I clearly did a bad bet if we look at it in the short therm. $POWL (+6,05 %) is 30% down in my portfolio, and like 40% in the last month, just as $STRL (+8,21 %) .

They both are AI infrastructure related stocks, so I understand the “why” they’re down, specially on the $STRL (+8,21 %) case as they were clearly over their historical valuations… but this fall from $POWL (+6,05 %) is not logic unless I am missing something.

Its valuation is almost the lowest in the 3 last years, (even low for their long therm historical average) while beating on everything. The company has no debt and still a correct growth profile. Is this just a momentum thing or I don’t know something?

I’m in between the feeling that I should double down, or cut my losses.

What am I missing?

Increasing my position in $POWL (+6,05 %)

Just in for a month now, this selloff deleted all my gains (35% at some point) and put me in negative territory.

Why am I adding more?

The price is 5% down from when I entered, and I belive it was already a good entrance point. The earnings and guidance made it jump higher and I belive the story behind its growth is still alive.

Why did it sell off 20% with the DeepSeek news? Well, $POWL (+6,05 %) performance is linked to data centers and energy consumption, as they provide power solutions, and with capex on that being questioned, it fell like a rock, just like $CLS (+1,67 %) or $STRL (+8,21 %) .

I belive that we have seen an overreaction on the market to sensacionalist news, without deep understanding of what is really going on (I also belive that some stocks are priced to perfection and that a selloff was necessary).

So what is going on?

DeepSeek trained a model and claimed it cost 6M to train, but how did they calculate that cost? By implying a cost of RENTING H800 NVIDIA chips (the “cheaper but not so cheap chips” from Nvidia that can be sold in China). So not considering the cost of the infrastructure, salaries, development and building the model itself etc… just the training hours on rented H800 NVIDIA chips. We don’t even know for sure the truth about this costs, this is a Chinese start up we are talking about.

What we do know is that they did achieve good answers from a LLM for a lesser price, as their model uses a different more efficient aproach, that will for sure be copied soon (they’re an open source model) by the other players if they feel like it is the way to go.

So, is this going to affect the spending in AI?

This is a step in the right direction to make AI something global, that will be used in every aspect of our lifes, and the AI infrastructure will be needed to be able to afford all this demand that will be comming. Everything is going to evolve faster than ever, and AI infrastructure will be key to that, even if improvement can be achieved with less computational power, don’t you think that the US and China are just in the beggining of an AI race and that any of them can afford being left behind?

I obviously don’t know, but I’m not selling my positions on the AI infrastructure related stocks, specially the ones with reasonable P/E like $POWL (+6,05 %) and $CLS. (+1,67 %)

Amazing what one single news event can make to the hole market... Is it such a big deal?

Well, an impressive AI comming from China, that is supposedly performing as good as ChatGPT or LLama with way less resourcess (and I say supposedly as no one really knows anything when it is comming from China...). Is this really going to change everything?

My worst performers today are $CLS (+1,67 %) , $STRL (+8,21 %) and $POWL (+6,05 %) all of them linking their exceptional performance this last months to the AI infrastrure. Will they suffer that much?

They're american companies, with mainly an american business, in a country that has put its focus on compiting with China for the AI dominance. China has an AI that can work as ChatGPT with less resources, so what...? Is USA going to stop investing in it for that reason? Or maybe the opposite?...

Does what DeepSeek has achieved, mean that the Big Tech are investing too much in "super chips" when they could do the same with less investing and less data centers? Maybe... But will they stop the investments in data centers? If something like what DeepSeek did can be achieved with less powerful chips, what will the Big Tech be able do with the "super chips" of Nvidia? Sky is the limit, and the american Big Tech have the capacity of burning cash trying to reach it, while they keep earning billions.

While I'm sad I lost more than 15% on this 3 stocks valuations and my profits are less nice than they were on Friday, I still see this companies in a very nice shape to keep growing, and the AI market to keep expanding. Even with the fall valuations are not super beautiful (considering the noise), but this might be an oportunity for the ones that didn't invest before in this companies to get a piece of them now.

I'm not going to use this to add more of this stocks to my portfolio at this point, as the 3 of them are at levels of 1 month ago and I already have them in my portfolio, but if this blood bath continues I'll fell I have to... This is the breather that at some point was going to come with the big rally they've been running.

update* bought some $CLS (+1,67 %) 25% down was too tempting for a stock I follow closely