SINGAPORE – US technology stocks, long the darlings of global markets, remained under pressure heading into Oct 15, after a six-day slide wiped US$1.6 trillion off semiconductor stocks.
Nvidia was down 3.2% on Oct 14, while Microsoft was down nearly 2%. Broadcom slipped 3.1%, and AMD fell 5.7%. Oracle lost 8.8%, while AI-infrastructure firms CoreWeave tumbled 15% and Nebius dipped 10%.
The S&P 500 fell 1.11%, the worst drop since July, while the Nasdaq slid 1.9%.
Asia markets also felt the heat from worsening tech sentiments in the US.
South Korea’s KOSPI index sank more than 7%, while Japan’s Nikkei stock index fell over 5%.
The Straits Times Index slid by a smaller margin, slipping 0.37% to close at 5,403.5 on Oct 15, although some of Singapore’s tech and semiconductor names, which are not components of the index, took a hit.
AEM Holdings dropped 9.9% to close at $8.80 on Oct 15, while Frenken was down 6.4% to $2.20. UMS Integration fell 9.7% to close at $2.33, and Addvalue Tech fell 6.9% to 12 cents.
What has caused the tech selloff and what should investors look out for?
Why tech stocks are falling
US inflation data came in “hotter than expected”, which raised expectations for further Federal Reserve rate hikes.
In September, the Federal Open Market Committee (FOMC) voted unanimously to raise the benchmark federal funds rate by 0.25 percentage points, bringing the target range up to 3.75% to 4%.
Further interest rate hikes by the Fed would hit artificial intelligence (AI) investment particularly hard, raising borrowing costs for one of the most capital-intensive cycles analysts have seen in the technology sector. said Saxo Markets chief investment strategist Charu Chanana.
Volatile US Treasury yields are adding to the uncertainty, having recently reached 24-year highs before retreating. But lower yields do not necessarily signal better conditions for tech stocks.
“Yields falling because markets are worried about growth or credit stress is quite different from yields falling because the Fed can comfortably ease,” Chanana said.
Investors are questioning whether companies borrowing heavily to build AI infrastructure can generate enough cash to service their debts. Those concerns can push up their borrowing costs even when Treasury yields fall.
Chanana pointed to specialised AI cloud provider CoreWeave, whose shares fell sharply as yields on some of its bonds approached 15%, signalling that investors were demanding greater compensation for lending to it.
Its financing has become a test of lenders’ willingness to keep funding the AI expansion, with implications for spending on data centres and chips.
“This is the first time in this AI cycle that the bond market, rather than the equity market, is beginning to dictate the conversation,” Chanana said.
Broadcom CEO’s ‘casino’ remarks
Broadcom chief executive Hock Tan added fuel to the fire when he likened AI investors to gambling in a casino during a DBS AI event on Oct 12.
“A lot of investors have no clue what they are investing in,” he said, adding that they “set very extreme expectations only to find out that it’s almost like they are in a casino.”
He also pointed to the importance of companies actually finding out what returns on investments they can generate from deploying AI technologies, rather than wasting money to experiment with AI.
Tan likened this to being a child playing in a video arcade with $10 worth of tokens.
“You finish it in 10 minutes and you haven’t learned anything … That’s what’s happening in enterprises today.”
Chanana said Tan’s comments mattered “because they crystallised something the market was already starting to worry about.”
But she noted that he also said that compute demand remains ahead of supply and projected that this would continue through 2029 or 2030.
“Someone eventually has to finance all this infrastructure, and investors increasingly want to know what return they are getting for it.”
Competition from China
Another factor has been competition from China, particularly in the memory space.
On Oct 10, at an industry conference in Shanghai. Chinese chipmaker CXMT reported progress in making denser memory chips, raising concerns that increased competition could erode the high prices and profits investors expect from Micron and SK Hynix.
“CXMT still has to prove that it can manufacture the technology reliably and at scale, so this is not an overnight fundamental threat. But when memory stocks have already rallied sharply, even a potential future increase in supply can hit valuations,” Chanana said.
Richard Tang, Julius Baer’s head of research for Hong Kong, said investors have grown increasingly anxious about the competitive threat from China, with CXMT paving the way for the commercial production of server memory chips by the end of this year.
He noted that the development triggered an immediate sell-off of US and South Korean memory makers.
What should investors watch out for?
Earnings will matter increasingly, said Billy Toh, CGS International Securities regional head of retail research.
“The first phase rewarded companies supplying the AI build out. Investors now want clearer evidence that all this spending can generate sustainable revenue, profits, and cash flow,” he said.
He added that investors should look beyond companies’ spending announcements to how they actually make money from AI.
For example, Microsoft reported in July that customers were paying for more than 30 million Copilot user licences, while Salesforce is charging for work performed by its Agentforce tools.
“The important questions are whether customers keep paying, usage grows and those revenues translate into better margins,” he said.
He noted that Nvidia, Broadcom and Micron remain important indicators of chip demand, while Microsoft, Salesforce, Amazon, Alphabet and Meta offer different ways to assess AI’s commercial returns.
Meanwhile, when asked whether the decline in Singapore tech stocks was linked to data centre designer AirTrunk’s delayed listing on the local exchange, Toh said both reflected the “same more cautious funding environment”.
Airtrunk’s planned US$1.5 billion Singapore real estate investment trust (REIT) listing was reportedly delayed until 2027 amid higher interest rates and market volatility.
“Data centres need substantial capital, so financing costs and investors’ required returns directly affect valuations,” he said.
“We wouldn’t attribute the delay solely to the US chip selloff or take it as proof that data-centre demand is weakening. It shows that strong long-term growth prospects do not remove the need to get pricing and timing right.”
OCBC chief economist Selena Ling said: “Ultimately, investors should distinguish between short-term valuation and financing concerns and the longer-term demand outlook, which remains constructive.”
She added that the key factors to monitor remain underlying semiconductor demand, AI-related capital expenditure and persistent supply constraints.
Lower share prices could offer opportunities in companies whose earnings continue to grow. But investors should be selective, as financial strength and the ability to deliver on business plans vary, Ling said.
