Samsung is set to become the clearest winner of the AI memory boom, with analysts pointing to an about 18-fold jump in second-quarter operating profit. An LSEG SmartEstimate pegs the April to June figure at 86 trillion won ($56.35 billion), signalling a third straight record.
AI-fuelled memory crunch reshapes profit outlook
The profit surge reflects a supply squeeze that AI has intensified. Demand for inference infrastructure is outpacing output growth from global memory makers, leaving the market undersupplied at least through next year, according to analysts.
Crucially, this is not just a high-bandwidth memory story. Broader appetite for conventional DRAM and NAND has strengthened as agentic AI systems handle more complex, multi-step tasks that require more memory per server and extra storage during inference.
Samsung’s position as a key supplier to Nvidia, Google and Apple puts it at the centre of this shift. It has signed multi-year binding contracts, disclosed in April, with customers seeking to lock in supply, though details were not shared.
Pricing power returns across DRAM, NAND and HBM
Tight inventories have restored pricing power. Citi Research said second-quarter average selling prices climbed 44% for DRAM and 53% for NAND quarter-on-quarter, underscoring how quickly the AI cycle has reset memory margins.
Nomura expects the trend to extend into the July-September quarter, forecasting commodity DRAM prices to rise 24% quarter-on-quarter and NAND prices to increase 25%, supported by demand from consumer devices and both traditional and AI data centres.
Key figures shaping the market now include:
– Operating profit estimate: 86 trillion won
– Prior-year quarter: 4.7 trillion won
– DRAM pricing in Q2: up 44%
– NAND pricing in Q2: up 53%
– Next quarter DRAM view: up 24%
– Next quarter NAND view: up 25%
Investors cheer, but bonus charges and capex raise questions
Equity markets have moved ahead of earnings. Shares of Samsung Electronics, SK Hynix and Micron have rallied 158%, 273% and 242% this year, pushing each company’s market value above $1 trillion amid the shortage-driven boom.
There are near-term accounting risks. Analysts cautioned that second-quarter results could miss consensus if Samsung books a larger-than-expected provision for employee bonuses. In late May, it averted a large-scale strike with a wage deal that allocates 10.5% of the semiconductor division’s operating profit to special bonuses.
Some analysts estimate cumulative bonus provisions could exceed 40 trillion won, making the timing of recognition a key swing factor for the quarter. Samsung will announce detailed earnings later this month.
Capacity bets, cloud capex and sustainability of AI spend
Longer term, the industry’s investment plans are enormous. Samsung and SK Hynix last week pledged to invest 3,200 trillion won ($2.07 trillion) to expand chip capacity in South Korea. Samsung expects to make that investment between 2026 and 2040, while SK Hynix has not provided a timeframe.
JPMorgan flagged a central risk: whether cloud providers can sustain AI memory’s rising share of capital expenditure, estimated at 52% this year and expected to exceed 70% next year. Any slowdown in AI spending could reverberate through the memory cycle, particularly for Samsung and SK Hynix.
Investors also want clearer proof that breakthroughs in AI services will accelerate cloud revenue growth, supporting memory’s expanding share of AI infrastructure budgets. For now, scarcity and contracts have given suppliers leverage.
Downstream pricing and competitive signals
Tight memory supply is filtering into devices. Samsung has already raised smartphone prices, and analysts said further increases may be needed in the second half. Rival Apple raised prices of iPads and MacBooks last month, signalling limited relief for buyers.
The stakes are clear. If AI deployment maintains its current pace and infrastructure builds stay on track, Samsung’s record-setting profit run could continue. If capex cools, the market’s new equilibrium will be tested just as capacity expansions loom.
