On Thursday, Apple did something it almost never does: it raised prices across nearly its entire MacBook and iPad lineup, effective immediately — and not by a polite $50. The 14-inch MacBook Pro jumped from $1,699 to $1,999. The 11-inch iPad Pro went from $999 to $1,199. A maxed-out 16-inch MacBook Pro now lists for a surreal $9,999. The entry-level MacBook Neo, the machine pitched at students and casual users, rose from $599 to $699 — a 17% hike on the cheapest laptop in the store.

The company blamed “surging memory costs.” That’s not corporate spin. It’s the spillover from something most consumers have barely registered: the world’s DRAM and NAND supply is being devoured by artificial intelligence, and the bill is now landing on everyone — including people who have never opened a chatbot in their lives.

The memory crunch isn’t coming. It’s already on the Best Buy receipt.

The numbers behind the memory market are almost absurd. Micron, one of the three major DRAM manufacturers, saw its revenue more than quadruple in its most recent fiscal quarter. High-bandwidth memory — the specialized DRAM that powers Nvidia’s AI GPUs — is in such short supply that Micron’s order book is effectively locked through 2027. The memory supercycle, as analysts are calling it, may not ease until 2028. These aren’t forecasts pulled from a trade-journal headline; they’re reflected in balance sheets. Micron’s market capitalization hit $1 trillion in May, doubling in 48 days.

DRAM and NAND are not infinite. The same fabrication capacity that produces the memory for a MacBook Air also produces the memory for a data-center GPU cluster training the next large language model. When one buyer shows up with effectively unlimited capital and a willingness to pay whatever it takes, the other buyers — including Apple, the most feared negotiator in the electronics supply chain — get squeezed. Apple finally blinked, and it blinked hard.

The predictable reaction online has split along well-worn lines. One camp calls it price-gouging: Apple is sitting on $160 billion in cash, why is it nickel-and-diming customers? The other camp blames Washington’s fiscal policy or tariffs. Both miss the point. Apple didn’t raise prices because it wanted to protect its famously fat margins. It raised prices because the marginal cost of memory has genuinely exploded, and even a company with Apple’s supply-chain leverage can’t outmuscle an entire industry in the grip of an AI arms race.

Sam Altman doesn’t buy his own RAM — but you do.

Here’s the uncomfortable truth the AI industry would prefer not to discuss: the real-world cost of building artificial intelligence is not confined to the subscription fees of ChatGPT Plus or the Azure bills of a startup. It is leaking into the price of every device that uses the same foundational components.

When OpenAI, Anthropic, Google, and every venture-backed AI lab on Sand Hill Road race to train ever-larger models, they consume physical inputs — silicon, rare-earth minerals, energy, and especially memory — that are not purely additive to the global supply. There is no separate factory that makes “AI memory” and leaves consumer memory untouched. It is the same pool. Every HBM stack that goes into a Blackwell GPU is DRAM capacity that cannot go into a laptop.

“The hyperscalers pay spot, and they pay in bulk,” one memory-industry procurement manager told me over a bad connection from a fab complex in Taichung. “Apple used to be the customer everyone feared. Now it’s the customer that’s just large enough to get told what the new price is.”

This is not an argument against AI. It is an argument for honesty about its costs. The industry has sold a story of abundance: intelligence too cheap to meter, a productivity boom that pays for itself. But the supply chain tells a different story — one of constrained physical resources, bidding wars, and costs that cannot simply be engineered away. That story is now visible on Apple’s online store.

The distributional question nobody wants to answer

The politics of this are almost radioactive, which is why nobody in Silicon Valley or Washington seems eager to acknowledge them. The memory crunch is a regressive tax on computing. The people training foundation models are not the ones paying $200 more for a MacBook Neo. The people buying high-end workstations for AI development can expense the bump. The student, the small-business owner, the school district refreshing its iPad fleet — they eat the increase with no offsetting benefit.

This isn’t a distributional effect anyone planned. It’s simply what happens when an industry grows fast enough, and concentrates enough capital, to bid physical inputs away from the rest of the economy. If you’ve wondered why the promised consumer benefits of the AI revolution have been slow to materialize while the costs seem to arrive ahead of schedule, Thursday’s price hike is your answer.

A trillion-dollar appetite, paid in installments

None of this is Apple’s fault in any meaningful sense. The company is a middleman between memory fabs and end users, and it held the line longer than most. If anything, the speed and size of the hike should be read as a warning about what’s coming for every other electronics manufacturer with less pricing power and thinner margins. The $499 laptop doesn’t have $100 of memory-cost headroom to absorb. Some of those OEMs will simply exit the market.

The AI industry is fond of talking about externalities — alignment risk, job displacement, the paperclip maximizer. But the most mundane externality is right here, on a Thursday morning in June, in the form of a price tag: when a handful of companies bid up a scarce physical resource to train models that may or may not deliver on their promises, the rest of us pay a premium on the machines we use to do everything else. That’s not a hypothetical. That’s $500 on a 14-inch MacBook Pro, and it’s probably just the start.

Sources