What Samsung's AI Chip Boom Means for Supply to Data Centres

Samsung Electronics is one of the main producers of the components that support AI infrastructure.

It is so popular, in fact, that rising demand for memory chips is helping the company forecast a major increase in second-quarter operating profit.

The South Korean technology giant expects an operating profit of ₩8.9tn (US$5.8bn) between the start of April and the end of June.

The figure represents a significant rise from Samsung's ₩4.7tn (US$3.1bn) operating profit, reported in the same period last year.

Once associated mainly with conventional hardware, Samsung now supplies components that support next-generation AI infrastructure used across data centres.

It was only recently that Samsung reached a US$1tn market capitalisation milestone (May 2026) alongside Micron and SK Hynix, which are its direct competitors in the memory chip industry.

AI demand reshapes memory supply for data centres

Samsung has released its earnings guidance before its full detailed report is set to arrive later this month.

The preview comes as semiconductor demand exceeds supply, which is creating pressure across the market and increasing memory chip prices.

Data centre operators expanding AI infrastructure require advanced memory components to support particularly demanding workloads – but there is one problem: demand outweighs supply.

Samsung has responded by increasing prices for its memory products.

The company recorded around ₩17.1tn (US$11.17bn) in sales during the quarter, more than double the amount from the same period last year.

Revenue has also risen by 129% year-on-year.

Marc Einstein, Industry Analyst at Counterpoint Research, told the BBC that the projected earnings mark one of "the best quarterly performances ever".

He added that the performance is close to the technology sector record set by NVIDIA earlier this year.

“This has everything to do with the AI boom as memory companies continue to ride a tidal wave driven by limited supply and unprecedented demand,” Marc said.

Meanwhile, the research firm IDC reports that demand for data centres and AI infrastructure is different from anything the memory industry has navigated before.

This ever-rising demand, correlating with an increase of high-density workloads, has had a major impact on the supply of chips used for everyday electronics.

How HBM production changes the data centre supply chain

The growth of generative AI depends heavily on high-bandwidth memory (HBM), a specialised memory component that provides faster performance and lower power consumption.

Samsung is increasing its production capacity for HBM as AI infrastructure development continues.

The company has moved more production capacity towards vertically stacked high-performance chips, reducing the market availability of standard storage components.

As AI investment expands beyond HBM into conventional DRAM and NAND products, memory prices continue to rise. Citi Research reported that average selling prices for DRAM and NAND have risen 44% and 53% quarter-on-quarter, respectively.

Analysts say the rapid growth of HBM production tightens the supply of conventional memory products. These components support smartphones, personal computers and enterprise servers, including systems used in data centre environments.

HBM chips integrate closely with AI processors, allowing faster data movement while using less power. Their role in AI systems gives suppliers greater pricing power compared with standard memory products.

Samsung’s AI growth faces future supply questions

Samsung made its profit forecast while also setting aside funds for employee bonuses.

Following union worker strikes, Samsung agreed to a wage deal in May linking employee pay to operating profit.

Analysts have told Reuters that without those provisions, operating profit would likely exceed ₩10tn (US$6.54bn).

Losses from its foundry and logic chip businesses are likely to widen because bonus expenses apply across the entire semiconductor division.

If spending on AI data centres slows, repeating similar profit increases may become harder to achieve.

The availability of advanced memory components remains closely linked to the expansion of AI systems and the ability of data centres to support growing workloads.