Key Takeaways
- Samsung projects Q2 operating profit of KRW 89.4 trillion ($66 billion), a 19-fold year-over-year increase
- AI memory chip demand has driven Samsung's market cap past $1 trillion, with shares doubling in 12 months
- Samsung and SK Hynix plan a $1 trillion domestic semiconductor investment to expand AI chip capacity
Samsung Electronics expects its Q2 operating profit to hit KRW 89.4 trillion (roughly $66 billion), a 19-fold increase from the same quarter in 2024. The driver is straightforward: AI infrastructure providers cannot get enough high-bandwidth memory chips, and Samsung makes more of them than anyone else.
Revenue for the quarter is forecast to exceed KRW 171 trillion ($113 billion), more than double the KRW 74.6 trillion Samsung posted a year ago. This marks the third consecutive quarter of record operating profit for the company, a streak built almost entirely on the AI chip boom.
Why AI memory chips are minting money
Every AI accelerator, whether from NVIDIA, AMD, or Google, needs High Bandwidth Memory (HBM) to feed data to the processor fast enough. HBM stacks memory chips vertically and connects them with silicon interposers, delivering bandwidth that standard DRAM cannot match. Training large language models and running inference at scale both require this specialized memory.
Samsung and SK Hynix dominate HBM production. The global HBM market is projected to reach $47 billion by 2028, growing at roughly 30% annually. Right now, demand outstrips supply. Data center operators are placing orders months in advance, paying premium prices, and still waiting.
Samsung's shares have doubled over the past year, pushing its market capitalization past $1 trillion. For a company that was struggling with memory chip oversupply just two years ago, the reversal is striking.
Samsung and SK Hynix bet $1 trillion on future demand
Last week, Samsung and SK Hynix announced plans to lead a domestic semiconductor and AI investment program worth around $1 trillion. The goal: expand capacity for advanced memory chips before competitors catch up.
SK Hynix moved fast. The company launched a US IPO on July 6, seeking to raise KRW 43 trillion (approximately $32 billion) to fund its own expansion. SK Hynix has been particularly strong in HBM3E, the latest generation of high-bandwidth memory, and counts NVIDIA as a major customer.
Samsung is playing catch-up on HBM3E qualification with some customers but holds advantages in overall production scale and vertical integration. The company manufactures its own logic chips, memory chips, and packaging, which should help margins as HBM production matures.
What the numbers actually show
Samsung's full Q2 results are due later this month. The guidance figures, while dramatic, are preliminary. Still, the trajectory is clear. Three quarters of record profits is not a blip. It reflects a structural shift in how technology companies allocate capital.
The AI infrastructure buildout is absorbing enormous amounts of high-end memory. Microsoft, Google, Amazon, and Meta are all racing to deploy more training and inference capacity. Startups like OpenAI and Anthropic need the same chips. So do sovereign AI projects in Saudi Arabia, UAE, and India.
For Samsung, this means pricing power. Memory chips are typically a cyclical business with brutal price swings. AI demand has temporarily suspended that cycle. The question is how long it lasts.
Risks worth watching
The obvious risk is a demand pullback. If AI investment slows, memory prices will drop fast. Samsung's Q2 numbers depend on hyperscalers continuing to spend. A recession or a shift in investor sentiment toward AI could change that.
Competition is intensifying too. Micron is ramping HBM production. Chinese memory makers, despite US export controls, are working on their own versions. Samsung's lead is real but not permanent.
Geopolitics adds uncertainty. US-China semiconductor restrictions have already reshaped supply chains. Further restrictions could affect Samsung's ability to sell certain chips or source certain equipment.
Logicity's Take
Samsung's earnings tell AI builders something important about their own cost structure: memory is the bottleneck. HBM prices are elevated because demand exceeds supply, and that feeds directly into the cost of training runs and inference deployments. Teams building AI products should factor in memory constraints when planning capacity. If you are evaluating cloud providers for AI workloads, compare not just GPU availability but memory bandwidth and pricing. Tools like [DigitalOcean](https://logicity.in/r/digitalocean) for standard workloads and specialized GPU cloud providers for training will have different cost curves depending on how Samsung and SK Hynix manage supply over the next 18 months.
Disclosure
Some links in this post are affiliate links — Logicity earns a commission if you sign up, at no extra cost to you. We only link products we have used or actively recommend.
Frequently Asked Questions
Why is Samsung's profit increasing so much from AI?
AI training and inference require High Bandwidth Memory (HBM) chips that only Samsung, SK Hynix, and Micron produce in volume. Demand from data centers far exceeds current supply, letting Samsung charge premium prices.
How does Samsung compare to SK Hynix in AI memory chips?
SK Hynix has led in HBM3E qualification with NVIDIA, while Samsung has broader production scale and vertical integration. Both companies are investing heavily to expand capacity.
What is HBM and why does AI need it?
High Bandwidth Memory stacks DRAM chips vertically with silicon interposers, delivering far more bandwidth than standard memory. Large language models and AI accelerators need this bandwidth to process data fast enough for training and inference.
Could Samsung's AI chip profits decline?
Yes. Memory is cyclical, and a slowdown in AI infrastructure spending would reduce demand. Competition from Micron and Chinese manufacturers could also pressure prices as supply increases.
Shows how AI investment translates to business results beyond chip makers
Need Help Implementing This?
If you are building AI products and need guidance on infrastructure planning, cloud provider selection, or cost optimization for memory-intensive workloads, reach out to the Logicity team for a consultation.
Source: TahawulTech.com / Daniel Shepherd
Manaal Khan
Tech & Innovation Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
