Global tech giants are showing no signs of slowing their AI expenditures, a record-breaking wave that is fueling growth for hardware suppliers like Samsung Electronics Co., Ltd. and SK Hynix. Despite lingering skepticism about the sustainability of AI demand and its ability to justify such massive capital outlays, Meta Platforms, Inc. alone has disclosed an ambitious spending plan of $135 billion for this year—one of the largest planned expenditures in the corporate world.
Their suppliers are responding in kind: SK Hynix announced plans to "significantly increase" capital expenditure, while Samsung also stated it would ramp up investments in memory chip production capacity. Concurrently, investors continue to reward this ambition as long as companies demonstrate growth. Microsoft's stock fell 6.1% after it revealed a slowdown in its cloud services growth, whereas Meta's shares climbed 6.6%.
Earnings reports this week from a series of industry titans underscored the immense market demand for AI hardware, a momentum likely to persist into 2026. Meta, Microsoft, and peers like Amazon and Alphabet—the hyperscale cloud providers—are driving a global procurement wave for chips, servers, and computing equipment, igniting enthusiasm among hardware suppliers worldwide, particularly in Asia.
Samsung and SK Hynix, the two primary manufacturers of the memory chips essential for NVIDIA's core AI accelerators and data center servers, both reported profits that surged severalfold. ASML, the sole supplier of the cutting-edge lithography machines required for producing advanced semiconductors, also delivered results that far exceeded expectations.
"The real headline is the 'continued climb' in capital expenditure. It's not surprising to see a cooling in the share price reaction as investors digest these aggressive investment plans," said Matt Britzman, an analyst at Hargreaves Lansdown, referring to Meta.
Meanwhile, this massive demand is exacerbating a global chip supply-demand imbalance, threatening multiple industries from smartphones and consumer electronics to automotive manufacturing. While demand for the NVIDIA and AMD accelerators needed to develop and run AI is already outstripping supply, investors are growing increasingly concerned about potential similar shortages for foundational memory chips.
Elon Musk recently stated in a podcast with X Prize Foundation founder Peter Diamandis that semiconductor supply will become a significant bottleneck for the growth of companies, including Tesla, potentially making the establishment of a "Tesla TeraFab"—a factory capable of manufacturing both logic and memory chips and providing packaging—a necessity.
"If we don't build the fabs, we will hit the 'chip wall.' We have two choices: hit the wall or build the fabs ourselves," Musk said.
Memory chip makers are shifting production lines toward high-bandwidth memory (HBM), a high-margin product tailored for AI data centers. Because producing the same capacity of HBM requires approximately three times the wafer capacity of standard DRAM, this shift is reducing supply for the consumer electronics industry.
This threatens PC manufacturers and smaller electronics firms, potentially leading to double-digit price increases for them. However, concerns about the end-demand for AI persist. Microsoft's capital expenditure grew 66% this quarter, beating expectations, but revenue from its Azure cloud division grew 38%—one percentage point slower than the previous three months.
On the other hand, Meta CEO Mark Zuckerberg spoke of a "massive AI acceleration" that has been brewing in the tech industry for over a year. "I expect our first models to be quite good, but more importantly, we are going to show the rapid trajectory we are on," he said.
During Thursday trading in Seoul, Samsung's stock declined about 0.4%, even as the company stated it plans to begin shipments of its next-generation HBM4 memory in February—a key step in catching up to SK Hynix in this lucrative segment. Shares of SK Hynix rose approximately 2%.
Samsung's chip division reported a fivefold increase in profit, significantly surpassing analysts' average forecasts. The company also announced a 3.57 trillion won share buyback and a special dividend, bringing its total fourth-quarter dividend payout to 3.75 trillion won.
In Asia, the focus is on the battle for leadership in next-generation HBM4, which will be integrated with NVIDIA's upcoming flagship Rubin processor. Samsung is close to securing NVIDIA's certification for its latest version of AI memory chips.
The AI construction boom is also helping to bolster Samsung's foundry business, which competes with industry leader TSMC. Company executives stated that Samsung's contract chip manufacturing division expects orders for 2-nanometer chips to grow by about 130% in 2026 and is actively engaged in discussions with clients in the US and China.
Comments