On the morning of October 28, SK hynix released record-breaking financial results, sending its shares sharply higher.
The South Korean stock market subsequently triggered a buy-side sidecar, halting program trading buy orders for five minutes.
Strong earnings and a large shareholder return plan became the key catalysts for this rally.
But another set of numbers is more worth pondering: previously, SK hynix shares had retreated by roughly half from their June high.
A mid-October report on a Bank of America research note showed that, based on 2027-2028 earnings forecasts, its price-to-earnings ratio was in the range of two to three times.
This was the forward valuation basis at that time, not a real-time quote during trading after the earnings release.
Why would a company with an operating margin approaching 80% still command such a low earnings multiple?
Financial statements answer how much can be earned now, but valuation must also judge how long those earnings can last.
SK hynix may be undervalued.
But the key to making that judgment is to break apart today's profits: how much comes from technological leadership, how much comes from industry shortage, and how long each of these two forces can persist.
First, the Report Card: What Kind of Margin Is 80%?
In the third quarter of 2026, SK hynix reported revenue of 104.6 trillion Korean won and operating profit of 83.3 trillion Korean won, up 32% and 38% quarter-over-quarter respectively, both setting quarterly historical highs.
The operating margin approached 80%.
Figure 1 | Revenue and operating profit hit new highs, yet net profit declined quarter-over-quarter.
The three metrics are calculated on different bases.
This means that for every 100 won of product sold, after deducting operating costs and expenses, nearly 80 won of operating profit remains.
Only after including non-operating items, taxes, and other items does net profit emerge.
Net profit for the quarter was approximately 39.8 trillion Korean won, down 58% quarter-over-quarter.
Foreign exchange losses, valuation losses on investment assets, and the high base from large investment gains in the previous quarter all contributed to this result.
Therefore, one cannot call 80% a net profit margin, nor can one conclude that the memory business has weakened based solely on the net profit decline.
To judge earnings in the next phase, one must return to product competitiveness and supply-demand dynamics.
The Value of HBM Lies in the Capacity It Crowds Out
When discussing SK hynix, HBM — the stacked memory that provides high-bandwidth data channels for AI chips — is the first thing that comes to mind.
But this quarter, there was another important growth engine: DDR5 used in servers.
Management disclosed during the earnings call that server DDR5 sales grew more than 40% quarter-over-quarter.
Two paths are at work simultaneously here.
Server expansion itself requires more memory; on the other hand, HBM and ordinary DRAM compete for limited advanced process nodes and wafer capacity.
As resources tilt toward HBM, supply of ordinary memory also gets squeezed.
Figure 2 | The impact of hot-selling HBM transmits to ordinary memory through limited capacity.
The diagram is a mechanism illustration and does not quantify the profit contribution of the two categories.
Thus, the benefits brought by HBM are reflected not only in its own sales but also potentially in the pricing room for other memory products.
But the technology premium and the shortage dividend may not last equally long.
Product leadership requires the next generation of technology to continue delivering; the degree of shortage will change with capacity expansion, yield improvement, and demand shifts.
Even if HBM continues to grow, as long as supply gradually catches up, the pricing dividend for ordinary memory may also thin out.
Seeing this relationship clearly is more valuable than simply extrapolating the 80% margin several years forward.
High Profits Attract Competitors and Also Change Customer Choices
According to Counterpoint data, from the first quarter to the second quarter of 2026, Samsung's HBM revenue share rose from 21% to 33%; SK hynix's fell from 58% to 50%, still ranking first.
Figure 3 | The comparison is of HBM revenue share.
Share changes cannot be directly equated with changes in sales volume or yield.
Leadership still exists, but peers are also sharing in the growth.
The richer the excess profits, the more incentive competitors have to invest capital, improve yields, and win customers.
In manufacturing, an 80% operating margin also looks like an invitation to expand capacity.
The customer side is also changing.
Nvidia's NVHBM solution announced in August integrates its own memory controller into the HBM base die and pushes multiple memory suppliers to provide a unified implementation, reducing the engineering burden on customers to adapt to and certify different suppliers.
This gives customers more choices and also means memory makers need to continuously prove the value of next-generation products.
But it is not enough to prove that SK hynix has lost pricing power, nor can it be simply deduced that memory makers are reduced to a foundry role.
New capacity must go through fab construction, equipment installation, yield improvement, and customer certification, and customers cannot switch at zero cost.
Intensified competition does not equal an immediate price collapse.
What is truly worth tracking is whether new demand or effective supply is running faster.
Long-Term Agreements Extend Orders, But Not Necessarily the High Margins
Long-term agreements are an important basis for optimism about SK hynix.
The company has signed LTAs with 14 major customers, typically for five years, locking in supply volumes covering more than 50% of total capacity for 2027 and 2028.
This makes future sales volume easier to judge and also provides a basis for capacity expansion.
But price terms are equally critical: the company disclosed that prices are adjusted within a range agreed upon by both parties, based on market conditions.
Figure 4 | Sales visibility improves, but price still has an adjustment mechanism.
The illustrative price band does not represent actual contract prices or upper and lower limits.
Therefore, describing these contracts as fully market-linked would underestimate their protective effect; describing them as locking in a future 80% margin would overestimate the degree of protection.
Customers willing to place orders in advance does not mean they are willing to bear the current high margins indefinitely.
How strong the specific protection is depends on the price range, adjustment mechanism, and performance terms that have not yet been fully disclosed.
Orders can have a long shelf life; the shelf life of excess profits still needs verification.
Behind Dividends and Capacity Expansion Lies the Same Question of Timing
The company judges that supply tightness may persist until 2030 and plans to significantly increase investment next year.
At the same time, it plans to return approximately 100 trillion Korean won to shareholders this year, including a special dividend from sharing in Kioxia investment gains.
These two arrangements can coexist.
The company must both prepare future capacity and return operating and investment results to shareholders.
Merely based on dividends and buybacks, one cannot infer that management is bearish on the future.
What is truly difficult is: with today's shortage, how much capacity should be built for the market several years from now?
Figure 5 | Capacity expansion has a time lag.
Price pressure is a scenario when demand fails to keep up with supply, not a certain prediction of future prices.
Building fabs, installing equipment, and improving yields all take time.
Today's high profits will drive multiple companies to expand simultaneously; by the time capacity is actually released, if demand growth falls short of expectations, shortage may turn into oversupply, and price increases may turn into price wars.
Therefore, to observe investment discipline, one cannot look only at whether capital expenditure has increased.
More importantly, one must see whether the company can adjust investment scale and production pace in line with order and supply-demand changes.
Expand too slowly and it may miss growth; expand too quickly and it may weaken the scarcity that supports high profits.
SK hynix also explicitly stated in its financial report that it will flexibly adjust investment scale and timing based on customer demand, product competitiveness, profitability, and investment efficiency.
Whether this commitment can be fulfilled will only be seen when supply and demand change.
Demand also has a price boundary.
Memory price increases raise the cost of terminals such as PCs and phones, which may in turn suppress shipments.
Therefore, judging how long the shortage can last requires looking not only at AI server expansion but also at whether other markets can bear higher costs.
True Cheapness Must Withstand Falling Profits
Returning to the opening question: why is the low P/E ratio still controversial when the financial report is so strong?
Because the denominator of the P/E ratio is earnings, and earnings in the memory industry change.
Here is a purely mathematical example: a stock trades at 300 won with earnings per share of 100 won, giving a P/E of 3 times; if earnings per share fall to 50 won and the stock price stays unchanged, the P/E becomes 6 times.
Figure 6 | Hypothetical example, unrelated to SK hynix's actual stock price, and not an earnings forecast.
A low multiple could mean the market is overly pessimistic, or it could mean current or forecast high profits are inflating the denominator.
Based solely on a P/E of two to three times, one cannot directly determine which it is.
But this does not mean that once margins decline, the undervaluation thesis necessarily fails.
If technological advantages remain, and profit scale and cash flow are strong enough, even if margins leave their peak, the current price may still undervalue the company.
Therefore, a more persuasive judgment requires answering three specific questions: Can next-generation HBM maintain its advantage? Can new demand continue to absorb new supply? After margins decline, can the remaining profits and cash flow support the current valuation?
SK hynix's current strength has already been proven by its financial report.
What truly needs verification is how strong it can remain when scarcity is no longer so scarce.
Do you value SK hynix's technological advantage more, or the duration of this memory shortage?
Feel free to share your judgment in the comments section.
Data as of October 28, 2026.
Financial figures are rounded; the stock price drawdown and low P/E ratio use the time points and forecast basis noted in the text.
The cover is an AI-generated conceptual illustration, and the charts in the text are drawn based on disclosed data and analytical mechanisms.
Comments