고란(KO RAN)

🚨This is why Asian markets are crashing. South Korea's KOSPI crashed more than 8% and triggered a circuit breaker that halted trading for 20 minutes. Japan's Nikkei fell more than 4% and Taiwan's TAIEX fell 3.41%. Here is what is hitting them all at the same time: 1. Fears of a memory chip oversupply. A column in Maeil Business Newspaper, a leading South Korean financial daily, warns the memory boom could end as early as 2029 as new capacity comes online. Memory is in short supply right now, so prices and profits are very high. When supply catches up, prices fall and so do the profits. The column says Korea has only 2 to 3 years to build new industries before the cycle could reverse. 2. Elon Musk added to the fear. He wrote: "China will be making a LOT of RAM in 2028." That is earlier than the 2029 date in the column, so the oversupply could start sooner than expected. China's CXMT plans to go from 350,000 wafers a month at the end of 2026 to 500,000 by the end of 2028. Morgan Stanley also estimates China will make up 30% of new global DRAM capacity by 2028. 3. US inflation came in hotter. September CPI rose to 3.7%, above the 3.6% expected. That pushed up the odds of a Fed rate hike, and markets reacted negatively. US chip stocks fell for a sixth straight day last night, erasing about $1.6 trillion in value. The same selloff is now following through in Asia. 4. War is keeping oil near $95. The US is considering using the "Mother of All Bombs" against Iran if President Trump orders new strikes.

🚨 TRUMP LOSING THE MIDTERMS COULD BE THE TRIGGER THAT POPS THE AI BUBBLE. The reason comes down to two things: Politics and CAPEX. The AI boom now requires an enormous amount of spending to keep growing. Major hyperscalers are on track to spend nearly $800 billion on CAPEX this year, per J.P. Morgan, roughly 10x their 2019 spending. UBS estimates overall AI spending could reach around $900 billion in 2026 and $1.2 trillion in 2027. That money flows through almost the entire AI trade. Microsoft, Amazon, Alphabet, Meta and Oracle build the data centers. Nvidia, Broadcom and AMD sell the chips. Then you have memory, networking, cooling, electricity and construction. This is why AI stocks don't need CAPEX to collapse. CAPEX growth only needs to disappoint Wall Street. And maintaining this level of spending is becoming harder. AI CAPEX consumed around 33% of hyperscaler operating cash flow in 2023. J.P. Morgan estimates that has reached roughly 93% in 2026. Hyperscalers issued around $121 billion of bonds in 2025, and J.P. Morgan expects around $250 billion in 2026. The full data center buildout could require trillions of dollars through 2030. So AI is becoming more dependent on debt and external financing at exactly the wrong time: U.S. borrowing costs are rising again. The Fed just raised rates to 3.75%-4.00%, while the 30-year Treasury yield has reached its highest level since 2002. Trump wants the opposite. He has publicly demanded rates of 1% or lower. Cheaper money would make it easier to finance data centers, power projects and the next round of AI expansion. But the Fed is independent, and some officials are discussing further hikes because inflation remains high. And this creates another risk around the midterms. Trump has been one of the strongest political voices pushing the Fed toward lower rates. Losing Congress would not remove Trump's pressure on the Fed because he would still be president. But it could weaken his broader ability to push a pro-liquidity, pro investment agenda through Washington while the Fed is moving in the opposite direction. And right now, the Fed is clearly not following Trump's preferred path. Most policymakers still expect another rate hike this year, while Fed Governor Michael Barr says further hikes will likely be needed to control inflation. That could leave Trump entering 2027 with a divided government, less room to advance new AI-supportive legislation, and a Fed still keeping borrowing costs high or potentially raising them further. For an AI industry increasingly dependent on outside financing, that's another major risk. Now add the midterms. Trump has made rapid AI infrastructure expansion a major part of his administration's policy. Executive Order 14318 directs federal agencies to speed up permitting for large data centers and related power infrastructure, use federal land and provide pathways for financial support to qualifying projects. But Republicans could lose control of Congress in November. Trump would still be president. His executive orders would remain in place. But a Democratic House would control committees and could increase oversight of the AI buildout through hearings, investigations and subpoenas. It could also create more fights over federal funding and push harder on some of the biggest political problems surrounding data centers: 1. Electricity bills. 2. Grid upgrades. 3. Water usage. 4. Environmental reviews. And who pays for the infrastructure these projects require. That matters because Trump's administration is trying to make the buildout faster, while a change in congressional control could increase scrutiny and political friction around parts of that expansion. And this could happen while Trump is pushing for much lower rates but the actual cost of borrowing remains extremely high. That's where the two risks meet. More political friction + expensive financing. For an industry preparing to spend trillions of dollars, that combination could be dangerous. A project already under construction doesn't suddenly disappear after the election. Existing chip orders don't automatically get cancelled either. The real question for markets becomes: What happens to the next round of CAPEX? If political pressure, higher power costs and expensive financing make hyperscalers even slightly less aggressive with 2027 and 2028 spending, Wall Street starts cutting future demand expectations. Nvidia doesn't need today's GPU sales to collapse. Investors only need to expect slower growth in tomorrow's orders. The same applies to Broadcom, AMD, Micron and the rest of the AI infrastructure chain. And this is where it becomes a broader market problem. Microsoft, Nvidia, Amazon, Alphabet, Meta, Broadcom and other AI-linked companies are among the biggest weights in the Nasdaq-100. So when the AI trade gets repriced, it isn't a small corner of the market falling. Some of the companies with the greatest influence over the entire index are falling together. That means a major AI selloff can drag the Nasdaq lower even if large parts of the market aren't directly involved. And investors are already becoming less willing to reward spending for the sake of spending. J.P. Morgan says higher CAPEX plans have recently been rewarded only when accompanied by stronger revenue expectations. Investors increasingly want proof that all this investment is actually generating demand and returns. That's what makes November important. AI enters the midterms with nearly $800 billion of hyperscaler CAPEX, spending consuming roughly 93% of operating cash flow, rapidly growing financing needs and borrowing costs already near multi-decade highs. Trump is pushing for faster AI infrastructure development and dramatically lower interest rates. A Republican loss wouldn't reverse all of that overnight. But it could add congressional pressure to an AI CAPEX cycle that is already becoming much more expensive to finance. And if that causes Wall Street to question the next trillion dollars of AI spending, the companies that drove this market higher would be the first place investors look.