Concerned about an AI bubble? Sign up for The Daily Upside for smart and actionable market news, built for investors.
You thought gasoline prices have gone up? That’s nothing compared with the soaring cost of memory chips, which have become not only incredibly expensive but also increasingly scarce relative to demand because of the AI buildout.
To see why, you only need to understand two things.
The first is compute, the most famous component of the AI buildout. In short, it’s the hardware that provides the processing power to speed up AI and machine learning. As Big Tech hyperscalers and the major AI labs have raced to obtain as much processing power as possible for their more and more sophisticated advanced models, high-end hardware like Nvidia’s graphics processing units (GPUs) has become one of the world’s most coveted resources. The demand transformed the semiconductor designer into one of the most profitable and valuable companies on earth.
Sign up for The Daily Upside at no cost for premium analysis on all your favorite stocks.
READ ALSO: Wall Street Looks to Zillow, Rocket Mortgage Earnings for Clues on K-Shaped Housing Market and Can Palantir’s Sizzling Growth Win Over Market Bears?
Running Into the ‘Memory Wall’
It has also led to a dramatic GPU shortage, enabling the valuations of rival GPU makers AMD and Intel to take off, too. Other competitors have emerged, like Google’s Tensor Processing Units (TPUs), which the Mountain View, California, company offers to select outside firms including Meta, with which it announced a multibillion-dollar pact in February.
The second thing, of course, is memory. It’s impossible for companies to scale on compute power alone, because compute hardware can only handle data processing. Memory hardware stores the vast amounts of data advanced AI models require to do all kinds of things.
It’s needed to generate responses to user queries, which requires sifting through troves of information for an answer. In AI agents, it’s needed to remember a user’s preferences and past interactions so they can offer consistent support. In complex coding and software engineering, it requires not just knowing a massive codebase, but also remembering all the debugging and troubleshooting that went into development so that it can fix problems. Basically, AI doesn’t advance from here without more and more powerful memory.
So, just as AI has created a gold rush for state-of-the-art compute, it has created a gold rush for state-of-the-art memory. And the top of the line when it comes to that is high bandwidth memory (HBM), a specialized form of dynamic random-access memory (DRAM).
Basically, the old-fashioned DRAM on your laptop is much too slow for the vast compute power of AI. So advanced HBM stacks layers of DRAM (in some cases a dozen or more layers) on an AI GPU.
Without enough memory, AI can hit what’s called a “memory wall.” Computing power increases but memory bandwidth doesn’t follow, resulting in a high-powered computing setup that’s frequently left idle while it waits for memory access speeds to catch up. At a certain point, it hinders the benefit of the compute itself. You can’t scale advanced AI without both: They are the two core needs of larger models.
AI hyperscalers are set to spend $750 billion this year and $1 trillion in 2027, according to Goldman Sachs. Much of that is going to be directed toward HBM and DRAM, which is currently creating both dreams and nightmares. Which one depends on who you ask.
Taking Stock and Trading Stocks
The global DRAM market is dominated by three companies, Samsung Electronics, SK Hynix and Micron Technology. The massive uptick in demand for their products is a dream scenario.
Money has poured in. Samsung Electronics, to take one example, reported record earnings for the third quarter in a row last week, with second-quarter revenue roughly doubling year-over-year to $117 billion and net profit multiplying 14-fold.
So, too, have investors. Shares in Samsung are up 267% in the past 12 months. SK Hynix is up 528%, and Micron Technology is up 690%.
The fact that this skyrocketing rally is powered almost solely by the AI trade, however, has summoned nightmares for others.
The memory market is now attached at the hip to investor sentiment regarding AI, which means when investors sour, things can get very bumpy. Early last week, chip stocks had at one point lost $1 trillion in market value in just a handful of days as investors fretted about the resilience of AI infrastructure spending.
Swept up in those concerns was the $24 billion, AI-focused hedge fund Situational Awareness, which sold most of its public equities portfolio to rival Citadel after its highly leveraged bets on volatile memory stocks brought it to the brink of collapse.
Another nightmare sufferer has been South Korea’s stock market, where SK Hynix and Samsung are heavily weighted, which has swung from meltdown to rebound depending on the current, oft capricious mood about AI. After last week’s selloff, the country’s financial regulators intervened to rein in excessive retail bets on the chipmakers by limiting access to highly leveraged (and wildly popular) ETFs.
Naturally, in keeping with the theme of volatility, South Korean chipmakers staged their biggest-ever rally on Friday, fueled by assurances in the latest quarterly reports by Google, Amazon, Microsoft and Meta that made clear the hyperscalers have no plans to dial down capital spending.
The demand those hyperscalers are creating, however, is also pushing Samsung Electronics, SK Hynix and Micron to their limits. Simply put, there’s not enough memory to go around and not enough manufacturing capacity to make up for it.
“We are continuing to increase production, but demand is growing even faster,” Samsung Vice President Kim Jaejune said on an earnings call last week. “Next year’s supply shortage will deepen further than this year’s, widening the gap between supply and demand.”
Paying the AI Tax
Of all the nightmares arising from the memory crunch, the one that’s most disturbing is the one that belongs directly to you.
Because AI has become so lucrative, the three major RAM manufacturers have shifted from mostly producing conventional DRAM to focusing more of their capacity on HBM, which is what AI data centers need.
DRAM, unfortunately for you, is what consumer electronics use for their computational memory needs. This has created a brutal shortage that has been not-so-lovingly dubbed RAMageddon or the DRAMpocalypse by observers.
The shortage means the price of memory is going up. DRAM contract prices rose as much as 89% in the second quarter of 2026 alone. Consumer electronics manufacturers can’t absorb cost increases of that magnitude, which means the costs are increasingly being passed on to consumers.
Last month, Apple hiked laptop and iPad prices by 15% to 25%. On an earnings call Thursday, CEO Tim Cook warned memory prices will only go up from here. Microsoft also hiked the price of its Xbox game console yesterday, asking $100 more for the starter model, citing the cost of memory chips.
Some have started to nickname this phenomenon an “AI tax.”
MacBooks and Xboxes, however, tend to have loyal customer bases with disposable income to spare. For consumer electronics companies that cater to the middle of the market, RAMageddon has been even more stark.
Take the smartphone industry. Counterpoint Research estimates global smartphone shipments fell 11% year-over-year in the second quarter to the lowest level in 13 years. Price increases resulting from the memory crisis were the main contributor.
While shipments rose from the big guns in the sector, like Samsung and Apple, mid-market competitors Xiaomi, OPPO and vivo each recorded double-digit percentage declines.
Further reflecting these dynamics, Counterpoint found that smartphone revenue rose 7% year over year to $109 billion in the quarter, despite falling shipments, reflecting the fact that more is being paid by fewer consumers.
The stress won’t be limited to the smartphone industry, either. When digital camera maker Canon reported last week, the company said it will “seek to mitigate the negative impact [of the memory crunch] as much as possible through price increases and additional cost reductions.”
Priced Out
According to Taiwan’s Economic Daily News, Hsiang Hsu, the chairman of computer hardware designer Micro-Star International, said the visibility of memory supply is only about a month. He expects consumer product shipments will fall 10% to 20% this year as prices increase.
MSI makes gaming laptops and gear, computer and gaming accessories, vehicle infotainment systems, computers for industrial use and motherboards, giving it a high-res view into the consumer electronics market.
Sharp, the Japanese consumer appliance firm that makes home electronics, appliances, smartphones and computers and thus also has a good view of the market, said it plans to focus on higher-margin offerings to offset rising memory costs. At a June business meeting, co-Chief Operating Officer Shigeru Kobayashi said the company “will accelerate the shift toward high-priced, high-value-added products to secure profits amid the rising price of memory chips.”
Research firm Gartner estimated earlier this year that DRAM and solid-state drive (SSD) prices would rise 130% by the end of 2026, in turn increasing PC prices by 17% and smartphone prices by 13%.
Apple shares fell 10% on Friday after it issued a disappointing sales forecast. The company noted it is proving difficult to acquire enough components, including memory chips, as the AI capex boom strains supplies.
In other words, your “AI tax” is probably going up. Look for those smartphone deals now.
This post first appeared on The Daily Upside. To receive razor sharp analysis and perspective on all things finance, economics, and markets, subscribe to our free The Daily Upside newsletter.
