Buddy, Can You Spare a GPU?

November 2, 2026

I wanted to write this blog post earlier in October but got hung up with work last month. For some reason, work has been crazy lately. So Ed Zitron, Paul from Paul's Hardware, Steve Burke from Gamers Nexus and others beat me to this topic, and Vex, a YouTuber, does a great job explaining it too. Just as I was finishing this, Brennan published I Am Relentless and Undaunted. Go read it. It's worth your time.

So I don't have to make this a long post, sorry, just kidding, it will be very long. I have to get this off my chest and vent. I believe there's a conspiracy here, and it deserves some attention.

I'm putting on my tinfoil hat for this one.

I try not to be a conspiracy theorist, but come on. How can you look at what's happening and not think these shortages and high component prices are being deliberately created? At the heart of it, I believe, is Nvidia and its bottom line. As we've learned from Ed Zitron and Business Insider's reporting, Amazon is sitting on warehouses full of GPUs, and by extension, I suspect there's memory and storage equipment sitting around too. Why? Because some of these data centers aren't ready, and others don't have enough electrical power to operate the equipment. Microsoft has acknowledged similar problems and has pulled back from some planned data center projects. So why are these companies buying billions of dollars in GPUs that could be outdated by the time they're installed?

I have a theory.

I believe Nvidia and major memory manufacturers Samsung, SK Hynix, and Micron may be pressuring companies to take delivery of equipment long before their data centers are ready. Ed Zitron makes the same case, calling it channel stuffing.^[1] They're also prioritizing enterprise AI hardware and leaving consumers fighting over whatever manufacturing capacity is left. I suspect something similar is happening with storage manufacturers like Western Digital, Toshiba, and Seagate.

Look at Micron. After nearly 30 years, it discontinued its Crucial consumer memory and storage business in February 2026. Not because consumers stopped buying RAM and SSDs, but because it wanted to concentrate on larger, more profitable customers.^[2] These companies are making deliberate production decisions, and they know the consequences. Consumers are left paying considerably more for RAM, storage, and graphics cards. Apparently, if you want to upgrade your computer, you might have to sell a kidney to afford it. Meanwhile, these enterprise customers can afford to buy the whole body. And we're supposed to believe there's nothing the manufacturers can do about it. Wink, wink.

If a data center isn't coming online for another two years, why does it need billions of dollars in equipment sitting in warehouses today? Why not schedule deliveries six to nine months before it's ready, or even a year if necessary? I understand semiconductor manufacturing is complicated, but surely these companies can plan for enterprise demand while maintaining reasonable consumer production.

Instead, we're watching this enormous circle of financing, hardware purchases, and advance commitments. Nvidia is also investing in AI companies and helping finance infrastructure that will purchase its own products. Are these enterprise customers being told to take delivery now or risk losing their place in line? We're not forcing anybody to buy anything. We're just making sure you'll have the hardware when you eventually need it. Wink, wink.

And before anyone tells me these companies would never manipulate a market, let's remember some history. Remember Archer Daniels Midland, better known as ADM? In the 1990s, ADM and competing manufacturers were caught secretly fixing prices and dividing up sales of lysine, an additive used in animal feed. ADM pleaded guilty and paid a $100 million criminal fine covering both lysine and a separate citric acid conspiracy. An insider working with the FBI helped expose what was happening.^[3] Samsung and Hynix, now SK Hynix, have already pleaded guilty to fixing DRAM prices. Samsung paid a $300 million criminal fine, and Hynix paid $185 million. Executives also went to prison.^[4] So we're supposed to forget that history and accept that today's shortages are entirely unavoidable? Maybe these companies are making independent decisions. Maybe everything is perfectly legal. But I suspect some learned quite a bit from their previous encounters with antitrust regulators.

Nvidia isn't stopping with GPUs. It invested $5 billion in Intel, entered into partnerships across the industry, and has agreed to acquire Hugging Face, one of the biggest platforms for distributing open AI models. The U.S. Justice Department is already asking questions about that deal.^[5] Nvidia says Hugging Face will remain open to competing hardware. I hope that's true, but I have to wonder how much of the AI industry Nvidia intends to influence. Now look at consumer graphics cards. Nvidia already dominates the market, AMD has pulled back from competing at the very high end, and Intel is putting considerable attention into professional AI graphics hardware. Maybe all this is just a coincidence, but consumers are certainly getting fewer choices at the high end, both price-wise and hardware-wise, with the rumored end of the RTX 5090, which Nvidia denies. And it's not just the 5090. The RTX 5080, which launched at $999, now has listings for $2,000 or more, and Nvidia's own store has quietly pulled it too.^[6] Even Intel's Arc Pro B70, originally priced under $1,000, has appeared at prices exceeding $2,000.^[7] AMD cards haven't escaped the increases either.^[8] These price increases are happening at every rung of the ladder. How is the average consumer supposed to afford this stuff anymore?

I wonder whether we're slowly being pushed toward a future where ordinary consumers aren't expected to own powerful computers anymore. Instead, we're expected to rent computing power from somebody else's data center. Just when I thought this couldn't get any more ridiculous, Sam Altman of OpenAI said something recently that really bothered me. During an interview on the Cheeky Pint podcast, Altman argued that it's “quite terrible” for people to have so much of the world's limited DRAM in personal computers, supposedly being used only 5% of the time. He believes cloud data centers can make better use of that memory.^[9] Really, Sam? Paul from Paul's Hardware had the perfect answer in a recent video: “I feel the same way about Sam's bank account. He's not using all that money all the time, so it should definitely be reallocated to people who will make better use of it.” I heard that and almost fell over laughing, because it's a very apt comparison. Gamers Nexus covered the same interview in its latest HW News episode, and Steve Burke summed it up: “Sam is saying the quiet part out loud.”

The CEO of one of the largest AI companies, whose industry is buying enormous amounts of memory and hardware, suggesting that consumers having RAM in their own computers is an inefficient use of resources. Apparently, it's perfectly fine for billions of dollars in GPUs to sit in warehouses waiting for electricity, but heaven forbid I have 32 or 64 GB of RAM in my own computer that isn't being used every second of the day. What arrogance. My computer belongs to me. I bought it with my own money, and I should be able to use it whenever and however I want. The attitude behind his comments shows an utter lack of understanding of ordinary people. It's a Marie Antoinette, “let them eat cake” moment.

Want to run powerful AI? Subscribe to a cloud service. Want to play games? There's GeForce NOW, Xbox Cloud Gaming, and other services. Why sell consumers powerful hardware when companies can sell it to enterprise customers and charge consumers for ongoing access? In other words, a subscription. Or as Brennan put it, we “will be compelled to rent and lease and borrow.” And look, everybody has agreements with everybody else, so obviously we're encouraging competition. Wink, wink.

Even Apple, with its enormous balance sheet and once-great ability to dictate terms to suppliers, is having problems with rising memory and storage costs.^[10] And Samsung? It competes directly with Apple in smartphones and makes many of the components that go into them. You would think Samsung would have a tremendous advantage right now. But it's chasing the same enterprise AI money, even as higher component costs hurt its own consumer products.^[11]

Apparently, even Macs aren't safe from this AI buying frenzy. According to The Information, OpenAI has purchased tens of thousands of Mac minis and Mac Studios for AI workloads, while Anthropic is reportedly renting Mac minis through Amazon. These are computers meant for consumers, independent developers, and small businesses.^[12]

So now even the computers people could use as alternatives to Nvidia hardware are being purchased in enormous quantities by AI companies. They're all riding this cash cow for everything it's worth. But what happens when the AI financial bubble finally bursts?

Let me make something clear. I don't believe AI itself is going to collapse. I believe demand will continue growing, whether that's local AI or cloud and frontier AI, because AI is useful technology. What I think will collapse is the financial house of cards being built around it. Eventually, Wall Street and private investors are going to demand returns that justify the billions, no trillions at this point being spent on GPUs, memory, storage, and data centers. Some companies won't be able to meet their obligations. Others may be acquired, restructured, or go bankrupt.

And what happens to all that expensive enterprise hardware when the spending slows down or data centers start replacing equipment? I hope consumers get to buy it for pennies on the dollar. If manufacturers want to concentrate on HBM and enterprise hardware, maybe we consumers should learn how to adapt that equipment for our own needs. I'm not talking about removing HBM chips and sticking them into ordinary motherboards. I'm talking about buying retired enterprise GPUs with their memory already attached and adapting them for personal use. Some have 80 GB, 192 GB, or more of high-bandwidth memory. Imagine what we could do with that for local AI. Sure, some will need special cooling, adapters, power supplies, and software support. But people are already figuring this stuff out. Look at what Lon Seidman did with this Tesla V100. Maybe equipment these companies consider obsolete will be perfectly good for the rest of us. And maybe we can stop buying so much new hardware directly from the companies that pushed us aside. Because brand loyalty used to mean something to me. I had it for years. But seeing how these companies treat ordinary consumers is changing my thinking.

Look at Microsoft. After years of complaints about Windows 11, it's finally emphasizing performance, reliability, and improving the user experience. Meanwhile, Linux has become a more practical alternative for gamers, and Macs are attracting people frustrated with Windows. I can't prove this caused Microsoft's changes, but I'd certainly think twice about taking consumers for granted when they have alternatives if I were Microsoft.^[13]

And that's something Nvidia, Samsung, Micron, AMD, Intel, and the others ought to consider. Consumers helped build these companies. They shouldn't assume we'll automatically come back when the enterprise money slows down. Now, before anyone gets the wrong idea, I think AI is great technology, and I want to see it succeed. I actually like OpenAI and what it has made available to ordinary people. That doesn't mean I agree with everything its leadership says or does, especially Altman's comments about RAM.

But I also believe the AI industry has some wrongs to make right. A tremendous amount of copyrighted material has been used to train AI systems, often without the creators' knowledge or permission. You can't simply tell an AI model to unlearn it, but the companies that used people's work without permission should compensate them.

And the same goes for hardware manufacturers. They're making production decisions that are creating or worsening scarcity for ordinary consumers, then expecting us to accept the resulting prices as though nothing could have been done differently. When the financial bubble eventually bursts, I hope we consumers remember who treated us as valued customers and who decided we weren't profitable enough to matter anymore.

Maybe my tinfoil hat is getting a little tight.

But I don't think I'm imagining the consequences of what's happening right in front of us. This is all one big circle of money, financing, hardware, and companies making deals with one another. I'm not against AI. I'm not against companies making money. But I am against consumers being pushed aside while these companies chase profits that may not be sustainable.

The technology isn't the problem. The way the industry is financing it, controlling it, and treating the rest of us is.

And that circle needs to be broken.

💬 Personal

Thanks for reading this long post and if you read it all the way through, you're a superstar of a reader! I promise some shorter ones.

Sources

#hardware #opinion #technology

💬 Discuss... ✉️ Email 🦣 Reply on Mastodon