Abu Dhabi will sell you energy for $0.0135 per kWh, or 74 kWh per dollar - that's twice the energy you can buy from a coal plant if they've already paid for the entire plant, and 7 times the energy you'd get from a coal plant paying for capital depreciation.

We don't typically talk about energy per dollar like this because most people don't consume energy based on its price. You aren't eyeing electricity auctions to pounce on good deals like eBay. You run your AC, washing machine, lights, and laptops then the utility sends a bill each month. You might turn the thermostat up if you live in California and PG&E's chronic mismanagement is starting to sting your wallet, but your usage is mostly fixed wrt price.

That "demand inelasticity" becomes a problem as more power generation transitions to solar and wind. The power grid is an eternal high-wire act: Every instant of every day power supply and demand must be perfectly matched. Too little power gets you brown-outs, too much starts frying expensive equipment.

We achieve that balance using a portfolio of power assets with varying levels of 'dispatchability,' which is the ability to vary the amount of power sent into the grid. Solar and wind are 'non-dispatchable' because we have no control over their generation. We can turn them off completely (and make no money), or we can let them run and their output is set by the sun's brightness or the wind's speed. Coal plants are 'dispatchable' because we can turn their burners up and down to make more or less steam, spinning the turbines with more or less torque. When there's more demand you turn your coal plant up and make more money. However, 'dispatchable' is a spectrum not a binary: Coal plants and nukes warm up and cool down over hours. Natural gas plants and hydro power can swing on and off in minutes to seconds, and grid-scale battery storage is the king of dispatchability with effectively instantaneous response. The more responsive sources are generally only turned on when real-time or short-term energy prices are high, while the slower sources trade on longer term auction systems with lower, less volatile prices and run most of the time.

The percentage of time a source spends running is called its 'capacity factor' and it's a critical component in coal's impending doom: As solar and other non-dispatchable sources push daytime prices lower, and energy storage systems start eating into the valuable evening hours, the profitable coal hours will continue to shrink, the amortized cost of coal infrastructure will be spread over less time, and the cost of coal in the remaining profitable hours will increase.

Storage and overproduction

Coal is dead, but it's not the only fossil we'd like to be rid of. Today, the non-dispatchable nature of solar and wind are providing some comfort to natural gas asset holders. The logic is that as we build more intermittent sources of power, the remainder of the grid will need to be more dispatchable to maintain balance, and that will lead to higher payouts for peaking natural gas plants. That might hold true short term, but there are two factors which will kill them in the medium term.

First, storage technology is accelerating fast. Lithium ion batteries are dropping in cost about as aggressively as solar panels, albeit with a later start. As batteries get cheaper, it makes more and more sense to drop them next to substations, suck in cheap solar power during the day, and sell it back at a markup in the evening and night.

Second, we can simply overbuild the renewables. At a certain price threshold it makes sense to just build so much solar that supply exceeds demand even on your darkest, rainiest days. The downside of this approach is on bright sunny days, you need to turn lots of your panels off to keep the grid balanced. If your darkest days produce a factor of 4 less energy than your brightest, you need to build 4x the capacity you otherwise would and - since solar is essentially free once built - that means your price goes up by a factor of 4 too. Cheers ring out from the fossil fuel lobby! Except, perhaps it doesn't have to. Turning off panels and throwing away almost-free energy is dumb. If you could find somebody willing to pay anything for your excess power on sunny days, it would make sense to leave the panels connected and sell power at rock bottom prices.

"Stuff that turns on when electricity is cheap" is called demand response. Smart grids are a form of consumer-oriented demand response where you consume energy you would have otherwise at more opportune times. Think running your AC extra hard while the sun is shining and then leaving it off during the evening demand peak.

The demand response I find more exciting is industrial: We're going to see a whole new market for insanely energy intensive but capital-light manufacturing emerge thanks to these power sources. Most big commercial electricity use today is inelastic as I said above: if you spent hundreds of millions on a server farm or a car factory, you'd be a fool to turn it off overnight to pinch pennies on your electric bill.

However, suppose you invented a new process for producing nitrogen fertilizers from air, water, and an obscene amount of electricity. The equipment required to run your process is fairly cheap, and it turns on and off quickly. It absolutely makes sense for you to turn that process off overnight, and just build extra equipment if your existing install can't meet demand running part-time. That company exists - it's called Nitricity and they're deploying their first systems on California farms right now. That category - demand response using cheap hardware to produce commodities - is going to bear a lot of fruit. Water desalination, hydrogen production, metal ore electrolysis, and more. Cheap hardware is also usually small hardware, so this revolution comes with a parallel shift from centralized to distributed production.

The other exciting possibility with overproduction is crappy storage. If power is effectively free all Summer but still has nontrivial cost in Winter, it makes sense to spend Summer energy for Winter energy at a 4:1 exchange rate. Energy storage with enough capacity to shift demand across seasons like that is an unsolved problem today, but I'm very optimistic someone will invent an obscenely cheap solution with say 25% round-trip efficiency. Direct-air-capture carbon into methane, or split water into hydrogen and oxygen, or pump insane quantities of air into pressurized caves deep undersea. The specific technology is an unknown quantity today, but I bet somebody will find a way to bank free energy.

Remember earlier where I talked about hypothetically building 4x as much solar as you need? The head of Australia's renewable energy program would like to do 7x. We're headed for a future of incredible energy abundance, and dramatically cheaper, cleaner sources for almost every meaningful commodity.