Samsung profit hits $39b on strong AI chip demand

Samsung Electronics said on April 30 that first-quarter operating profit rose more than eight times from a year earlier to 57.2 trillion won (US$38.9 billion).

It is setting a record as strong AI server demand and tight memory chip supply lifted prices and earnings.

The result beat analysts’ estimates tracked by LSEG SmartEstimate and topped Samsung’s full-year 2025 operating profit of 43.6 trillion won (US$29.6 billion).

Revenue also hit a record and rose about 70% year on year.

Samsung said its memory business set a quarterly sales record as customers bought more higher-value chips for AI systems.

It expects server memory demand to stay strong in the second half as hyperscalers expand AI capacity.

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

Samsung’s profit rests on a product where it still trails

High Bandwidth Memory (HBM) powers the profit jump. AI servers and accelerators use this memory because it moves data fast with lower power use 1.
Samsung’s memory unit logged a second straight quarterly earnings record, even as the company still trails rivals in HBM 2.
SK hynix and Micron moved to a more advanced process for HBM3E. Samsung faced delays getting its own chip approved by large customers 3.
Samsung also appears to get fewer usable HBM chips from each batch than peers. That raises production risk as strong demand pushes prices and profit higher 1.

The memory bottleneck will shape the next wave of AI

Samsung’s results make clear that HBM supply now limits AI hardware growth along with graphics processing unit (GPU) output 3.
Next-generation AI accelerators, including Nvidia’s Rubin design, rely on HBM from a small group of suppliers that includes Samsung 3.
Output cannot ramp quickly. Standard memory lines need specialized tools to add the chips’ vertical connections, and that has become the tightest capacity constraint 1.
Samsung expects capital spending to climb, part of a wider industry effort to add advanced memory capacity for AI demand 2.