How much NVDA did we all buy today?

[johnson86-1]

I'm thrilled that MS is getting in on an emerging industry. Elon has one going up in Memphis as well. What worries me is the energy consumption part hasn't been adequately communicated and how it will impact regular consumers. Allegedly there's capacity but MLGW asked consumers to scale back during the cold spell last week when I assume the source generation was running at peak. What happens when these facilities go live? Don't get me wrong - there is a world where it all works out just fine and we will get there but it feels like we intend to just wing it given how large the investments are. Am I overreacting or is this a legit concern?

I'm not familiar with MLGW's assets, but I would be very surprised if they are providing any capacity to a data center. They may technically be providing service, but there would likely be a contract with an utility with significant generating assets that addresses capacity/demand and pricing and if MLGW is involved, it would be getting a small markup on the power without taking a lot of the risk of investing in capacity. Just a guess on my part though based on how most municipal utilities operate.

As far as how it impacts regular customers, it probably won't have a big impact. Not sure what their load profile looks like, but these things use so much energy, I would assume there is going to be capacity more or less dedicated to them. If the utility is long enough to serve them without new investment, it should put downward pressure on prices of other consumers at the expense of sucking up the excess capacity. Most people would probably consider that a win as it's expensive to pay for capacity you don't use.

 

[johnson86-1]

My rule of thumb for any new technology has always been I want to be in the 2nd quartile to adopt it. Let the early adopters spend all the money and make all the mistakes and then get the kinks ironed out, then I'll get on board with a better and cheaper product.

I feel like that's a rule that's more applicable when it's unclear how a new technology is going to be monetized. It's not clear how all these LLM chatbots are going to make money yet. So I wouldn't be comfortable betting big on any of them, because a new player may be the one that actually figures out the commercial side.

But unless something changes, they are going to use the tech provided by NVDA I think? If the Chinese AI can really operate without NVDA type tech, then yes, that's a game changer. If they are just using reverse engineered versions of NVDA tech, that's a problem that's going to cut into margins, but not make them obsolete or anything.

The best case scenario for NVDA I think is that the Chinese really did make a much cheaper AI that is still dependent on products NVDA makes. THen you'll probably have much more widespread adoption of AI and a higher demand for NVDA products.

ETA: If it's not clear, I know nothing and am talking out of my *** on this subject and all of the above is more of a question for somebody that knows more than I do than it is a statement of fact or even opinion. That's just how I am thinking about it but no clue if that's close to correct.

 

[ckDOG]

I'm not familiar with MLGW's assets, but I would be very surprised if they are providing any capacity to a data center. They may technically be providing service, but there would likely be a contract with an utility with significant generating assets that addresses capacity/demand and pricing and if MLGW is involved, it would be getting a small markup on the power without taking a lot of the risk of investing in capacity. Just a guess on my part though based on how most municipal utilities operate.

As far as how it impacts regular customers, it probably won't have a big impact. Not sure what their load profile looks like, but these things use so much energy, I would assume there is going to be capacity more or less dedicated to them. If the utility is long enough to serve them without new investment, it should put downward pressure on prices of other consumers at the expense of sucking up the excess capacity. Most people would probably consider that a win as it's expensive to pay for capacity you don't use.

MLGW sources from TVA. I believe the data center is running off temporary local gas turbines but eventually will be on the TVA (via MLGW) platform. I agree with the idea that excess capacity needs to be used for efficiency purposes but the projected 300 MW draw is pretty large. TVA was running pretty thin a couple years ago but that may have been a perfect storm situation when a largish plant went down during a high consumption time.

Honestly if MLGW wasn't involved at all I'd feel better. This is the same utility that built up a many thousands of miles and hundreds of millions of dollars of tree trimming backlog over the last decade. There's so much wood hanging over key infrastructure in this city that we get power outages during summer breezes. Reminds me of the random outages we would get when I worked in Brazil.

Side note but related: as I'm googling some of this stuff, I've noticed their AI tool seems to have improved significantly as of late. I used to ignore it but now find it very useful. It provides some really nice topic summaries with handy links.