The PC era is dying. Welcome to the collective computer era 

 “The purpose of computers is human freedom.”

– Ted Nelson, Computer Lib/Dream Machines (1974)

The computer is as emblematic of the American dream as the automobile. Perhaps it’s only natural that Apple, HP, Adobe, Google, and Amazon were each launched out of a garage. 

It was inside the garage that the modern era of personal computers was born, where anyone could own the power to calculate millions, and then billions of processes per second. PCs are a tool designed to move us faster, with a hood you can pop open to soup up. We insist that our computers speed up every year if only because it’s proof of progress. The very term “personal computer” promises liberty and autonomy; this isn’t the bus, but a transistor-powered rocket carrying a payload of rare earth minerals and rainbow hued headlights. 

The PC shrunk whole industries of work to our desktops, driving our ambitions anywhere they wanted to go. Whether you were publishing without a publisher, creating art without a studio, balancing books without an accountant, or mailing without a post office, the computer offered an all-in-one device for self-starters—a business in a box. Like cars, we invest in new computers because they are more expensive to fix than keep. And they enable us to pursue the two most important American ideals: self-expression and capitalism. 

But half a century since the idea of the PC went mainstream, the personal computer as both a product and an ideal has never been more at risk. In the age of AI, companies are acquiring unprecedented amounts of hardware, driving up prices, and affecting the entire PC market along the way. It’s affecting everyone from Dell dudes to kids dreaming about a Playstation 6. 

You can call the PC endangered, or on the precipice of anthropogenic evolution. Either way, we’re currently shifting toward a world of consolidated processing, where the PC is trending toward a luxury item. Meanwhile, computation itself is becoming a utility, priced and positioned as intelligence that we rent on demand.

Buying a computer today

Computers of all types are generally more expensive to acquire today than they were a year ago. As companies including OpenAI, Google, Meta, and Amazon plan unprecedentedly massive data centers to process AI, they’ve gobbled up the available stock of chips—all while tariffs add an extra dollop of pain to one’s wallet.

DRAM in particualr—the semiconductor inside computers that holds information used by the CPU—is the culprit, with prices ballooning by 172% in 2025. Then they grew another 90% in Q1 of 2026 alone. On top of that, the NAND flash memory powering most solid state hard drives has also been on the rise, up by 50% in Nov 2025 and projected to keep growing.

In practical terms, this means a midrange gaming PC will cost hundreds more this year than last, as computer component prices have become one of the richest sources of memefodder on the internet. (Wedding gift, anyone?) 

Major OEMs like Dell and Acer are warning that 20% price increases are coming later this year. One analyst expects a 12% drop in desktop and laptop shipments. On one hand, 262 million PCs still sell every year, and sales were up 9% last year (though some of that was due to defensive purchasing as prices skyrocketed.) On the other, the PC hardly dominates mainstream computing these days. Smartphones already outsell PCs by 5 to 1.

Hit first and hardest will be the low end of the PC market, where margins are tightest. Chromebooks shipments could drop by 28% due to component shortages. Meanwhile, our cheapest, most specialized proxy for a PC—game consoles—are also feeling the sting. Nintendo, Sony, and Microsoft all raised prices on their products in the last year, while Sony and Microsoft in particular need to plan how they’ll build their next generation of Playstation and Xbox. The next Xbox is rumored to cost $1,000 (almost doubling the price). Sony may delay the PS6, anticipated for 2027, until 2029 in hopes that hardware gets cheaper. 

PCs as we know them seem like they’ll be forced to evolve—they may offer you a familiar mouse and keyboard, but the insides will be a little bit different (with less power and customization). 

There may be no better sign of the times than the Macbook Neo. Apple’s latest laptop uses the guts of an iPhone to stay cheap, and it limits RAM to a mere 8GB that’s soldered right onto the motherboard. For a majority of people, the Neo has plenty of power, much like a 4-cylinder sedan can get you onto the same highway as a Ferrari. But as a “new” approach, its single core architecture is inherently limited for the most compute-heavy tasks—tasks like advanced media production and running local AI—even compared to Apple’s other budget machines like the Macbook Air or Mac Mini. 

Apple seems to be tacitly stating the obvious while it camped outside Samsung to negotiate lower RAM prices: People can’t afford PCs anymore. We need something neo.

Intelligence is the new utility

At the same time that PC ownership is costing us more, modern AI providers are stepping in with a rent-to-never-own alternative. They are structuring computers as a utility rather than a product. The PC is becoming less something that you buy a couple times a decade, and more a phenomenon that you pay for through the wires in perpetuity. 

That means your Gemini, Anthropic, or OpenAI subscription probably cost you about $20 a month today. NVIDIA, knowing its GPUs are in high demand to train AI, even lets you rent a GPU in the cloud through its GeForce Now service (also about $20/month). 

The tech industry has dreamed about this sticky subscription revenue for decades—consider how everything from the Adobe Suite to Spotify are now a monthly line item. Steve Ballmer promoted the future of the “cloud” back in 2009, as tech companies framed the consolidation of computing power from individuals to corporations as an event every bit as natural as the weather. As Sam Altman himself recently put it: “We see a future where intelligence is a utility, like electricity or water, and people buy it from us on a meter.”

The true cost of this seemingly efficient setup isn’t just directly out of our pockets, though. These cost scale when you realize that modern LLMs are the very engine behind modern software and services. Apps, instead of running on their own servers (their own PCs hooked up to the internet, if you will) or even using servers from Amazon’s AWS, need to pay the meter every time their AI tools pick the brain of ChatGPT. Their expenses grow with your queries. You might pay for your own ChatGPT subscription, then pay for an app that’s paying for its ChatGPT subscription.

In Silicon Valley, we can get an early peek at how this adds up. I’ve heard of one popular AI assistant that was paying $150/user per month, eating a majority of that cost to build market share. Meanwhile, the coders who are pushing AI to its limits discover just how quickly those Claude Code credits can go. A single employee can easily spend four to five figures a month on AI coding. The value of that expenditure is proving worth it since AI can effectively amplify their workforce (or in Amazon’s recent case, effectively reduce their workforce). But companies don’t own this power. They rent it. 

This trendline isn’t without pushback. As AI positions itself as a utility, we are seeing a countertrend from vibecoders fighting for autonomy. As AIs get better, they also get more efficient. These AIs can theoretically be run locally. But in current practice, they’re actually hybrids, running some things locally, and some in the cloud. 

OpenClaw arrived in 2026 as something of a Model T for personal AI…an AI for the PC era, if you will. All you need is a Mac Mini and an openness to risk to build your own customized assistant. (Oh, but you still need to connect to a subscription LLM to really put it to work—and Claude is a popular choice.) Perplexity created its own version of OpenClaw with what they call, wait for it, “Personal Computer.” Also built for a Mac Mini, it’s a Perplexity agent that can access your files and coordinate tasks for you. (Oh, but there’s a catch again: The AI processing still happens on Perplexity’s servers.) 

I get the desire people have to operate our own AIs. It’s the lingering promise of the PC: computational autonomy, or homesteading in the digital age. Especially in the long term, it feels like these homesteaders could use hyper processing efficient, open source AI models so well that the Anthropics and OpenAIs can’t compete on a majority of daily tasks. 

But in the nearer term, while the most technically proficient will certainly keep experimenting with local AIs with increasing capability, anything short of an AI crash seems to ensure that building these PC systems will only get more cost prohibitive. Because none of the frontier model providers has any incentive to ween you off its drip.

The rise of the collective computer

Later this year, we’ll be getting a look at OpenAI’s first piece of consumer hardware, something spiritually akin to the iPhone for the AI age. While we know close to nothing about it, it does seem part of a larger suite of AI devices from Meta, and likely Google and Apple, that are designed less like self-reliant PCs than they are sensors. Ever listening, seeing, and recording, they promise to bring AI’s brain to every part of your life. 

In this scenario, we’ve more or less abandoned the PC completely, as we’ve abstracted the computer from PC (ours), to smartphone (half ours, half a company’s), to something new (and theirs). 

Such AI hardware as we’ve imagined it is little more than a data collector—a nerve ending—for a greater, centralized machine. Our many AI computers distributed around the world would be like a massive, receptive body for corporate brains like Gemini and ChatGPT.

That grand computer will not be a personal computer. It would be a collective computer, technically owned by the many but controlled by the few—the equivalent of installing a Nest thermostat but letting the power company set your home’s temperature. 

And while I’m not arguing that such a computer could be more capable than anything we’ve yet realized in the age of silicon, it is one we would own even less than the systems and services filled with strange terms and conditions today.

That collective computer is not a hot rod you wax in front of the neighborhood, nor is it a sensible sedan that quietly ferries your children to school. It’s a train that will go only as fast and as far as you can shovel the coal. Unless it all goes off the rails, first.

View the full article