Unlock Editor's Digest for free
FT editor Roula Khalaf has chosen her favorite stories in this weekly newsletter.
We all know that when investors jump on the slightest change in earnings estimates, stock prices soar. This can be seen in the sharp movements of AI stocks in recent weeks, particularly the sector's top names such as Nvidia, ASML, Arm, and US and Asian chipmakers.But they don't have to move in step with each other.
Let's take ASML as an example. A Dutch chip equipment maker's weaker-than-expected orders have triggered extreme volatility in AI stocks around the world. The logic was that new orders for machines made by the world's largest advanced chip equipment maker were lower than market expectations, so growth prospects for artificial intelligence chips should also deteriorate.
Certainly, the conditions for a correction are in place. The AI stock market boom is priced in too quickly. This hype has disproportionately benefited even some stocks unlikely to benefit greatly from AI-driven growth. Meanwhile, the sector remains highly exposed to geopolitical risks. China is an important market for most chip companies, including Nvidia and ASML. It's the latter's largest market, accounting for nearly half of its system sales in the first quarter.
Still, the recent wild swings in AI stocks highlight more than anything else the market's overreaction to relatively benign cyclical movements. The reality is that the AI chip market is likely to continue expanding rapidly despite the apparent weakness in the broader semiconductor sector. That's because, despite impressive sales growth over the past year, chips made for AI applications still only make up a small portion of the world's chip supply. For example, consider high-bandwidth memory chips, which are a critical component of all AI chips. It takes about six of these advanced memory chips, including those made by Nvidia, to power a single AI chip.
However, based on sales volume, it only accounts for about 1% of the total memory chip market. Even at today's growth rates, AI chips can offset cyclical downturns in the vast remaining market, which is still dominated by traditional memory, storage, and processing chips used in smartphones, cars, and electronic devices. You can not.
Here, companies are currently trying to deal with excessive inventory levels as a result of overproduction by chipmakers and hoarding by device makers during pandemic-induced shortages. This higher-than-normal inventory continues to suppress end-market demand. Therefore, there has been little urgency for chipmakers to increase production capacity or order new chip-making equipment in recent years. Memory chip prices have fallen more than 50% since 2021, and the drop in spot prices has continued this month. Recovery in demand has been particularly slow for automakers and smartphone makers, the largest consumers of semiconductors.
Additionally, ASML's quarterly order fluctuations are typically not a reason for concern. ASML sells only about 100 new lithography systems every quarter. The quarterly numbers are volatile because about three companies buy most of the company's advanced machines. Failure in one quarter does not necessarily indicate results for the rest of the year.
On the contrary, an extended period of lower-than-expected equipment orders from chipmakers means that large orders are likely to occur in coming quarters. More than 70 new chip manufacturing projects are being built around the world. TSMC and Samsung plan to begin mass production of next-generation 2-nanometer chips next year. This will require new equipment (which only ASML can manufacture) for the new production line to be ordered this year.
TSMC's latest results confirm this. Although it lowered its growth forecast for the chip market, excluding memory chips, this year's capital investment plan remained unchanged at $28 billion to $32 billion. The company is currently building new factories in countries such as the US, Japan and Germany, with plans already in place to begin producing its ultra-advanced 1.6 nanometer chips in 2026.
The underlying driver of the AI stock boom of the past two years has little to do with the cyclical ups and downs of the traditional chip and equipment sector, and more to do with excitement about AI's business-transforming potential. It's worth remembering. Those looking for signs of peaking AI hype focus on the discrepancy between soaring stock prices and the slower-than-expected pace of widespread AI adoption by companies, rather than normal cyclical forces. Should.
june.yoon@ft.com
