Battery trade war hits booming datacenter industry Battery energy storage systems (BESS) could become standard at datacenters as AI infrastructure expand, with analysts forecasting 20 GW of capacity deployed over the next decade. Hyperscalers building AI facilities progressively view BESS as essential to their energy mix, according to financial analyst Jefferies in a report …
In the Uk, you're better off with Givenergy than Tesla. Givenergy batteries play nicely with Octopus energy, they have a publicly available API, so you if you desire, you can set up your own custom charging/ dishcharging routine. (The controls for the battery can be accessed through Home Assistant, so you can even set up some simple "if / then" type controlling using a point and click interface.)
Givenergy? The company who doesn't answer phones, messages, letters, doesn't honour warranties and lies ALL THE TIME?
Octopus binned them as a supplier nearly 2 years ago because their batteries are junk and "support" doesn't exist. Ditto EON. The very best of luck finding an installer because every single one around here (East Mids) no longer deals with them - several have court cases pending against GE.
https://uk.trustpilot.com/review/givenergy.co.uk
Just don't do it - I bought their crap and I know I'll have to take legal action in order for the warranty to be honoured. They're scum.
No, not the Depeche Mode song. If the raw material your business consumes is electricity, you have to make sure the shipments keep coming. That could mean Grid, solar, wind, battery, thermal sinks and a market for your waste heat. Relying on just one source could mean problems. It also means locating somewhere with easy access to all, not just the place where the local government is going to pay for the facility.
If you were deploying a significant fraction of your datacentres daily energy use in battery storage quite a few possibilities arise.
In AU we have an embarras de richesse of electricity from solar photovoltaics during the middle of the day, so a battery equiped datacentre sucking some of this excess at a serious discount to use during peak demand times would be a win-win. If at any time the battery storage exceeded the datacentre's needs the excess could be supplied to the grid when the demand (hence price) is high.
With batteries you would be running a DC system viz grid AC to DC conversion followed by battery DC to AC conversion then back to DC by the systems switchmode PS. I would wonder if significant energy savings could be made by optimising or eliminating some of these conversions. I don't what the AC to the rack is today but I have seen a refurbished US HPC system with 120VAC three phase (later converted to 240VAC three phase) but I wonder whether higher frequencies (I think aircraft use 400Hz) and voltages might be more efficient.
Amusing saying Tesla batteries are made in the US when I am fairly sure the cells are still manufactured in the PRC. Tesla robotically assembles the batteries in the US and adds its dubious software. This might have changed in the last few years (not the dubious software) but I would think it unlikely. These cunning Chinese could instrument each cell with a homegrown embedded CPU which could talk to all the other cells in the battery and subvert the Tesla battery management systems (paranoid enough ?:) Good luck to them I say.
The Hornsdale Power Reserve in South Australia is effectively doing the same job you describe in paragraph one, saving daylight energy for use later - using Tesla batteries. However, it's for smoothing out the general mains supply surges and reducing overnight generator usage rather than dedicated to a DC as a backup/aux power system. The HPR also improves the stability of the mains supply, reducing the incidence of brown outs or cuts in the infrastructure. Sadly it can't stop the problem of an over-eager workman with a JCB digging near your DC.
In the large scale of power distribution, AC is better as it can transfer higher power over longer distances - there's lots of history to support that. Within the DC, AC to the row, or set of rows, and then conversion to DC is possible. With future DC rack designs aimed to cater for water cooling, the supplies are moved either to the bottom of the rack, or to the end of the rack along with the cooling systems.
I don't trust that anything "Made in the USA" is truly that, and the same here within the UK. Maybe "assembled" would be a more correct word as the components are usually built somewhere in Asia.
Individual cells already have a control processor of sorts, to handle faults and supply status information to the upstream battery management. I think this is a legal requirement in cars, and in many countries for solar hybrid power installations. What's in the software of that cell, your guess is as good as mine - it's amazing what can be done even with a few KB of memory and a "random wire". ;-)
> "I don't trust that anything "Made in the USA" is truly that, and the same here within the UK. Maybe "assembled" would be a more correct word as the components are usually built somewhere in Asia."
Logically, I suppose you could stick "Made in the UK" on anything, and if they wanted to dispute that you could argue that part of the final assembly *was* done in the UK.
Specifically, the part of the final assembly where you stuck on the "Made in the UK" label.
(Also, what happens if the "Made in the UK" labels aren't made in the UK? Doesn't that mean they're technically illegal? Shouldn't they say "Made in China", or whatever, underneath where it says "Made in the UK"? And how would one distinguish those from "Made in China" stickers that *were* made in the UK?)
> AC is better as it can transfer higher power over longer distances
I know that is true of overhead power lines, so curious why subsea interconnects are usually DC- DC..
I can partly guess as it is unlikely that UK/France are synchronised grids (!) and there is some fancy interaction between AC cables laid close to each other (you have to de-rate them)
> "I know that is true of overhead power lines, so curious why subsea interconnects are usually DC- DC.."
Wiki sez:-
"Unlike overhead powerlines, many submarine power cables tend to operate with DC current. Electrical phases must endure close proximity inside the cable, increasing parasitic capacitance. It is more economical to use AC only with lines shorter than 100 km in length, in which case losses at the landing point grid interfaces dominate."
Presumably something that's not a problem with electricity pylons, where the separate wires can be kept further apart.
Only historically because you could easily use transformers. At large size they are efficient and long life. Before high voltage semiconductors one solution to DC was simply a motor driving a dynamo (WWII rotary converters/ dynamotors) or an alternator. At lower power 6V or 12V DC to 200V DC (Valve car radios) the vibrator pack was used. Basically an oscillating relay and transformer. Rectification was either by metal rectifier, valve rectifier, or an extra pair of contacts on the vibrator pack as a synchronous rectifier (now done with FETs in SMPSUs). Vibrator packs are short lived compared to dynamotors.
There is less loss with high voltage DC than high voltage AC as AC is more affected by inductance and capacitance. AC also has skin effect, but it's likely negligable at 50Hz or 60Hz.
The DC incidentally means the two grids don't need synchronised.
The high power over long distances is due to high voltage, not AC.
Maybe 50 years ago there was optimism about superconductors.
>> AC is better as it can transfer higher power over longer distances
>I know that is true of overhead power lines, so curious why subsea interconnects are usually DC- DC..
>I can partly guess as it is unlikely that UK/France are synchronised grids (!) and there is some fancy interaction between AC cables laid close to each other (you have to de-rate them)
AC is not better over long distances. The frequency limits the maximum length of power lines, Light (and electrons) is slow in this context. And the constant change in current direction and strength generates a magnetic field that adds to the power loss of the resistance of the lines and being an inductor further restricts the maximum length of a line. Add to this the long lines being (bad) big capacitors that constantly have to be charged and discharged. This makes very long AC lines not a viable option.
Long distance high power lines are preferred DC lines. And very long distance must be DC, because the would be no power coming out of the end of the line.
It depends, what you call long distance. Above several hundred km the technical burden / cost of generating high voltage DC and regenerating AC from that is less than the losses on the AC line. And if you connect grids, you avoid possible long distance harmonics.
Even middle distance AC networks are an art to prevent blackouts from harmonics. See the Spanish blackout this year, that started with failing production capacity and then two generators in different locations going into a pendulum catastrophe, after that all dominoes fell.
"Individual cells already have a control processor of sorts, to handle faults and supply status information to the upstream battery management."
In a big battery pack, each group of cells will be monitored. Often it's per parallel string. It's analog rather than a CPU/MPU of some sort that's doing the monitoring at the lowest levels. Even in cases where there some sort of active control, it's very simplistic.
> but I wonder whether higher frequencies (I think aircraft use 400Hz) and voltages might be more efficient.
ICL kit in the 1970's ran of 400Hz supplies and my first job after school was at a manufacturer who built 50Hz to 400Hz motor-generator sets for them. Plug your 50Hz mains on side and get 400 out the other.
Just don't do it backwards....
I read your comment with interest, as I'd recently ordered parts for a new PC including 32GB (2 x 16GB) of Corsair DDR5 RAM, which cost me £109.99 inc VAT when I ordered it on 12 October. (Even that didn't seem all *that* cheap and I assumed some AI-demand-driven price increase had already been factored in. (*))
I checked again just now and, just one month later to the day, the price for the *exact* same SKU/spec has gone from £109.99 to £173.99...!!!
I'd also mentioned to my boss just a couple of days ago that the price we were paying for completely bog-standard Kingston SATA SSDs had jumped from (e.g.) circa £40 to £50 for a 960GB model on our most recent order.
(Not to mention that this is on the UK market so, at the risk of stating the obvious, Trump's US tariffs don't, or shouldn't, have an effect on prices here).
(*) I don't normally keep up-to-date with RAM prices. However, in late 2011 (i.e. fourteen years ago) I was already able to buy 16GB of DDR3 for £75 inc VAT, or around £130 in todays' prices. Admittedly that was quite a lot back then and I did so purely *because* RAM was cheap at the time and I'd wanted to take advantage of that while I could. But if RAM prices had continued falling and capacities increasing at the same rate they were 15-25 years ago, we'd expect to be getting a lot more- and a lot cheaper- than 32GB for around £100. And while it obviously wasn't going to continue at that rate forever, that didn't seem too impressive. But it looks like a bargain compared to the price a month later!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
