In case you missed it, AI is apparently the next great technological revolution. You’d be forgiven for not noticing, because outside of deep-fake scams, eye-watering energy bills, terrible adverts, polluted water supplies, and chatbots confidently making things up, the actual value delivered so far has been slim.
That hasn’t stopped the tech industry from becoming completely obsessed.
Right now, the biggest winners in the AI gold rush aren’t the companies selling “intelligence.” They’re the ones making the hardware that powers it. More specifically, the [GPUs](https://amzn.to/45fV9Cd) used to run large models, and the [RAM](https://amzn.to/4qhb5we) that those GPUs need in huge quantities.
And that’s where things start to get ugly.
The situation has become so extreme that OpenAI, the company behind ChatGPT, has reportedly signalled its intention to buy around 40% of the world’s raw, unprocessed DRAM wafer output through 2029. That’s not finished memory. It’s RAM that still needs to be processed before it can be used at all.
Even more absurd, there’s growing suspicion that many completed AI GPUs are simply sitting in warehouses. Not because they aren’t needed, but because the data centres meant to house them either aren’t finished or haven’t even been built yet.
In other words, the industry is hoarding components on a massive scale, just in case the AI dream finally pays off.
So far, it hasn’t.
For all the talk of transformation, AI remains a money pit for most of the companies deploying it. Even OpenAI, one of the most prominent and most visible players on the planet, is reportedly losing around $12 billion per quarter.
That’s not a temporary dip. That’s sustained, industrial-scale cash burn.
Despite this, the spending continues. More GPUs. More RAM. More power. More infrastructure. All are chasing a future payoff that may never arrive.
The consequences are already starting to show up where you least want them: consumer hardware.
Emulation handheld maker Retroid has been the first to publicly ring the alarm bell, announcing that it’s ending Retroid Pocket 6 pre-order pricing early due to RAM shortages. That’s not a marketing gimmick. It’s a warning.
And it won’t stop there.
Smartphones could slide back toward 4GB of RAM as manufacturers struggle to source affordable memory. PCs, laptops, and even consoles will face the same pressure. Either onboard RAM drops, or prices go up to keep specs where they are now.
There’s no clever workaround. Memory shortages don’t magically fix themselves.
The end result is painfully simple. Regular consumers will pay more, or get less, to subsidise an AI boom that most of them didn’t ask for and don’t particularly benefit from.
Lower-spec handhelds. Pricier phones. Compromised computers. All so data centres can generate inaccurate chat responses and churn out low-effort content.
Unless you genuinely enjoy watching endlessly recycled AI slop, there’s minimal upside here.
Looking ahead to 2026 and beyond, expect an unwelcome trend. Any emulation handheld you buy is increasingly likely to be held back by limited RAM or pushed into a higher price bracket. Not because the technology requires it, but because the supply chain has been hijacked by AI speculation.
The industry has gone all in on a technology that hasn’t proven its worth, and everyone else is being dragged along for the ride.
Whether we like it or not, we’re paying the price.
That hasn’t stopped the tech industry from becoming completely obsessed.
Right now, the biggest winners in the AI gold rush aren’t the companies selling “intelligence.” They’re the ones making the hardware that powers it. More specifically, the [GPUs](https://amzn.to/45fV9Cd) used to run large models, and the [RAM](https://amzn.to/4qhb5we) that those GPUs need in huge quantities.
And that’s where things start to get ugly.
A scramble for memory
The situation has become so extreme that OpenAI, the company behind ChatGPT, has reportedly signalled its intention to buy around 40% of the world’s raw, unprocessed DRAM wafer output through 2029. That’s not finished memory. It’s RAM that still needs to be processed before it can be used at all.
Even more absurd, there’s growing suspicion that many completed AI GPUs are simply sitting in warehouses. Not because they aren’t needed, but because the data centres meant to house them either aren’t finished or haven’t even been built yet.
In other words, the industry is hoarding components on a massive scale, just in case the AI dream finally pays off.
So far, it hasn’t.
Burning cash at record speed
For all the talk of transformation, AI remains a money pit for most of the companies deploying it. Even OpenAI, one of the most prominent and most visible players on the planet, is reportedly losing around $12 billion per quarter.
That’s not a temporary dip. That’s sustained, industrial-scale cash burn.
Despite this, the spending continues. More GPUs. More RAM. More power. More infrastructure. All are chasing a future payoff that may never arrive.
Consumers get the bill
The consequences are already starting to show up where you least want them: consumer hardware.
Emulation handheld maker Retroid has been the first to publicly ring the alarm bell, announcing that it’s ending Retroid Pocket 6 pre-order pricing early due to RAM shortages. That’s not a marketing gimmick. It’s a warning.
And it won’t stop there.
Smartphones could slide back toward 4GB of RAM as manufacturers struggle to source affordable memory. PCs, laptops, and even consoles will face the same pressure. Either onboard RAM drops, or prices go up to keep specs where they are now.
There’s no clever workaround. Memory shortages don’t magically fix themselves.
A lose-lose situation
The end result is painfully simple. Regular consumers will pay more, or get less, to subsidise an AI boom that most of them didn’t ask for and don’t particularly benefit from.
Lower-spec handhelds. Pricier phones. Compromised computers. All so data centres can generate inaccurate chat responses and churn out low-effort content.
Unless you genuinely enjoy watching endlessly recycled AI slop, there’s minimal upside here.
What this means going forward
Looking ahead to 2026 and beyond, expect an unwelcome trend. Any emulation handheld you buy is increasingly likely to be held back by limited RAM or pushed into a higher price bracket. Not because the technology requires it, but because the supply chain has been hijacked by AI speculation.
The industry has gone all in on a technology that hasn’t proven its worth, and everyone else is being dragged along for the ride.
Whether we like it or not, we’re paying the price.