Overnight in U.S. stocks, the global semiconductor sector suffered a violent sell-off. The Philadelphia Semiconductor Index fell sharply, with Micron Technology and SanDisk both falling more than 6% during the session, and Western Digital falling more than 7%. During the Asia-Pacific trading session, SK Hynix plunged nearly 10% intraday, Samsung Electronics fell nearly 9%, and South Korea's Kospi index triggered an intraday circuit breaker for the second time in a week.

There are several direct incentives: Apple has rarely raised prices on all Mac and iPad series, with a maximum increase of US$500; Microsoft announced a third price increase for Xbox game consoles on the same day; AI giant OpenAI is considering postponing its IPO.

But what these news really touches is the most sensitive nerve in the market: when the chips required for AI are so expensive that even Apple can’t afford them, the entire incremental logic of AI may have to be rewritten.

What is DRAM?

First explain a basic concept.

DRAM, called dynamic random access memory in Chinese, is one of the core components of electronic products. You can think of it as the "short-term memory" of computers and mobile phones - when you open an app, the data is transferred into DRAM first; when you play games, the screen data is also temporarily stored in DRAM. Without it, the device can do almost nothing.

No one paid attention to this chip in the past. But today, how much your next device sells for is largely determined by it.

Source: DRAMeXchange

As of June 26, DRAMeXchange quotations show that the spot price of DDR5 16Gb closed at US$46.73, and DDR4 16Gb soared to US$72.32. DDR4 is more expensive than DDR5 - this is the first price inversion in DRAM history. For example, if you buy a used car from last year, it will be more expensive than a new car this year. It sounds absurd, but this is the reality of the DRAM market.

It becomes more exaggerated when viewed over time. From the third quarter of 2025 to the second quarter of 2026, the price of ordinary DRAM has increased by approximately 4.5 times. Citigroup predicts that the average price of DRAM will increase by 200% this year, and Goldman Sachs predicts that it will increase by 176%. Morgan Stanley believes prices will rise another 20% to 30% in the third quarter.

AI "eats" production capacity

Earlier this week, Micron handed over a financial report worthy of being recorded in history: revenue of US$41.46 billion, a year-on-year increase of 346%, and gross profit margin of 84.9% - higher than Nvidia. This once forgotten chip company is now the brightest star in the capital market.

But every penny of this "exploding" financial report was taken out of other people's pockets.

Who is paying the bill? Google, Amazon, Meta, Microsoft, Tesla. These "old giants" in the AI ​​era are burning money at an alarming rate. Goldman Sachs predicts that their combined capital expenditures may exceed US$770 billion in 2026. Where did the money go? A large part of it went into the pockets of Micron, SK Hynix, and Samsung.

The consequences are already beginning to show.

In order to take the lead in the AI ​​race, these technology giants have to spend heavily on expensive GPU and memory equipment, causing the real disposable free cash flow on the company's books to shrink significantly. According to the chart data shown in the Nomura Securities research report, the free cash flow forecasts of cloud vendors including Amazon, Microsoft, Google, Meta and Oracle in the next 12 months will show a cliff-like decline, marking that AI investment has officially entered the "cash-burn verification period."

In other words, every dollar Micron makes corresponds to a piece of free cash flow that disappears from the accounts of a certain technology giant.

Apple raised prices and blew the whistle

The news of Apple’s price increase is the most critical turning point in this game.

Why? Because Apple is the company with the strongest supply chain management capabilities in the world, bar none. If even Apple cannot internally digest the cost pressure brought about by DRAM price increases and is forced to pass the price on to consumers - it means that the entire industry chain has reached its limit.

Apple’s stock price fell more than 6% after news of the price increase. The market reads the signal: When a Mac costs $500 more, what will consumers do? Postponing replacement, foregoing upgrades, or even switching to cheaper alternatives. The cumulative effect of these decisions is shrinking demand.

Following this, Microsoft also announced a price increase for Xbox. The game console market is extremely price-sensitive, and every price increase will directly affect sales. Microsoft has obviously made the calculation: if it doesn’t increase the price, it will lose a profit on each unit it sells; if it increases the price, it will at least lose a little less.

Deutsche Bank estimates that the surge in storage costs is being translated into a "chip inflation tax" - consumer electronics terminal market revenue will fall by 15% year-on-year in 2026. Morgan Stanley also warned that by 2027, the storage wafers available for consumer electronics will be at least 15% short of demand.

When the cost of AI infrastructure begins to erode terminal profits, the market no longer asks "who makes the most money in AI", but instead asks "who is paying for AI and who is receiving money." Under this new logic, there are only three winners - Micron, SK Hynix, and Samsung. They monopolize the supply of DRAM and HBM. No matter how the giants fight, they only collect money.

What is the market afraid of?

Back to the plunge overnight.

Micron actually just handed over a record-breaking financial report-both revenue and gross profit margin exceeded expectations. But the market still crashed. This shows that what investors are worried about is not short-term performance at all.

The question they are really asking is: Will the profits made by chip stocks from rising prices completely crush end demand?

PC manufacturers, mobile phone manufacturers, and automobile manufacturers no longer have the ability to absorb costs and can only pass on prices to consumers. When consumers face price increases, the simplest reaction is: not to buy.

A strategist at Singapore's Longao Bank said: The news of Apple's price increase and OpenAI's postponement of its IPO "have caused the market to restart discussions about where the limit of the memory bottleneck is."

In the second half of the year, we speculate that the following factors may trigger a larger correction in U.S. stocks:

• ROI verification pressure during the earnings season: If cloud giants are unable to provide corresponding returns after huge investments in AI, investors will lose patience due to profit margin compression, triggering a valuation correction.

• Credit financing triggers pressure on valuations: Large-scale debt issuance by companies may push up financing costs and suppress the overall valuation of technology stocks.

• Liquidity siphoning from giant IPOs: As star AI companies go public in the second half of the year, institutions will have to sell off existing technology positions in order to subscribe for new shares, thus causing significant liquidity suppression in the secondary market.

• Inflation and disturbances in the Fed's stance: If energy and other factors trigger a rebound in inflation, the Fed may be forced to maintain a hawkish stance, which will directly interrupt the expansion of stock market valuations supported by expectations of interest rate cuts.

• Risk of cutting AI spending: This is the most lethal risk point; once giants cut spending, it will directly cause the AI industry chain (such as semiconductors) to face serious profit forecast revisions.

The incremental era of “whoever touches AI will rise” in the past two years may be coming to an end. What will succeed it is a more cruel era of redistribution - only the giants at the top of the supply chain will have the last laugh, while most players are paying for this infrastructure competition.

As for you and me, the consumers? The first round of inflationary impact of AI has quietly climbed into the shopping cart.