Computers are More Expensive than Ever: Why the Entire PC Market is Up in the Air

Corporate greed.

Of course, it is more complicated than that, but everything eventually traces back to giant technology corporations attempting to increase profit and shareholder revenue. The largest direct cause of this price spike is AI, as RAM especially is being bought up millions of dollars at a time for use in AI data centers. This exacerbated a preexisting shortage of RAM and other computer parts caused by the COVID-19 pandemic, a shortage the industry was unable to recover from prior to the sudden AI boom. The result is that RAM prices have gotten three to five times more expensive over the past few years.

Average Price of DDR4-3200 2x8GB RAM over the past 18 months, courtesy of PCPartPicker.com

Let’s break all this down: What is RAM? Why do PCs need it? Why do AI data centers need it? Why are all these data centers being built in the first place? What is the end goal for all these giant technology companies?

It is a lot to cover, but I promise I’ll be concise.

The PC keeps running, running and running, running…

RAM stands for Random-Access Memory, and it is essentially your computer’s short-term memory. Think of your computer as a retro office with a desk surrounded by filing cabinets. You would use the desk as your working, functional space, while the filing cabinets are more long-term organization and storage. In this analogy, the desk is your computer’s RAM, and the filing cabinets are your PC’s hard drives. The more RAM, the larger the desk, meaning your PC can run more operations at the same time. Nearly every operation on a computer uses RAM in some way, so without enough RAM, your PC will run slower and crash more often. In the office analogy, having larger and nicer filing cabinets (better hard drives or SSDs), will not help if you are trying to operate with a desk that’s way too small (not having enough RAM).

Disconnect from all Intellect

If you have read any of my other articles, you may be under the impression that I am anti artificial intelligence. That is entirely accurate, but I have so many gripes with modern AI that I cannot get into it in this article. Suffice to say that generative AI is a very useful tool, but it has issues, both practical and ethical in nature. An issue that spans both those categories are data centers.

An artificial intelligence data center is essentially a building filled to the brim with computers specifically designed for AI processing, including both training and answering prompts. The devices in these data centers use some of the best and most powerful computer components in the world, and each of these centers use thousands upon thousands of computers per location. The result is that AI data centers use an absurd amount of power and water. Power is obviously necessary to, well, power the facility, but water is also needed to cool the devices and keep the location operational. However, the focus here is on the material demands of these data centers, and how that is impacting the supply of commercial RAM.

I just can’t get enough

Major tech companies are estimated to spend $650 billion on AI data centers in 2026, with well over a hundred announced, in production, or already built in just the United States. That’s a lot of money, and a ridiculous amount of hardware that will need to be produced. Now, circling back to RAM, the computers in these data centers will require a lot of RAM, way more than the average consumer device. In fact, AI data centers primarily utilize a more advanced form of RAM known as High Bandwidth Memory, aka HBM. That’s good, right? Use this advanced, expensive stuff for the data centers, and leave everything else to the consumer. Sadly, it’s not that simple.

HBM memory on an AMD Radeon R9 Nano graphics card's GPU package

There are not enough places in the world making computer parts. Specifically in this case, the semiconductor memory chips (more commonly known as microchips) needed for RAM and the RAM sticks themselves. Roughly 60% of the world’s microchips are produced in one of the most politically fragile locations in the world: Taiwan. The situation and history surrounding the legitimacy of Taiwan as an independent nation separate from China is way beyond the scope of this article, but know that China has laid claim to Taiwan for decades, and is always threatening to make more drastic moves to officially take over the island for good. Such an outcome would give the People’s Republic of China such a stranglehold on international computing that some politicians and strategists in the west suggest destroying the facilities in Taiwan in the event of a proper Chinese invasion. Naturally, such an outcome would devastate the global economy indefinitely.

While horrible in theory, the situation between Taiwan and China is not quite so apocalyptic currently. However, the instability of said situation certainly contributes to rising prices globally, though it may also have limited benefits. Many nations, including the United States, are making moves to establish their own production of microchips, which would eventually ease dependence on Taiwanese production. The bad news is that it will take quite a while to establish these facilities from scratch, so don’t expect relief from that outcome anytime soon.

The more immediate stress on computer prices is on the RAM production side. The vast majority of RAM production stems from just three companies: Samsung, SK Hynix, and Micron. These few producers cannot meet the demand for RAM. So, in order to maximize profit, they must pick and choose what RAM they will produce, and who they will produce it for. These AI companies are buying millions of dollars in RAM at a time for each of these AI data centers. Not only are they purchasing large quantities of RAM, they’re opting for the most expensive variety of RAM (HBM). The result is that the companies making RAM are incentivized to make nearly exclusively HBM RAM for AI companies and their data centers. Now, the less powerful RAM consumers and businesses want and need is barely even being made, and certainly not being made to meet global demand. In fact, Micron has completely abandoned consumer RAM, leaving just two companies to meet the world’s needs.

To make matters worse, even with Micron only focusing on AI, even with Samsung and SK Hynix focusing primarily on AI, it’s not enough for these AI companies. Only the wealthiest tech giants like Google and Microsoft can afford to get the HBM straight from the factories, so the rest of the many, many AI companies today are scrambling to get what they can. They are sweeping up RAM wherever they can get it; HBM and consumer-grade. There is so little left for the consumer that it has become obscenely expensive.

New days are strange, is the world insane?

Artificial intelligence is the reason why RAM and computers in general are so expensive right now, but what is the point? What is the endgame for all these tech giants? Is there any hope for consumer electronics?

The theoretical endgame is what it often is for the wealthiest corporations: maximizing profits by minimizing cost. One of the highest costs for nearly every business is labor, especially white-collar jobs where people are expected to make six figures or more. Robotics automated many jobs in manufacturing and put a lot of blue-collar workers out of work, and these tech giants are hoping that AI can do the same for white-collar workers. Ironically, the blue-collar jobs that robotics cannot easily replace (think plumbers, electricians, mechanics, etc.) are some of the few jobs relatively safe from AI automation. However, if you are a worried white-collar worker or a business owner seeing dollar signs, AI is not there yet, and there are a lot of signs that it won’t get there for a long time.

Generative AI, which is what most people mean when discussing “AI” these days, is prone to mistakes, hallucinations, and just plain low-quality work. There was a notion a few years back that AI was only bound to get better, and that was true to an extent, but the AI companies have hit a wall. Generative AI requires a ton of training data, and frankly, they are running out. Lately, some AI companies have fallen back on training AI data models on AI-generated content, which is a circular arrangement that will inevitably eat AI alive. AI trained on AI will reach a point where it is no longer useful to us. The path to true artificial intelligence rivaling human intelligence does not lie through generative AI technology.

Don’t hear what I’m not saying; generative AI is not useless. It is a powerful tool with a wide range of applications in a lot of different industries, but it is just that: a tool.

Selfishness got us following the wrong direction

This too will pass. This AI boom is largely unsustainable. The truth is that most of the money flowing into AI can be traced back to a single industry: AI. That’s right; nearly every dollar spent on AI is either from another AI company, or a technology company funding AI so the AI companies can afford their products. Absent from this story up to this point is the largest offender: Nvidia. Why? They create GPUs, which are needed for PCs, but primarily for gaming computers (and other similarly powerful PCs), but RAM is more universal. GPUs are needed for these AI data centers too, but they are incredibly and increasingly expensive. So, Nvidia pumps money into these AI companies so they can afford to buy Nvidia products.

The problem? People are not paying for AI. Consistently, ever since ChatGPT exploded onto the scene a few years ago, the average consumer has proven that while they like to use these chatbots, they won’t actually pay for them. These tech giants may be dreaming of a world where they don’t need to pay humans for labor anymore, but they are truly just blowing up a massive bubble. Time will tell if it pops, gradually deflates, or consumes us all.