Samsung’s Profit Just Jumped Eightfold on AI Chips. Here’s What That Number Actually Means

In the financial district of Seoul, a figure that stops people in their tracks is 57.2 trillion Korean won. That represents Samsung Electronics‘ anticipated operating profit for the first quarter of 2025, which is more than eight times what the company made during the same period last year and nearly three times the previous quarterly record. By all measures, it’s an incredible turnaround.

However, observing the stock’s response—an 11% increase that swiftly leveled off—there’s a persistent feeling that the market had already subtly come to terms with the notion that Samsung would take such a step.

The market for memory chips has always been cyclical. The pattern is familiar to anyone who has followed the semiconductor markets for more than a few years: supply spikes, prices plummet, businesses suffer, capacity is reduced, demand recovers, and the cycle repeats itself. Samsung has frequently benefited from that trend.

However, the current situation is more than just a cyclical recovery, which is what the eightfold profit figure is actually measuring. In a way that is truly difficult to dispute, it is the first obvious financial fingerprint of the AI infrastructure buildout to appear on the balance sheet of a conventional chip manufacturer.

High-bandwidth memory, or HBM, served as the driver. These are the chips that sit next to AI processors in data centers, supplying them with data quickly enough to meet the computational demands of workloads like large language models. Over the past year, demand for HBM has been so great that it has caused actual supply shortages, not manufactured scarcity, but a real inability to produce enough of the product quickly enough.

It was predicted that commodity memory prices would increase by over 50% in the second quarter alone, with no signs of a significant slowdown. Over half of Samsung’s operating profits come from its Device Solutions division, which has been operating in this environment.

However, it’s possible that the eightfold number hides as much as it makes clear. A year ago, Samsung was still struggling through one of the worst memory downturns in recent memory, reporting only 6.69 trillion won in operating profit for a quarter, making the comparison base exceptionally weak.

Eightfold sounds dramatic, and it is, but the low starting point contributes to the drama. While reading the headlines, it’s important to keep that context in mind, even though it doesn’t lessen what has happened.

The response from analysts, or more accurately, how expectations had changed prior to the announcement, is more instructive. With almost religious fervor, forty analysts follow Samsung. Prior to the report, 93% of them had a buy rating. The consensus price target was higher than ₩250,000, which is about 21% higher than the stock’s pre-earnings level.

These figures don’t indicate that a market was taken aback. These figures represent a market that priced in precisely the type of quarter that Samsung delivered for months before receiving confirmation that it did.

Beneath the earnings headline, there is a structural shift that may be more significant than the Q1 figure. The two leading memory manufacturers, Samsung and SK Hynix, are both pursuing long-term contracts with significant clients. Three- to five-year supply agreements with customers like Microsoft and Google, rather than the traditional quarterly and annual contracts.

A five-year DRAM supply agreement is reportedly being negotiated by SK Hynix. Similar frameworks are being developed by Samsung. Micron is going in the same direction. In a subtle way, the memory industry is attempting to break free from its own cyclical pattern by shifting from the traditional produce-then-sell model to something more akin to a foundry business, where capacity decisions are made with actual customer commitments in hand and demand is visible years ahead.

When Professor Kim Jeong-ho of KAIST pointed out that the industry is moving toward built-to-order manufacturing due to AI’s desire for personalized memory products, he put it simply. Samsung’s earnings profile changes significantly if that holds true, though it’s still unclear because AI demand is not a given at scale. The predictability of revenue increases. Inventory risk decreases.

Planning for capital expenditures becomes logical instead of reactive. The premium multiple that analysts are modeling begins to resemble arithmetic more than optimism.

However, the hard work starts right here. The price of Samsung’s stock, which is currently trading at ₩206,000, is certain. Within 1% of that range is the DCF intrinsic value. The easy re-rating, which occurs when a stock is truly cheap and the market just catches up, has already occurred.

The slower, more difficult work of execution is what’s left: maintaining high HBM yields, concluding long-term contracts, overseeing supply chains during any geopolitical disruptions, and producing quarters that support rather than contradict the AI narrative. Complications are imminent. Tensions in the Middle East affect material supply chains used in semiconductor manufacturing in ways that analyst models do not fully account for.

It’s difficult to ignore the fact that the market’s 11% celebration was measured—almost courteous—for a business that recently announced a quarterly record of this magnitude. That self-control says something. Samsung fulfilled its promise. Whether this was a peak or a new floor is the question that the market is currently silently asking, and it will ultimately determine whether analysts‘ 20%+ upside targets are correct.