Elon Musk Says China Will Make “a LOT of RAM” in 2028

Musk Points to Chinese RAM in 2028

Elon Musk has turned a domestic policy debate in Seoul into a global supply question for the memory chip market. The Tesla chief executive replied on Thursday to a summary of a financial column published in Maeil Business Newspaper, one of South Korea's leading business dailies, with a single line: China will make “a LOT of RAM” in 2028. A summary of the argument circulated through the P Equity Research account on X; that summary post is what drew Musk's reply. His post did not identify which Chinese companies he expects to drive that output, nor whether it would target the segments Korean suppliers dominate. The column behind that reply argues that Korea's economy now leans heavily on Samsung Electronics and SK Hynix, and that this concentration leaves the country exposed once the current boom fades. Its author puts the inflection point around 2029, when new production capacity starts running, and gives Korea only two to three years to build new growth industries before the cycle potentially reverses. The timing is what makes the warning hard to dismiss. Korean chipmakers are posting record profits, demand still outruns supply, and rising prices, not unit growth, are driving the revenue surge. Peak earnings arriving alongside a supply threat is rarely a stable mix for any single-industry economy. His one-line reply carried no numbers of its own, but it lands on a debate that already has them. He has also praised Micron's pricing restraint, a sign he tracks memory costs closely as an input for his own companies. If his 2028 timeline proves right, Chinese RAM would arrive at scale a year before the column's oversupply estimate, and Korea's diversification window would shrink accordingly.

@pequityresearch · X post

That summary post.

View on X

Gartner's $837 Billion Memory Year

The numbers behind the bet are large. Gartner forecasts memory revenue of $837 billion in 2026, equal to 54% of global semiconductor sales, and credits rising prices for much of the jump from a 27% share in 2025. Memory is the volatile half of the chip industry: revenue swings with pricing cycles rather than unit demand alone, which is why a price-driven jump of this size reads as both strength and fragility. Even so, Gartner expects supply to stay tight in 2027 as extra capacity enters the market, and it projects memory revenue will cross $1 trillion that year. The warning also lands on receptive ground in Seoul. Seoul's own central bank has already framed the AI chip trade as a financial stability problem rather than a pure growth story. On the supply side, the threat is concrete: ChangXin Memory Technologies (CXMT) recently disclosed progress on a denser DRAM design, and the news weighed on Samsung and SK Hynix shares, which together account for roughly half of the KOSPI index. Concentration compounds that exposure, since a pricing turn at two companies carrying half a national index reaches pension savings and exporters at the same time. CXMT held about 8% of global DRAM revenue in the first quarter, according to Counterpoint Research, a figure that predates the denser-design disclosure. DRAM is the commodity-grade memory that data centres buy in bulk, which makes the Chinese challenger's roadmap a direct question for Korean margins. Traders outside the Korean cash market can express the timing view through stock perpetuals on large-cap chipmakers, products that track a share's price without an expiry date. None of this makes a glut imminent. Gartner's own 2027 tightness call is the reason the column's author still sees a runway before the turn. Every forecast in the chain, from $837 billion this year to $1 trillion next year, assumes pricing power holds.

Our reading is that the 2029 date deserves less confidence than the column gives it. Two of the three variables that decide this cycle sit outside Korea: Chinese capacity, which Musk now flags for 2028, and global demand, which Gartner expects to keep absorbing supply through 2027. The lever Seoul controls, diversification, is the slowest to move. Record profits and a forecast $1 trillion memory year raise the cost of complacency precisely because they make today's concentration look safe. The trigger to watch is a production line rather than a calendar date: once CXMT's denser DRAM reaches volume output, the oversupply argument stops being a column and becomes a forecast. Any broad Korean index ETF carries much the same two-company exposure as the KOSPI itself.