The U.S. stock market rose on a rebound-buying spree after the Independence Day holiday. Technology stocks led the way when trading resumed on the 6th local time in New York. The Nasdaq Composite gained 1.12%, the S&P 500 rose 0.72%, and the Dow Jones Industrial Average increased 0.29%. The Dow set a new all-time high.
The leading driver was AI semiconductors. AMD surged more than 6% right after Goldman Sachs raised its target price from $450 to $640. Broadcom rose 3.7% on news that it had extended its partnership with Apple. Western Digital jumped 7%, while TSMC and Qualcomm also posted strong gains.
Big tech also added momentum. Tesla soared on news of an expanded robotaxi service. Meta climbed nearly 3%. Microsoft, however, fell after announcing 4,800 layoffs, equivalent to 2.1% of its workforce. About one-fifth of the cuts will come from its Xbox gaming division alone.
SpaceX, which is set to be added to the Nasdaq 100 index on the 7th, slipped by nearly 1%. The company, which went public on the 12th of last month, had surpassed a market capitalization of $2 trillion as the largest IPO ever. Neo-cloud company TeraWulf rose 4.86% on the back of a long-term lease agreement with Anthropic.
A crack in the “AI chip shortage” narrative
The rebound marked a sharp reversal from the previous day’s market moves. The trigger came on the 1st of this month. Bloomberg reported that Meta was preparing “Meta Compute,” a cloud business that would lease out surplus AI computing resources. Meta’s share price rose, while semiconductor stocks fell. The same news was interpreted in the opposite way.
At the heart of the matter is the logic of scarcity. The foundation of the semiconductor rally over the past three years has been the belief that demand far exceeds supply. When the company that has been buying the most chips said it had spare server capacity, that premise came into question.
The shock was greater in South Korea. On the 2nd, the KOSPI plunged 7.89%, wiping out about 569 trillion won in market capitalization in a single day. A circuit breaker was triggered intraday. Foreign investors net sold Korean stocks for eight straight trading sessions. The won weakened to around 1,550 per dollar, approaching its weakest level since 2009.
The shortage narrative the market had relied on had no way to be verified. Real-time GPU utilization rates and data-center vacancy rates have never been disclosed. The Meta report exposed that gap.
Over the weekend, the mood changed. Major investment banks consecutively expressed a bullish outlook for the AI industry. Capital spending plans also remained intact. Meta’s annual capex forecast stands at $125 billion to $145 billion. Alphabet has guided for $185 billion, and Amazon has put its figure at around $200 billion. Combined spending by the four hyperscalers exceeds $650 billion.
Macroeconomic conditions also supported buying. International oil prices edged lower after Saudi Aramco cut its Asian selling prices. WTI settled at $68.55 per barrel. Earlier, U.S. June payrolls increased by only 57,000, missing expectations. As oil prices and hiring both cooled, expectations for interest-rate cuts rose. The U.S. 10-year Treasury yield fell to 4.47%.
The investment cycle is shifting from “building” to “monetizing”
The market is now shifting its focus from spending size to the speed of monetization. On the 6th, Anthropic signed a 20-year lease agreement with TeraWulf. The deal covers a 401MW data center in Kentucky. The contract is expected to generate about $19 billion in revenue over its term, or more than 29 trillion won. TeraWulf, which shifted from Bitcoin mining to AI infrastructure, secured long-term fixed revenue. Meta’s move to sell surplus capacity fits the same pattern: turning accumulated infrastructure into revenue.
In the AI industry, a circular capital structure has taken hold. Firms raise money, buy chips, rent out excess capacity to other AI companies, and then use that demand as justification to expand facilities again. Growth and bubble concerns are emerging at the same time. GPUs are assets that lose value over time, and some argue that depreciation costs are becoming a burden.
Investment criteria are also changing. The market is reacting less to the scale of investment and more to the pace of monetization. Capital is flowing toward companies that lease out excess capacity or lock in revenue through long-term contracts. By the end of August, Nvidia’s data-center revenue and margins, as well as Meta Compute’s actual pricing and customer base, are expected to become the next major turning point.
The same questions are being raised in the Korean market. The KOSPI is heavily reliant on the semiconductor narrative. SK Hynix is pushing ahead with a U.S. listing worth about $28 billion, or 43 trillion won. Samsung Electronics is also preparing to announce preliminary second-quarter results. The point at which AI demand is confirmed in actual earnings is being seen as a key variable that will determine the market’s direction.