I do quite a bit of research on topics I find interesting. One of the subjects I have been particularly focused on recently is computer pricing forecasts.
Tonight I did my regular thing and found something worth sharing.
There is clearly a broad consensus among manufacturers of computer components that prices will stay elevated for the remainder of the decade. With some players expecting high prices for the next 10 years. Explicitly because of the AI infrastructure growth cycle.
This “broad consensus” is coming from Micron, Asus, Nvidia, SK Hynix and others. Who are saying the sub-$500 laptop market may completely disappear before 2028.
What I am wondering is, how will prolonged high prices impact the homelab sphere?
I have put off upgrades already, but can I afford to keep kicking the can down the road?
If that does turn out to be true than the homelab will become even more niche, only for those who can afford RAM prices. That’s the bottleneck, I’m not gonna buy anymore kit (new or old) until prices fall.
I’ve got a Lenovo desktop PC (ebay) I fitted with 64GB of ram running Proxmox, would not be able to afford the very same same thing right now. My Lenovo laptop is 10 years old also with 64GB ram, it will have to do what I need.
In the UK, I have noticed, refurbishers, are selling laptops with meager amounts of ram. I have a feeling that Raspberry Pi’s are now costing more, that was my gateway drug to setting up my home network, so that might be a barrier to entry for some with limited funds.
My home network is 1GB, I can’t see myself moving towards 10G at all, the switches are still expensive, it will take ages before 1GB kit becomes obsolete at home.
There is no going back anymore. By 2036 everything is going to be cloud and ai based. The times of buying computer components and building your own stuff are long gone. Web will be DRM protected, everything will require account and personal identification because we have to “protect the children”. These astronomic prices we have today are just introduction to a very grim and gloomy future. If you think that ai bubble is going to pop and everything will get back to normal, you are delusional. And trust me when i say this. There is nothing i would like more than you guys quoting this post in the future and showing me how wrong i was.
The AI model is unsustainable as it stands now. None of these AI companies are making money. They are all over valued. The entire stock market is propped up by these artificial valuations. This can only go on for so long. At some point investors are going to need a return on their money.
In terms of home labs, you can still get some decent prices on old hardware on places like eBay. You can always pivot and start using some hosting providers like OVH to rent servers or VPS to deploy your software without having to make that initial investment.
Thats because AI models of today are not very efficient. We brute force it with massive power and water consumption. Of course something like that can not be sustained for a long time and it would come to an end. However, before it comes to an end, it will spawn a new and more efficient models that will use much less power. Dont forget ai hardware advancements as well. It doesnt take a genius to see where ai is going to be in the future. Just compare first version of ChatGPT to Mythos. Its advancing very fast. You are just coping hard.
Edit: Just to be clear. Im not talking about small ai companies that are promoting ai garbage. Im talking about big ai players like Palantir, Nvidia, AMD, Google, Microsoft and their impact on your life.
Just take a look at Micron, SK Hynix, Nvidia…etc. There is a good reason why they are all leaving consumer market. We are dying breed. You have few more years left until all the hardware we use today becomes completely obsolete and non compliant with future networks and standards. Worst of all, people are only focusing on AI and they are blaming AI for everything. They are obsessed with it so much, that they fail to see other problems we have today. Take a look at open source community and all the political crap happening over there. And i wont even go into real politics.
I don’t think the situation is as doom and gloom as you predict. Things will work out eventually it just going to take time until the dust settles. In the meantime you have to pivot and figure out to do things differently.
It is exactly doom and gloom and its only going to get worse. This horrible reality is unfolding right in front of you and you are either unable or unwilling to see it. Big players dont care about you anymore. They care for hyperscalers only because thats where the money is. Home lab basement dwellers and gamers are no longer profitable. Everything is going to be on the ai powered clouds. You will own nothing.
Thats just coping and hoping things will get better on its own based on absolutely nothing. Hope dies last.
I already do everything differently. And have been doing it for the last 35 years since i started. You learn a lot in that time span.
The high prices probably won’t really come back down unless a new manufacturer comes in to replace the void in the market. Not many sellers bring the prices back down
But there’s not all doom and gloom. Someone will want to start making money (maybe Texas Instruments?) by creating consumer level computer hardware that others left (Micron).
bretbernhoft:
I have put off upgrades already, but can I afford to keep kicking the can down the road?
I would plan out a budget for what you want to build. Then think about if you can scrapyard wars (second hand markets) the hardware or shrink the hardware. I really think most homelabs only need a mini PC- maybe two for learning High Availability- to do most of the projects that are posted on the internet. And then again, most of us “really” don’t need/use all of these projects (e.g., Huginn, n8n, ollama, etc.). But the homelab is a hobby for learning and creating- so to each their own
For example, I use an Asus PN53 with a mobile Ryzen 7 7735H and 64GB RAM (because this is the most important for VMs) and 3TB of NVME (which I only use about 400 GB of) as my main homelab node. Before that I used an old Home Theater PC I built in 2011 until 2023 that used a Core i5-2400 and only 16GB RAM with a 512GB NVME and 2TB of spinning disks. I would probably still be fine on the HTPC except the spinning disks started failing after 15 years.
Even repurposing consumer level NAS devices that allow for installing secondary operating systems can fit the bill for running home docker/podman services. See Buffalostation, Asustor, Terramaster, Ugreen, etc.).
The fancy youtube homelabs might be a disservice to the homelab community. I suggest specifically watching Hardware Haven for ideas and examples of a homelab running on older/smaller/affordable hardware versus something needing dozens of TB of storage and many dozens of AI token horsepower.
Not a feeling, fact! RPi5 with 16gb of ram is now $300usd RPi4 8gb is $120+ and even the little RPi3b+ with a whopping 1GB of ram is $55. The RPi3 is in a bad place because it is DDR2 and everyone quit making that chip (or about to quit).
As far as home lab, I’m still trying to muddle along, but most of the good stuff has all the ram pulled out, easier to sell and far easier to ship. I’m looking but not finding things to buy. Also thinking I need to just get out of the tech area for my last few years, it just seems dreadful lately.
Think about this for a moment… What will the financial crash look like when AI comes crashing down? I bet it looks a lot worse than the .com crash that everyone in finance seems to have forgotten.
