AI Capex Spending is still going strong.
1 Like
My exposure is relatively small. As I’ve noted in other threads, i think even if AI pulls back and projects fail, I think previously suppressed consumer purchases of memory products keeps these companies in a good spot.
They are still investing enormous sums in AI Capex.
Its so big that is actually surpassed US Personal Consumption in terms of size.
I agree that AI companies will eventually fail, but thats not going to be in 2026. I still see growth for the next 6-12 months.
No doubt. It makes one wonder about this Iran excursion thing. A lot of Middle East oil money is pouring in.
An unprecedented influx of foreign capital is also backing U.S. tech infrastructure. Countries of concern aside, the U.S. has received trillions in investment commitments from governments in the Middle East—such as Saudi Arabia and the United Arab Emirates —along with Japan and India to finance semiconductors, cloud computing, and energy infrastructure. [1]
Please don’t totally abandon simple web searches in favor of things like ChatGPT. It’s pretty bad for the environment and not everything needs that level of analysis!
3 Likes
Of course, Google searches now include call outs to Gemini.
Sometimes the Gemini result is even somewhat useful.
2 Likes
I’ve just started putting a tiny bit into the AI boom and I’ll probably add a bit more in. No telling when the top will be in, but it feels like the ride to the top has more room to go yet…
1 Like
Sounds like some famous last words there. Bull markets end when the last skeptics buy in.
3 Likes
This has been part of my reasoning, though. I could do DuckDuckGo I suppose, but that’s just Bing with a wrapper. Google is becoming increasingly useless. It automatically spits Gemini results of dubious accuracy at you (I assume the automatic prompts use low-effort models?) and otherwise is heavily SEO’d away from whatever you actually wanted.
To be clear, I don’t set Claude on Opus 4.8 on Max Thinking mode to ask for a recommendation on a restaurant or what the forecast will be (the latter being an especially horrid example I keep seeing people do.)
However, I was on vacation and did ask things like, “It’s 6:30 AM right now and I’m located at XYZ. I’d like to hit a local bakery/coffeeshop, then walk toward X to be there by around 8 AM when it opens. Along the way, I’d like to see X. Build a path for that - include any viewing attractions within a reasonable detour that make sense at this time.”
It’s often wrong, but is a great launching point for things that aren’t super important to be precisely correct like that.
1 Like
I have 24k to drop into an ETF and I am struggling on where to put it that doesnt seem overvalued. I might split it between healthcare and financials as they are both down YTD.
We’re having theoretical work discussions around how to prevent people spending potential company tokens on such searches.
4 Likes
But allow it for work-related dinners. Simple, burn some tokens to have Claude review your company’s use of tokens.
1 Like
Throw $4k at DRAM. It gets you better access to some of the difficult to invest in Korean tech manufacturers.
I would consider Korean tech to be overvalued at this point.
Depends.
SK is an interesting case.
Domestic investors are driving the tech shares up because they have made it easier to invest (retail) and pension funds have increased their target allocations for domestic investment.
Foreign investors are currently net sellers but the domestic demand is so strong it is driving the main ones (SK Hynix, Samsung) upwards.
The non-tech shares in SK are meh. No growth there.
1 Like
You got me onto Chinese ETFs, thanks. I picked up some KARS (Chinese EVs). They’ve gone up 12% in the last 4 months.
Let’s hope so. I’m up 6K this week on my 100K investment (about 5% of my portfolio). That a few dollars on DRAM is all I’m going to play with.
I posit we are not at the peak just yet…
https://en.sedaily.com/news/2026/06/03/goldman-sachs-raises-kospi-target-to-12000-on-chip-cycle
Hearing more and more stories about AI costs going up and above the savings they are perceived to be generating. People are burning tokens on bad processes which is more expensive than paying a human to fix it. It will make your productive employees much more productive, but your average employee much more expensive. Gotta think this is going to go South here pretty quickly when the economics don’t pan out.
2 Likes
Mostly agree that the long-term economics are still net negative at this point.
They are burning serious capital ($ Trillions) building the infrastructure, as well as subsidising the computing power. The user side is still happy to go along with this at current prices, even if the productivity gains are not quite clear just yet.
Main aim in my view is to create some degree of dependency to a specific vendor, that way they will be able to reduce the subsidy/increase their prices…in order become profitable long-term. Thats always been their standard playbook.
We are seeing some vendors go down that path (smaller players) but the larger players are still spending big on capex and computing subsidies.
2027 is when I think the capital crunch will come because at that point debt raising will be expensive, and an equity raise via stock would be a problem for existing shareholders (who wouldnt want to be diluted).