Otherwise known as "Skate to where the puck will be, not where it has been. -- Wayne Gretzky."
This is fairly common in other software-related industries, eg. mobile. Uber was not physically possible when the company was founded in 2008 - most phones didn't have GPS, and when they did, it drained the battery too quickly for the phone to charge from a car. Same with Whatsapp's founding in 2009 - the push notifications that let it catch on weren't actually available until a year after the product first came out. I know a number of Silicon Valley drone transportation companies that know their product isn't feasible today, but are betting on improvements in battery technology.
Oftentimes the market (and underlying technology) pivots into the startup, the startup doesn't pivot into the market. Many of the big pandemic successes like Zoom, Toast, Marco Polo, Peloton, etc. already existed before the pandemic, they just happened to luck out when everybody got shut up at home.
I've had this experience myself. I usually find that I was overlooking important inputs to, constraints on or goals behind the decision.
> Buy 100 now at full price and 50 servers that are twice as powerful two years later
As someone who buys a lot of hardware for my employer, I can tell you that performance is only one of several interrelated considerations that goes in to purchasing decisions, and is far from the most important.
For that matter, exactly the same thing can be said about the price.
Of course! Unfortunately and somewhat sadly, later on I discovered that these constraints and goals boiled down to one of a few categories:
1) Got to spend the budget this FY, otherwise we get our budget cut.
2) A big enough sale at once means the sales person is more likely to give me a kick-back.
3) The overhead of the procurement process is so obscenely enormous that it swamps any potential cost saving. A single procurement cycle instead of two is literally cheaper overall than, say, a $500K saving on the hardware.
4) The procurement process is not that expensive, but I am that lazy. It's not my money, so why bother optimising it? It is my time and effort to do the procurement twice however...
5) I can't count, and that 10% discount looked really sweet. There is some techo telling me to buy most of this stuff later, but what does he know? I studied finance! I'm an MBA! I'm a business person now!
All Intel-compatible servers have been essentially perfectly compatible with each other for decades.
All major brands will keep selling server kit. It's not like you have to throw away your IBM brand Intel-based servers and replace them with ARM chips from some no-name vendor!
My point is that in next year and the year after, and the year after that IBM will sell you faster and faster servers. Also... bigger and bigger. At roughly the same price point. With largely (if not entirely) compatible management systems, drivers, etc...
Speaking of drivers: VMware and similar hypervisors have almost completely eliminated driver compatibility as a concern. Clusters can contain wildly different gear without issues.
There's a lack of understanding in the industry by older managers that gained their experience in the "before times" when even the firmware had to be consistent for a cluster of servers.
For example, the local budget airline Jetstar would buy used servers from Ebay.
Why?
Why not!
Why would you need "support" if a server dies? It is stateless (diskless!) and just 1% of your capacity! Just throw it out.
Why would you care about driver updates if all of your VMs are using emulated hypervisor drivers only?
Why would you care if a server is "used" if chips don't "wear out"?
Etc...
Where can we sign up?
Also the machines have to go somewhere and that costs a lot because if someone can do it cheaper they'll just sell VPSs for cheaper than you can host them. Doing it cheaper means more compute per rack/watt which means newer hardware.
There are also sometimes nontechnical business considerations that matter enough that you make tradeoffs around them.
That is why so little storage is being built, thus far. Costs for storage are falling even faster than wind or solar ever did, because there is nothing to be overcome except limits on scaling manufacturing. The physics of energy storage, chemical schemes excepted, are a matter of freshman physics, so the drive is only how to apply the same principles more cheaply.
Simplicity is the greatest virtue: if you can make a storage system with just one moving part, you have the makings of a winner. Many involve anchoring something to the sea floor: an air bladder with a hose to an air pump and turbine onshore, charged by pumping in air; a float with a cable down to a pulley, thence to a winch and motor/generator onshore, charged by dragging the float down to the bottom; an air tank with an electric pump/turbine just wired to shore, charged by evacuating the tank.
The overwhelming majority of storage will not be batteries, ever. But molten metal batteries have massive advantages over lithium, longevity and heat tolerance probably most important. Iron-air batteries are much cheaper.
Hydrogen and ammonia synthesis will not be as cheap as the others, but have the massive advantage that, because tankage is cheap and transportable, storage capacity is unlimited; and when as much tankage as you care to keep is full, any extra generating capacity yields a high-demand product, thus extra revenue. Furthermore, if your local storage gets looking likely to run dry, you just buy more.
Compatibility with existing natural gas generators is a plus; thus, compressed air, liquified nitrogen, hydrogen and ammonia. Burning something with compressed air increases efficiency.
Energy Vault is bald grift, the Theranos of storage.
Why not? Lithium iron phosphate batteries work fine, don't cost too much, don't have a thermal runaway problem, are good for 10,000 cycles, require little maintenance, have good energy density per unit volume, and you can buy them right now. They're heavier than lithium ion batteries, but in stationary installations, that doesn't matter. Something better may come along, but they're good enough to deploy now.
Utility batteries, where used, will be iron-air or molten metal. Light weight is no virtue in that use.
If those numbers are correct and we want cars that each have 50 kWh batteries, we can have up to about 2.2 billion cars, so we either get more lithium (seawater is 1000x more in quantity but IIRC not economical at this time), or we don’t all get to have electric cars.
This is making me go back to an older preference of mine, which is to have a really fat wire (order of 1m^2 cross section) going around the planet. Take years (literally) to mine and process that quantity, but in principle you can get winter solstice pre-dawn energy from your antipode. The losses, while large, are not catastrophic.
Long distance transmission has become very practical as the top-line cost of generation has fallen far below historic lows.
You don't need a square meter in cross section. There will be a hell of a lot of long distance transmission lines. Singapore is all in on one to northern Australia. UK has signed a deal for one to Africa.
If there’s any single entity I expect to have both the economic power and originational capacity to do that, it’s the Chinese government.
The main reason is that the cables are much cheaper than the storage (so a massive incentive if you can do it), but 15000 km * 1m^2 * density of copper is 1.3e8 tons, and current global annual production is 2e7 tons, and a similar problem for aluminium. Need a lot of organisation of the wider economy to pull that off.
https://www.pv-magazine.com/2021/11/15/chile-wants-to-export...
Pretty big categorical statement.
The problem with the "one moving part" storage systems is that their energy density is surprisingly bad. Gravitational potential energy in particular is terrible. Compressed air systems can be OK, especially the ones using underground reservoirs, but have an efficiency limited by the physics of gas expansion.
It's pretty hard to beat redox reactions for energy density. As you say, tankage is cheap, so can this be done with "flow batteries"?
But the real reason lithium is winning is the incredible power of path dependence and manufacturing efficiency. Is it the optimal chemists's whiteboard solution to churn out a billion 18650 cells? No. Is it the one that's easier to actually do in the world as it exists today? Possibly.
Chemical synthesis systems suffer on cost per watt to charge, but hit it out of the park on joules stored per dollar. Further pluses are bulk-shippable storage media that may be bought and sold, and can be discharged using existing gas turbines, even in places where you have not installed any synthesis apparatus.
It's probably best if we increase the carbon price until the market figures out how much storage is economical.
Huh? Why would you not charge your utility battery bank with surplus solar or wind power?
Bonus: at the point when solar or wind power would be curtailed (when it is generating more supply than the grid demands), is also when the wholesale price per MWh is zero or negative, making it free or profitable to charge your batteries.
Because you spent the money on more generating capacity, instead, and displaced exactly that much more carbon emission.
You put any excess power on a transmission line to someplace else that would otherwise be burning fossil fuel, and collect revenue. Maybe spend that on more generation.
If you are maxing out your transmission lines, spend money building those out. The more of those you have, the more places you can get revenue from.
When it starts to look like there will be nobody left to sell your excess to, then build storage.
Just from an economic standpoint it can also be a valid choice to build storage next to unpredictable generating capacity, for example if you build a wind farm and want to sell during daytime (when prices are highest) as much as possible.
(Technically, a 100% renewable grid could go negative I'd there were separate subsidies, but as soon as renewable goes below that level it would just be shut off rather than pay money to generate.)
That is better though for when the renewable plant isn’t running.
But I do like the idea that these industries might actually be motivated to progress faster now if people hold off on getting on board. Seems like its enough of a reality that they can justify r&d to make the last push toward real competitiveness.
Then again my brains fried from work and I don't know what I'm talking about.
All I know is I want me some solar panels and an electric F150.
The best thing for a house power bank is a used-up electric car.
Instead, for the maximum cleaning up of energy, we should build nuclear for the grid and use batteries on transportation.
But in particular, the iron-flow battery story going around recently stinks pretty badly. It's being heavily promoted almost entirely by one company, ESS, and their first real client is -- wait for it -- SoftBank [1]. The academic reports on iron-flow batteries [2] make the technology sound a lot less mature than the ESS website [3], which incorrectly refers to vanadium and lithium as "rare-earth metals".
A 2018 publication [4] from Narayan's group at USC boasts that:
>Thus, by operating at 60°C and a pH of 3 with ascorbic acid and ammonium chloride, we achieved a coulombic efficiency of 97.9%. While this value of coulombic efficiency is among the highest values reported for the iron electrode in the context of the all-iron flow battery, further improvement in efficiency is needed for supporting repeated cycling.
However, further work by Narayan's group led them to replace iron chloride by iron sulfate in 2020 [5] which was celebrated by USC in a press release [6].
It was shortly after this that ESS burst onto the scene claiming iron chloride batteries with extremely long cycle life using "carbon composite" electrodes, "porous polyethylene separator" and a "polypropene spacer" [3], which are suspiciously similar to the graphite electrodes, mesoporous hydrocarbon-polymer-not-disclosed (Tokuyama A901 [7]) anion-exchange membrane, and polypropylene housing used in the Narayan group's 2016 paper [8] proposing all-iron-flow batteries for grid storage. It's worth noting that chemically unmodified polyethylene is probably not a suitable material for an ion-selective membrane, but it wouldn't even be the second-worst mistake on the page.
Yet ESS, despite having supposedly solved major problems that are obviously of scientific interest to active researchers, does not appear to have any names on its website, and cites no publications. Frankly, it sounds like another EEStor.
1: https://cleantechnica.com/2021/10/07/first-ess-iron-flow-bat...
2: https://dornsife.usc.edu/labs/narayan/all-iron-redox-flow-ba...
3: https://essinc.com/iron-flow-chemistry/
4: https://www.sciencedirect.com/science/article/pii/S245191031...
5: https://iopscience.iop.org/article/10.1149/1945-7111/ab84f8/...
6: https://news.usc.edu/166306/flow-battery-renewable-energy-el...
7: https://watermark.silverchair.com/jeecs_18_2_024001.pdf
8: https://iopscience.iop.org/article/10.1149/2.0161601jes/pdf
A dollar spent on renewables buys several times what a nuke could produce, and immediately, not ten years from now and buying fuel in the meantime. The money spent just on the fuel over that time would mostly pay for building the renewables.
Is it possible to just spend infinite dollars today and solve the climate crisis by tomorrow?
https://www.lazard.com/media/451905/lazards-levelized-cost-o...
Every year the cost of nuclear increases because we have fewer examples of successful construction and more examples of failed construction. The industry is in shambles, effectively dead. The US attempts at construction of AP1000s resulted in 2/4 failing, and the other two reactors being several multiples behind in schedule and pricing. The latest excuse for the failure is that they began construction before design was complete, so of course they failed. However this was the request of the nuclear industry, in an attempt to bring down prices, and the entire regulatory approval process was changed to accommodate this, which was supposed to bring down prices and prevent the failure of construction. Look at any attempt to build nuclear in a modern economy and you will find failure, not success.
As for spending infinite dollars to solve climate change, no, that is not possible. There are real productive limits to capacity to build things. The solar, wind, and storage industries are growing at massive rates, but still its only barely enough to meet the speed needed for our energy transition.
If we had infinite money to spend on nuclear, we still would not be able to build sufficient new restore by, say 2040. In the US alone we would need to build ~100 reactors simply to replace those reaching their end of life. We do not have the construction capacity for that, much less a design to build, or willing financial backers.
For the foreseeable future, nuclear is a dying industry in the US, not because of regulation or public backlash, but because the industry can't build.
The only hope for nuclear in the US or Europe is for small modular reactors, a design that in the past has been rejected for being too expensive. But since it's closer to manufacturing (like a plane) than like construction, there's hope, even if it's a long shot.
I.e., if you are trying to scare up money for a big enough nuke plant to be worth installing, the stakeholders you would need on board will see noplace to skim off the money they demand to greenlight the project.
Thus far solar and wind seem thus far resistant to graft, for reasons that are easy to speculate about, but hard to prove.
It's much easier to take some graft off a super size construction project with few bidders and massive transaction costs compared to small repeatable transactions that happen with smaller projects.
Nuclear construction often ends up with people in jail. It's happening in South Carolina, and happened in South Korea too, and up until the corruption was found, SK had been touted as a modern nuclear success story that could maybe be replicated in the US.
So we are left with only China and Russia's Rosatom as the only builders that claim to be able to deliver at a reasonable cost. We just need to trust the builders enough to construct in our countries, with our workforces, and somehow get a hugely complex construction project with lots of high-precision welding and construction pours done on time and accurately.
Solar projects are useful at smaller sizes, so need fewer stakeholders, making it easier to find honest ones. People choosing to be involved with renewables are more often self-selected for idealism.
That's because the graft happens earlier in the process before the actual "build the thing" portion so you don't notice. The developer typically pisses away money directly or indirectly getting on the good side of the local powers that be before actually pulling the trigger on the project.
Contrast with nuclear or any other centralized power generation where the state gets involved. Sure, money gets pissed away in similar ways on those projects (pay off special interest X, promise a favorable rate for Y, etc) but it tends to not technically be graft because it's all done through the official processes.
Prices for big solar installations are in the public record. And for nukes. Recently North Carolina and Georgia spent, what, $15B for exactly 0 watts out. They were quoted another $10B to get the 2GW they had signed up for, which they had expected to pay, what, $8B for, total? They won't get any of it back.
The corruption tax on nukes is withering. Nobody involved wants the money to ever stop flowing, as actually delivering would cause.
The best I have seen for nukes is $2B/1GW, but nobody knows how to get that with any reliability; and that is discounted by a huge government disaster-insurance subsidy, and excludes ~$1B end-of-life decommissioning. I see $1B/1GW for recently finished solar projects, but prices are still falling fast.
No. Factories need to be build, infrastructure needs to be build with infinite money and an coordinated centrally planned effort it should be possible within 15 years to be net neutral.
This would include: electrifying all of africa, world wide giga grid, replacing all combustion motors, building a new fleet to replace all cargo vessels, build rail to curb all non trans ocean flights, radically cut down militaries all over the world, build a lot of heat pumps, building a lot of buildings in a carbon neutral way, also a lot of other environmental concerns (species protection, eco system protection) would need to be curbed for it to happen in 15 years, more flexibility if you relax that timeframe.
Modular thorium reactors would be a huge win if realized.
You need to watch some thorium debunking videos. I live near Indian Point, recently shut down. They tried thorium, early on. It cost too much. Every single thing about nukes costs too much.
I skimmed the Indian Point reactor -- it appears to be a non-LFTR reactor, and that seems to be where the excitement continues.
But yeah, renewables are great and only getting better. If we had taken a trillion dollars out of the fiasco of the Gulf Wars (ostensibly for "energy security") we could have done significant things. For example, I'm enamored with the possibilities of geothermal around the Yellowstone caldera -- if we could figure out how to do that without destroying the local environment.
Rebuilding The Grid with HVDC would help too, as well as an ammonia economy to utilize excess power from wind. It all seems very technically doable, it's the politics and petrol people that stand between us and a carbon free energy ecosystem (well, with reasonable exceptions for aerospace and other special cases)
Really, anything that needs a steam turbine is going to cost too much to compete.
If your electricity generation is fossil moving to EVs is of questionable benefit.
> Instead, for the maximum cleaning up of energy, we should build nuclear for the grid and use batteries on transportation.
We are talking about solar power costs exponentially dropping and your suggestion is to build nuclear, the slowest to build and already much less economical than solar. By the time your nuclear power plants are build you could buy ~10 times the capacity in solar and likely would not need any battery storage.
Several of the established players are close to achieving dry-low-NOx 100% hydrogen gas turbines. Then you are talking about >500 MW power per unit and a thermodynamic efficiency of >65%. If you are in the "extremely abundant but too variable renewable power" scenario, these things will be the major stabilisers. You can easily imagine smoothing out even seasonal fluctuations with them.
There has been some work on modding how well this could work for changing EV rebate incentives, but getting legislatures to adopt such complicated ideas is nearly impossible.
Another approach would be to add a realistic, risk-adjusted carbon tax to gasoline (probably north of $200/ton co2, or $2/gallon gasoline) and let the market sort it out. Unfortunately when it comes to car purchasing, consumers are less economically rational than even legislators.
Over the long term, any batteries sold are good batteries, because batteries exhibit large economies of scale. Profit on existing sales can fund new production facilities; incremental improvements in battery technology; R&D into capacity increases; improved distribution networks; R&D into new sources of lithium; and so on. The way you get dirt-cheap storage is to build lots of it.
In the more exotic architectures and chemistries, what's moving the needle is the exotic military or medical need, where the battery can be 10x more expensive on the OEM BOM and barely move the actual end unit price to the final user.
There is no realistic way to GENERALLY determine the price of the kWh coming out of a grid scale battery due to the large number of variables, such as battery lifepan, wholesale production rates, land and permitting cost, etc.
Capacitors are capable of delivering high levels of power but have relatively small energy storage capacity compared to batteries.
Grid scale batteries need to be sized in terms of both charge/discharge rates (kW/MW) and energy storage (kWh/MWh)
Short-term storage wants high charge and discharge rate, efficiency and durability. Longer-term storage mainly favors cheap capacity, and tolerates low charging rate and low efficiency.
I don't quite understand what you think is missing though, every industry and company I've had experience with has forecast things and planned based on some estimated intersection of many moving parts.
The thermal coal industry has to forecast electricity demand, competition, coal demand, coal prices, building and mining prices, regulatory environment, etc. to build and operate mines and power plants., out to about 50 years.
I don't know how much renewables have really caught the industry off guard, but rapid advancements in technology has caused many to come unstuck, especially those in the technology industry (Nvidia et al killed SGI, Intel killed DEC and Sun, and so on).
It's not that it's just renewables catching the old dinosaurs off guard here, it's going both ways because energy is obviously very volatile and highly complex. The Moore's law era gains in transistor performance and density were obviously vastly larger changes but they were actually very consistent and easy to predict by comparison, some guy can't remember his name now even made up a trivial formula for it that held up extremely well for 50 years! With a few relatively minor tweaks. In contrast there have been proclamations for many years now that wind and solar have killed coal. That the price points have crossed over, that coal is dead and coal infrastructure is a worthless asset. Today, thermal coal prices are at historic record highs. Someone is buying it and someone is making a killing digging it up. And a lot of executives are kicking themselves for divesting at the bottom of the market thinking it was worthless except for the free kudos they thought it would get them boasting about something they were going to do for financial reasons anyway.
It's hardly anything to boast about predicting silicon device improvements when others struggle to predict energy markets.
I learned quickly that the constants were the slopes of the best fit on a log graph. And that you could never intersect that curve with $today, you had to estimate your delivery date and use the intersection at that point for all planning. E.g.: vram budget for textures, polygon count limits, etc…
I’ve never seen this kind of thinking by management at any level, in any industry outside of computer game development.
I’ve tried explaining it to IT managers to no avail.
E.g.: don’t get a 10% discount now by buying 200 servers up front in a single sale! Buy 100 now at full price and 50 servers that are twice as powerful two years later.
Nope… got to get those sweet volume discounts…