Samsung Electronics expects the global shortage of artificial intelligence (AI) memory chips to deepen next year and persist into 2028, pointing to sustained demand from hyperscale data centre operators, even as the South Korean technology giant reported a more than 250-fold jump in semiconductor profit for the second quarter.
The upbeat demand outlook helped ease investor concerns that the rapid pace of AI infrastructure spending by major technology companies could slow in the coming years. Samsung shares rose as much as 8 per cent in early trade on Thursday before paring gains to trade 1.1 per cent lower.
“The supply shortage in 2027 is expected to worsen compared to this year, and it is expected to continue in 2028,” Jaejune Kim, Executive Vice President of Samsung’s memory business, told analysts during the company’s post-earnings conference call.
Kim said Samsung has secured long-term supply agreements with the world’s five largest data centre operators and is close to signing similar contracts with five additional major customers. The agreements, which run for at least five years, are expected to account for 60-70 per cent of the company’s long-term production capacity and include upfront payments and minimum pricing mechanisms to reduce investment risks.
Related: New Businesses On The Rise In India
Samsung’s semiconductor division posted an operating profit of 89.2 trillion won ($61.7 billion) in the April-June quarter, marking a dramatic turnaround from the same period a year ago and reflecting strong pricing and demand for AI memory chips.
Overall operating profit came in at 89.5 trillion won ($61.98 billion), broadly in line with the company’s earlier estimate of 89.4 trillion won, compared with 4.68 trillion won a year earlier. Revenue more than doubled, rising 130 per cent year-on-year to 171.5 trillion won.
Challenges in the Mobile Division
However, raised chip prices weighed on its smartphone business. Samsung’s mobile division posted a loss of 700 billion won, marking its first quarterly loss.
Samsung said revenue from its high bandwidth memory (HBM) chips—critical components used in AI accelerators—will more than triple in the third quarter as production ramps up. The company, which supplies HBM products to customers including Nvidia and Advanced Micro Devices, said stronger HBM sales should help align its share of the HBM market with its broader share of the dynamic random access memory (DRAM) market during the second half of the year.
Related: Chatbots May Be Talking to Competitors
Samsung has been working to narrow the gap with domestic rival SK Hynix, which established an early lead in supplying HBM chips for AI applications.
It is a familiar pattern for the manufacturer. In previous hardware cycles, competitors who secured early manufacturing advantages often held market share for extended periods, requiring significant internal restructuring and capital investment to close the gap. This time, the focus appears to be on securing long-term volume commitments rather than chasing immediate market share gains.
Samsung also said its contract chip manufacturing business is expected to return to profitability in the near future as factory utilisation improves and semiconductor prices strengthen. The company remains on schedule to begin operations at its Taylor, Texas fabrication plant later this year and plans to break ground on a second facility, which is expected to begin mass production in 2030.
Chief Financial Officer Park Soon-cheol said Samsung has no plans to issue American depositary receipts (ADRs), despite SK Hynix’s recent listing in the United States.