Myrient, a 390TB video game preservation archive, is shutting down due to skyrocketing RAM, SSD, and hard drive prices driven by AI data center demand. This is a perfect example of AI's hidden costs - and cultural preservation is getting squeezed out.
The technology is impressive. The question is whether anyone asked about the tradeoffs.
Myrient was one of the largest online video game archives, preserving decades of gaming history from consoles and computers that are no longer manufactured. We're talking about 390 terabytes of ROMs, disc images, and game files - cultural artifacts that would otherwise be lost.
Now it's closing because the operators can no longer afford the hardware. RAM, SSD, and hard drive prices have spiked as AI companies vacuum up storage capacity for training data and inference workloads. What was affordable last year is prohibitively expensive today.
This is what happens when entire markets get distorted by a single technology boom. AI companies are competing to build bigger training datasets and larger model deployments. That requires massive amounts of storage - not just for the models themselves, but for the petabytes of training data and the infrastructure to serve billions of inference requests.
The result? Storage prices increase across the board, and projects that can't match AI companies' willingness to pay get priced out. Digital preservation efforts, scientific research projects, and independent archives can't compete with Microsoft and Google's procurement budgets.
Myrient's closure won't make headlines the way a new AI model release does. But it represents a real loss. Video game preservation is already precarious - many games from the 1980s and 1990s exist only in private collections or archives like Myrient. When those archives disappear, so does access to gaming history.
The irony is that AI companies claim they're preserving and advancing human knowledge. OpenAI trained its models on massive text datasets. Google scanned millions of books. They positioned this as cultural preservation through technology.
But actual preservation work - the kind that doesn't generate revenue or train models - is getting crowded out by the same AI boom. We're not just talking about electricity or chip shortages. The entire storage market is being reshaped, and cultural preservation is collateral damage.
I've talked to people running digital archives and scientific data repositories. They're all seeing the same cost pressures. Storage that used to be a minor line item in their budgets is now a major expense. Some are reducing redundancy to cut costs, which makes their archives more vulnerable to data loss. Others are shutting down entirely.
This is part of a broader pattern where AI infrastructure deployment is creating shortages and price spikes in adjacent markets. We saw it first with GPUs, then power consumption, now storage. Each time, projects that can't justify AI-level spending get squeezed out.
The video game industry generates billions in revenue, but preservation isn't profitable. Commercial game companies rarely maintain their back catalogs - it's enthusiasts and archivists who do this work. When they can't afford the hardware, decades of gaming history becomes inaccessible.
Some will argue this is just how markets work. High demand drives up prices, and resources flow to their most valuable uses. But "valuable" is defined by willingness to pay, not cultural importance. AI training might be worth more to Microsoft than game preservation is to archivists, but that doesn't mean society benefits from losing the latter.
The technology behind large language models is genuinely impressive. Inference at scale requires massive storage infrastructure. But we're making tradeoffs, even if we're not explicitly choosing to make them.
Myrient's closure is a canary in the coal mine. If storage costs keep rising, other preservation projects will follow. We'll have increasingly powerful AI systems trained on cultural data from the past, while simultaneously losing the ability to preserve that culture for the future.
That's not a tradeoff we consciously decided to make. It's just what happens when one technology boom distorts an entire market. The question is whether we're okay with that.
