The promise of artificial intelligence has always carried an implicit economic assumption: that more computing power, applied intelligently, would drive costs down. Cheaper drug discovery. Cheaper logistics. Cheaper software development. A more productive economy with lower prices for everyone. That assumption just ran into reality. And reality, on 25 June, took the form of two back-to-back price announcements from two of the world’s most valuable companies.Giving rise to AI-flation.
Apple raised prices on MacBooks and iPads by up to $300, citing what CEO Tim Cook called “unavoidable” cost increases driven by surging memory and storage chip prices. Its shares closed 6% lower – one of the largest single-day drops in years. Within hours, Microsoft announced Xbox console price increases of $100 to $150 per model, effective August 1. Microsoft fell 3.5%.
The back-to-back announcements confirmed what the data has been showing for months: the AI chip shortage is no longer a problem contained within server farms and hyperscaler balance sheets. It has crossed the line from infrastructure cost to consumer price.
What is AI-flation and why is it happening now?
AI-flation is the transmission of AI infrastructure costs into consumer prices through shared supply chains. It’s happening because the AI buildout and the consumer electronics industry use the same fundamental inputs – memory chips – and the AI buildout is winning the allocation battle comprehensively.
Here is the mechanism in plain terms. Nvidia’s latest AI GPU requires eight modules of High-Bandwidth Memory. Each module uses 12 stacked DRAM chips. That’s 96 individual DRAM chips per AI accelerator. A single moderate-to-large data centre contains thousands of these accelerators.
The three companies that manufacture virtually all the world’s DRAM – Samsung, SK Hynix and Micron, who together control over 95% of global supply – have responded rationally: they’ve shifted production toward HBM for AI, because HBM generates three to five times more revenue per wafer than conventional DRAM.
The result is a zero-sum allocation problem. Data centres now consume an estimated 70% of all memory chips produced worldwide. For every ten memory chips rolling off a production line, seven are heading to server farms. That leaves the entire consumer electronics industry – smartphones, laptops, tablets, gaming consoles, smart home devices, cars – fighting over the remaining three.
In some cases, spot prices for DRAM have jumped nearly 700% in the past year – with analysts projecting an additional 30–40% climb through the rest of 2026.
A cycle or structural shift?
Memory chip shortages aren’t new.
The industry runs in cycles – oversupply drives prices down, undersupply drives them up, manufacturers invest in new capacity, the cycle resets.
Investors who’ve seen this pattern before may be tempted to treat AI-flation the same way: a temporary squeeze that normalises in 12–18 months.
That reading is almost certainly wrong this time. The International Data Corporation’s (IDC) assessment is unambiguous:
“This is not just a cyclical shortage driven by a mismatch in supply and demand, but a potentially permanent, strategic reallocation of the world’s silicon wafer capacity.”
The reason is physics and economics combined.
Building a new semiconductor fabrication plant costs $20–30 billion and takes two to three years from groundbreaking to first wafer. High Bandwidth Memory (HBM) production is already sold out through 2026 by major producers Micron and SK Hynix, confirming that the technology’s production readiness is lagging far behind market needs.
Even the emergency capital expenditure programmes now underway at Samsung, SK Hynix and Micron will not produce meaningful new supply before late 2027 at the earliest.
Meanwhile, AI demand is not plateauing. By 2030, AI servers are projected to account for more than 60% of global memory consumption – up from 50% today.
The second-order effects
AI-flation doesn’t stop at the consumer electronics shelf. It has second-order effects that reach further into the economy than most investors have modelled.
Chip prices are raising the price of downstream tech, consumer goods and automotive products – echoing the chip-shortage-driven price surges of the COVID era.
Honda has already flagged a volume reduction of 110,000 vehicles in North America due to semiconductor shortages. Yamaha Motor has described the business environment as facing “headwinds” from DRAM and NAND price increases.
The auto industry, which only just recovered from the COVID-era chip crisis, is being hit again from a different direction.
AI-flation creates a clear set of winners and losers, and the market is still in the process of pricing the distinction correctly.
#1: The memory manufacturers are the most direct beneficiaries
Samsung, SK Hynix and Micron are charging three to five times more revenue per wafer for HBM than for conventional DRAM. Their margins are structurally elevated for as long as AI demand outpaces new fab capacity, which, on current timelines, means at least 2027. Micron has significant US investor accessibility and direct HBM exposure to Nvidia’s GPU supply chain.
#2: Consumer electronics companies are structurally squeezed
Companies that sell consumer hardware – smartphones, PCs, tablets, gaming consoles – are caught between rising input costs and consumer price sensitivity that limits how much they can pass through.
#3: Semiconductor equipment companies are the picks-and-shovels play
The only long-term solution to AI-flation is more fab capacity. Building that capacity requires ASML’s lithography machines, Applied Materials’ deposition tools and Lam Research’s etching systems. Every new fab that breaks ground in the next 24 months is a multi-year equipment order.
The shortage that is causing AI-flation today is simultaneously funding the capex cycle that benefits semiconductor equipment manufacturers through 2029 and beyond.
Not a subscriber to Money Morning?
You can get free daily recommendations like these with Money Morning eletter. Just sign up here.
