This is the exact scenario they are being used in at scale. The company removes their gas peaker plant infrastructure and replaces them with batteries. Already have the grid interconnect and now can dispatch power on the millisecond level instead of hour level.
A "peaker" or Open Cycle Gas Turbine is much less efficient which is doubly expensive in Europe. Firstly you need more fuel, and the fuel is expensive, but also you're making more pollution and that's expensive too.
The UK for example basically doesn't have "peakers". Right now it's early evening, demand is high as people cook evening meals but haven't yet retired to bed where they stop using electricity, but renewable generation is reduced as the sun approaches the horizon. So there's 8GW of combined cycle gas power plant production, but only about 100MW of "peakers" and it might grow to 200MW or so at absolute peak.
This is not true. Batteries are cheaper than peaker power plants using fossil fuel. They also allow the operator to fulfill market demands at the minute level versus the hours previously that it took to turn on a peaker plant.
This is being done at scale in California and Texas.
Fore more than 2 hours capacity, batteries are expensive than most other and cost keeps increasing as more hours of capacity needed. Without gas or coal to burn when 1hr battery capacity runs out, battery storage is expensive.
To scale battery storage to a level that is capable of bridging, say, 48 hours of "Dunkelflaute" (darkness and no wind) on a regional scale (e.g. the entire Scandinavia) is probably unrealistic. Just the amount of lithium needed would be insane. And there were longer Dunkelflautes in recent history.
New advances in nuclear is what I hope for. First experimental SMRs are being installed in several places of the world, others are in design stage. Looks like a hopeful technology.
> To scale battery storage to a level that is capable of bridging, say, 48 hours of "Dunkelflaute" (darkness and no wind) on a regional scale (e.g. the entire Scandinavia) is probably unrealistic. Just the amount of lithium needed would be insane. And there were longer Dunkelflautes in recent history.
48 hours in Scandinavia is roughly equivalent to turning all their road vehicles electric. And that's even with Norway using the second highest per-capita rate of electricity in the world let alone Scandinavia (second to Iceland, whose electricity is 100% renewables thanks to abundant geothermal): https://www.statista.com/statistics/383633/worldwide-consump...
Given nobody is suggesting an instantaneous transition, this is not at all unrealistic, and I don't know why anyone might consider it to be.
Good luck with new nuclear, but with all the politics in that domain, I don't expect that to work out even if e.g. Helion Energy supplies working shipping-container-sized aneutronic fusion.
You can't be serious? I hope I'm just missing the joke
To have an equivalent capacity of DCs in space would require such astronomical costs there is 0 chance it would be profitable esp considering how fast GPUs deprecate
How about we start with actually finishing all the datacenters we've started (many if not most are unstarted, paused or outright cancelled)?
So by your logic if a massive company came and said they want to setup an oil refinery or coal power plant you'd say they should say yes just because they could get tax revenue?
You always have to way the pros and cons of such massive projects.
Plus, this isn't a ban it's just a one year moratorium so impact etc can be studied.
"AI" data centers consume massive amounts of energy per square foot, have been caught using LNG generators for power, have polluted ground water, material usage, habitat destruction, etc
How about you provide some proof?
(PS I'm not advocating that golf courses are good for the environment - they're not - but data centers are absolutely not "zero on the pollution scale" that is laughable)
Minimal impact as defined by whom? I'm sure they impact those that live around them plenty as well as the price on electricity for those on the same grid.
They provide mostly temporary jobs (and majority imported to boot), after construction they're run with very little staff.
New York needs to upgrade its power grid (source?) so they should force it upon themselves by primitively overloading the grid?
A one year moratorium while impact is investigated, esp considering the current state of datacenters buildout, especially considering many in the US are either unstarted, on hold or abandoned, seems reasonable.
Lets say they build 100 Data Centers in Oswego county. Per Capita income is only $20k and one of the poorest in new york state. There is a lot of space that is not used for anything. 100 Data Centers is going to provide some amount of jobs at each one, remote hands, janitorial, security, etc. Those all pay more than the per capita income. At scale you need some amount of plumbers, facilities, electricians, etc. Not to mention the local labor that gets hired to build them. I know plenty of guys working Steel that leave to go build Data Centers in other places, and go back to upstate new york. Enough Data Centers means more power plants for even more jobs.
In poor rural areas, 50 jobs paying not that much is a big improvement.
Local labor is often not used to build them. Most of these companies have crews they bring in during build out. Plus the actual setup of the facility is highly skilled: everyone is imported. The only local jobs we're talking about would be things like security, cleaning, and maybe some site prep work. The actual technical staff will not be hired locally in the majority of cases.
The argument here tho is the amount of jobs provided does not do a good enough job counteracting the downsides (noise, ground water pollution, generator air pollution, grid load and offset costs to customers, etc). In most cases these far outweigh <50 jobs if you look at overall cost/benefits.
Local labor is almost never used for these projects. Data centers are built by huge national or multinational construction companies, not a local general contractor.
Then the local utility companies have to undertake giant projects to keep up with demand. They issue many millions in bonds. Also, all of the construction trucks tear up the local roads. The local municipality has to allocate a bunch of money to fix the roads.
Then, next year, the AI bubble pops, and most of those data centers shutter their doors and stop contributing tax revenue.
Now you have the same $20k per capita income, but many millions more in debt!
It's pretty well documented that datacenters (esp the AI variety) are offloading grid expenses to customers as higher baseline costs.
The issue is that these DCs are not paying for the cost of the necessary grid upgrades but are instead having the power cos pass these costs off to all consumers as base increases.
I agree with your take on tradeoffs when it comes to DCs that are built with tax abatements, which are a terrible deal for consumers. Those that are not receiving tax benefits will make up for short-term energy prices with long-term property tax benefits, that will eventually pay back the grid upgrades and fund even more upgrades in the future.
Are you talking backup generator vs solar for a home?
If so, solar continually supplies power without paying for an input vs a backup generator which is only meant to run infrequently and is costly to run and requires you to pay for inputs and of course maintenance of an ICE.
- The workers platform is quite pleasant to work with compared to competitors.
- Globally deploying edge workers which have access to their many services (D1, R2, DO, etc)
- Having the ability to assemble globally distributed workers using bindings is dead simple
- Their CI pipeline, while limited, is easy to setup and run and keeps improving
- Their pricing is extremely competitive
For your second:
- That's my biggest conflict with using any service (Cloudflare, AWS, Azure, etc). Don't have a good answer with what to do about it considering for many projects I don't have the time/energy to fully self host everything.
One thing is hosting, which obviously comes with centralizing risks but a different one is just deciding to add a layer of "protection" in front of every website so that as much traffic as possible goes through one single company.
The dashboard UX has improved a lot lately but one thing that drives me absolutely nuts is that I get rate limited all the time using it.
For example, I had to recently change an env var we had on a handful of apps and opened them all into new tabs and made the changes and about half way through I started getting rate limited. This has happened to me many times and I've reported it to support and in Discord but it still happens.
One other big complaint is support is non-existent. We sent many support emails (on business plans) and I'm pretty sure we've never gotten a reply. Same for posting in Discord. It's pretty disheartening to build your business on Cloudflare and have no confidence support will help you when you need it.
yeah you should definitely not be getting rate limited, sorry this is annoying you're not the first to report i will dig in.
as far as support, i know there is a huge effort going on right now to improve response time and support in general, also I'm not as active in discord as I ought to be there's just so much noise, feel free to ping me on there directly if I can help brandon/@ygwyg. can't promise it'll be an instant response but I will respond
You really think an individual employee trying to help a customer when they likely have little impact on org wide issues like how customer support is run is tarpitting? Seems a bit uncharitable
reply