Google to buy power from fusion energy startup Commonwealth - if they can ever make it work Google has agreed to purchase 200 megawatts of fusion energy from Commonwealth Fusion Systems (CFS). That's assuming, of course, the Massachusetts-based startup can actually get the miniaturized sun to make more power than it consumes, something even the Chocolate Factory admits is a bit of a "moonshot." CFS's ARC fusion power …
"is expected to produce about 400 megawatts of energy when fully operational sometime in the "early 2030s." "
It is expected that "fully operational" fusion energy will start a revolution like coal + steam did with industrialization two and a half century ago. We are literally unable to imagine what this plentiful power will bring us.
Powering AI will be the least useful application for which we can use the plenty of fusion power. But if it helps financing the development, so be it.
Fusion will be hugely important if and when it happens, but it sure as f*** won't be ready and powering Google's data centres by the "early 2030s" and no-one involved believes that.
We're still at the experimental stage. We don't have anything remotely resembling even a prototype for something that would be remotely practical for real-world power generation.
Even if they can consistently get more power out than put in for a brief period in the lab, transforming that into a usable commercial design is a massive task in itself.
Fusion has been 15-20 years away for the past seventy years. It will probably happen, but no-one with half a brain would bet their life- or business- on that being the case with certainty by (say) 2040, let alone 2035.
And it will always be 15-20 years away until someone makes a massive breakthrough.
On the other hand we already have a fusion power source, its 93 million miles away - we just need to use it better. Imaging if the investment into fusion had been spent on better solar systems (systems in this case including storing and distributing that energy)
Yep,
not a single datacenter / AI slinger puts down the full amount to plan for, get approved and invest in a current proven nuclear fission power plant. They either:
* pay to reopen a mothballed one on the cheap.
* pay to extend an end-of-life existing one on the cheap.
* invest now small amounts compared to what the few new nuclear power plants build currently in the West cost in fussion or SMR research with a promisse to buy more when (rather then the more accurate if) the prototype (if there even is one today) goes in production.
* invest even smaller pocket change / advertising money now in startups claiming they invest in fussion power as if they were all in on it.
Meanwhile, they'll use any bit of electricity they can get their hands on to "produce" their AI in "AI mega factories". Who cares if that electricity comes from coal, gas, nuclear, renewables or having households and retirement homes power down airconditioners during heatwaves... it's not as if that would get people killed /s.
Google knows only too well that the project is unlikely to succeed in time, but it doesn't really matter, because the investment is really just an accounting trick: an expense that can be offset against tax.
As for the technology itself: while all the research is good, the LLNL's result is not the closest we've come: stability of reaction and being able to keep it going is more important than the "more in than out" argument and I think ITER's precursor may still hold the record for this. Either that or other experiments are marginally better but of the same magnitude: I think there was a recent announcement from China, the Oxford tokamak was doing good work, and the stellarator in Greifswald has done some interesting stuff. But all these systems are nasty state-sponsored liberal conspiracies which don't give freedom-loving VCs a seat at the table!
Even when they reach more *net* energy out then in (and it all be referred to either primary energy or all refered to electric energy because 1 MWh electricity in and 1.1 MWh heat out is a highly unfair comparison) AND have it stable AND keep it going then it still is going to be very far from commercially viable. Such first prototype may reliable produce electricy but at a staggering cost per kWh. Then it still needs many orders of magnitude in cost reduction by reducing the cost of the plant and increasing net output per plant. That last phase might take many more decades.
To make a not entirely fair comparison: the Curie family already achieved net output of fission power with their first discovery, albeit in a very inefficient way.
Very valid points, but given the potential returns, I think continued investment a good idea. (I could add that nuclear fission is still terribly expensive per kWh if you account for all associated costs).
If, and it's also valid to sceptical about a project that continues to overshoot budgets and deadlines, ITER is able both exothermic and stable, then we'll have opened the door for the kind of optimisations that commercial strategies excel at, like discovering you can use plasticine instead neodymium for magnets and silk stockings are excellent at stopping high energy neutrons… ;-)
In another flawed comparison, this time with the space race, let's put us in the early 1960s…
"That last phase might take many more decades."
There was almost a century between Newcoms first commercial piston steem engine and the first steam loc.
The industrial revolutionized the world, but that too took a while to take off.
It takes time to learn to simulate the center of the sun, actually, at 150M k, much hotter. So what?
Cheap power can always be sold. All this says to me is that Google is willing to pay now for first dibs on that energy down the road to the exclusion of better uses for it.
I just picked up 1kW (nameplate) of solar panels a couple of days ago for $120. That's all I could fit in the car so I'm hoping they'll have some more later and I'm guessing they will looking at their yard. They happen to be some of the smallest ones they have so I expect they won't sell as fast as some of the others. My problem is that I can't fit any wider than around 800mm in the car and hiring a truck would negate any small cost/watt savings to fetch back the higher rated panels. The panels are going to wind up the cheapest component of my system. The swamp cooler has been on solar this season (off-grid) and it's saving me a ton. To test a couple of the new panels, I charged up 3 of my drone batteries and they worked great. Cheap power means I can spend more money on things other than utility bills. By keeping my eyes out for deals, it's not costing me a mint to implement. If supplies of grid leccy gets limited by datacenters, I'll be in reasonable shape to have the HVAC going and food kept chilled/frozen. I hope to pick up some 24v 5kWh batteries by next month for storage.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
