SK Hynix's new SKHY Nasdaq ADR: what investors need to know

What SKHY actually represents#

SK Hynix's final SEC prospectus says it offered 177.9 million American depositary shares, or ADSs, at US$149 each. Every ADS represents one-tenth of one SK Hynix common share. Put another way, ten SKHY shares represent one Korea Exchange common share before allowing for depositary terms, fees, taxes and currency conversion.

The Korean common shares continue to trade on the Korea Exchange, which the prospectus describes as the principal trading market. Citibank is the depositary that holds the underlying common shares through a custodian. ADS holders receive rights through the deposit agreement rather than holding the Korean shares directly; for example, they may instruct the depositary on voting and receive declared distributions after applicable fees and expenses.

The temporary ticker added to the confusion. Nasdaq's trader notice says SKHYV was the when-issued symbol on 10 July. It changed to SKHY for regular-way trading on 13 July, with settlement of the when-issued trades expected on 14 July. SKHYV was not a different class of operating business or a second permanent ticker.

What changed for SK Hynix's capital structure#

The offering was a primary capital raise, not merely a wrapper placed around existing shares. SK Hynix issued 17.79 million new common shares to support the 177.9 million ADSs. The prospectus estimates net proceeds of about US$26.2 billion after underwriting discounts, commissions and offering expenses.

The SEC filing sequence provides four point-in-time documents for the transaction: the original F-1 filed on 24 June, amendments filed on 30 June and 6 July, and the final 424B4 prospectus filed on 10 July. The 6 July amendment estimated about US$28.0 billion of net proceeds for the same 17.79 million common shares; the final prospectus reduced the estimate to US$26.2 billion after the US$149 ADS price was fixed. Transaction analysis should therefore anchor to the latest filing rather than carry a draft proceeds figure forward.

The prospectus says SK Hynix expects 728.87 million common shares to be outstanding immediately after the offering, including shares represented by ADSs. Relative to the implied pre-offering count, the new shares increase the share count by about 2.5%. Existing owners therefore hold a smaller percentage claim on the company, while the company receives a very large pool of cash for investment. Both sides of that equation matter; describing the deal only as dilution ignores the new capital, while describing it only as funding ignores the additional shares.

SK Hynix allocated the proceeds toward Korean production infrastructure and equipment. Its prospectus lists approximately US$17.8 billion of planned investment for the first semiconductor fabrication plant, or fab, at the Yongin complex and US$12.6 billion for the Package & Test 7 (P&T7) advanced-packaging plant in Cheongju. It also expects to acquire about US$8.0 billion of extreme-ultraviolet (EUV) lithography scanners for delivery by December 2027. The filing reports those amounts as KRW26.6 trillion, KRW18.9 trillion and KRW11.9 trillion respectively; the US-dollar figures use the article's presentation rate. Costs beyond the offering proceeds are expected to come from operating cash flow, credit facilities, debt securities or other funding.

The raise gives SK Hynix more financing capacity to build advanced memory and packaging capacity without relying entirely on future operating cash or debt. That can support long-term output, but it is not immediate revenue. The prospectus schedules planned use-of-proceeds spending through 2030, but that is not the construction timetable. It separately says the phase-one cleanroom of Yongin's first fab is expected to open in the first quarter of 2027 and P&T7 construction is expected to finish by the end of 2027. Later spending, equipment installation and capacity ramps can continue after those milestones.

What the filing-backed data says about the business#

The final prospectus reports 2025 revenue equivalent to about US$65.0 billion at the article's presentation rate, from KRW97.147 trillion as filed. Dynamic random-access memory (DRAM), the working memory used in servers and devices, generated about US$50.1 billion, or 77.1% of total revenue, after growing 67.5% year on year. NAND flash, the non-volatile memory used for persistent storage, generated about US$13.8 billion, or 21.3%, after growing 7.3%. The filed product amounts were KRW74.904 trillion and KRW20.690 trillion respectively. In the first quarter of 2026, the mix remained similar: DRAM was 77.3% of sales and NAND flash was 22.0%.

Calling SK Hynix an AI-memory company is directionally useful but incomplete. High-bandwidth memory (HBM), the stacked DRAM used alongside AI processors, sits inside the much larger DRAM revenue engine, and DRAM supplied most of the 2025 growth. Investors therefore need to separate DRAM and NAND shipment growth, average selling prices and product mix rather than treat every memory category as if it were contributing equally.

The filing also highlights the cash-versus-buildout tension. SK Hynix generated about US$35.7 billion of operating cash flow in 2025 and spent about US$18.4 billion on capital expenditure. The simple difference is about US$17.3 billion; that subtraction is an analytical operating-cash-flow-minus-capex proxy, not a separately reported company free-cash-flow line or a forecast. Capital expenditure had risen from about US$5.6 billion in 2023 to US$10.7 billion in 2024, and the prospectus says the company plans to increase it considerably again in 2026.

That evidence shows a company entering the offering with substantial cash generation rather than using the raise to cover current operating burn. It does not establish the return on the next investment cycle. New fabs can consume cash years before reaching useful output, and the prospectus warns about approval delays, equipment-vendor problems, raw-material constraints and cost overruns. It also says lead times for some high-end equipment can exceed one year when demand is high.

Customer concentration is another boundary on the growth story. SK Hynix says its two largest customers represented 14.8% and 12.4% of first-quarter 2026 revenue, while its largest customer represented 23.9% of 2025 revenue. Those are disclosed exposures, not evidence that a customer loss has occurred. They do mean that order levels and purchasing decisions at a small number of large customers can materially influence how quickly new capacity converts into revenue.

Growth, revenue and valuation across memory stocks#

A peer comparison can sharpen the picture, but only if it is not treated as a ranking. SK Hynix reports calendar periods in Korean won under IFRS, while the US-listed peers below report different fiscal periods in US dollars under US GAAP. Their figures are operating cross-checks, not directly comparable valuation inputs. They neither replace company-specific research nor constitute recommendations.

The revenue scale is the first useful distinction. SK Hynix's first-quarter results show revenue increasing 198% year on year to approximately US$35.168 billion after translation, from KRW52.576 trillion as filed. Micron's latest quarter grew 346% to US$41.456 billion, while Sandisk grew 251% to US$5.950 billion. On a TTM basis, SK Hynix's translated US$88.350 billion and Micron's US$90.274 billion were much closer in revenue scale than either was to Sandisk. These are extraordinary rates from different reporting periods and depressed comparison bases; they do not show equivalent volume growth or make the companies interchangeable.

Micron Technology (Nasdaq: MU) is the closest US-listed mixed-memory comparison. For the 13 weeks ended 28 May 2026, Micron reported US$41.456 billion of revenue: DRAM supplied US$31.328 billion, or 76%, and NAND supplied US$9.943 billion, or 24%. That mix is strikingly close to SK Hynix's first-quarter split of 77.3% DRAM and 22.0% NAND, although the reporting periods do not align. Both companies include HBM within DRAM; Micron also said its fourth-generation high-bandwidth memory product, HBM4, was in high-volume shipment for its lead customer's platform.

The more useful Micron comparison is what drove the growth. Its latest 10-Q says sequential DRAM sales rose 67% as average selling prices increased in the low-60% range and bit shipments rose only in the low-single-digit range. NAND sales rose 99% as prices increased in the mid-80% range and bits rose in the mid-single digits. Micron's GAAP gross margin moved to 84.6% from 37.7% a year earlier. It generated US$25.39 billion of operating cash flow and reported US$7.1 billion of company-defined net capital expenditure and US$18.3 billion of adjusted free cash flow. Those last two measures incorporate Micron's definitions and incentives, so they should not be placed directly beside the simple SK Hynix operating-cash-flow-minus-capex proxy used above.

Sandisk (Nasdaq: SNDK) is the cleaner current NAND-only comparison after its 2025 separation from Western Digital. For the quarter ended 3 April 2026, it reported US$5.950 billion of revenue and a 78.4% GAAP gross margin. Datacenter contributed US$1.467 billion, Edge US$3.663 billion and Consumer US$820 million. Sandisk's 10-Q supplies the critical context: revenue rose 251% year on year as average selling price per gigabyte rose 248%, while total exabytes sold were flat. Sandisk therefore corroborates a sharp improvement in NAND economics, not uniform volume growth across memory end markets.

Sandisk's capital expenditure also cannot be read like SK Hynix's. Sandisk generated US$3.038 billion of operating cash flow in the quarter and reported US$45 million of property, plant and equipment purchases, but it procures substantially all of its flash wafers through manufacturing ventures with Kioxia. It generally bears half of those ventures' fixed costs and is committed to fund roughly half of their capital investment when venture cash flow is insufficient. Its top ten customers also represented 46% of quarterly revenue, with one customer above 10%. The accounting perimeter and customer mix matter as much as the headline cash figures.

The valuation snapshot deliberately uses equity value divided by TTM revenue rather than enterprise value because SK Hynix had just raised about US$26.2 billion net and Sandisk's joint-venture funding obligations do not behave like conventional fab debt. Using the latest filed basic share counts, SK Hynix's 14 July Korean close implied about US$932.4 billion of translated equity value, or 10.6 times TTM revenue. The 13 July US closes of US$979.31 for Micron and US$1,916.74 for Sandisk implied about US$1.106 trillion and US$283.8 billion respectively, or 12.3 and 21.5 times TTM revenue. These are dated market snapshots, not estimates of intrinsic value.

SKHY itself shows why listing mechanics matter. Its US$152.35 close on 13 July, extended across the ADS equivalent of the full post-offering share base and translated using the article's dated won-dollar rate, implied about US$1.110 trillion of equity value and 12.6 times TTM revenue—roughly 19% above the Korean-share equivalent. The two values describe the same company and must not be added together. That gap is an early-listing price dislocation, not a conclusion about relative value.

None of the sales multiples identifies a "cheaper" memory stock. They omit margins, cash, debt, capital intensity and the durability of current pricing, and a strong cycle can make annualised or trailing comparisons look more settled than the underlying business. Their useful role is narrower: showing how much equity value the market was assigning to each dollar of recently reported revenue at a specific time.

The relationship is not purely competitive. SK Hynix and Sandisk began an Open Compute Project workstream in February to standardise High Bandwidth Flash (HBF), a proposed layer between HBM and solid-state drives (SSDs) for AI inference. That is a confirmed technical-development programme, not reported commercial revenue. Taken together, the peer evidence does not identify a "better" memory stock or establish relative value. It creates more precise tests for SK Hynix: separate selling prices from shipped bits, track HBM ramps and yields inside DRAM, examine NAND end-market mix, and adjust capital-spending comparisons for how manufacturing is financed.

What the Nasdaq listing does not change#

The listing does not create a new US subsidiary, transfer the operating headquarters from Korea or replace the Korean common shares. SKHY remains an interest in the same consolidated SK Hynix business. Nasdaq access may broaden the shareholder base and dollar liquidity, but it does not alter the underlying demand, manufacturing yields or profitability of DRAM, high-bandwidth memory and NAND.

An ADS price also need not match exactly one-tenth of the Korean share price at every moment. The two securities trade in different currencies, market hours and settlement systems. Foreign-exchange moves, depositary mechanics, conversion restrictions, fees and temporary supply-and-demand imbalances can produce a premium or discount. The economic comparison therefore needs the Korean price and the live won-dollar exchange rate, not the two quoted prices in isolation.

The offering itself did not provide a new earnings forecast. SK Hynix's latest completed quarter remains the operating baseline: for the first quarter of 2026, the company reported figures equivalent to about US$35.2 billion of revenue, US$25.2 billion of operating profit, a 72% operating margin and US$27.0 billion of net profit under the article's presentation rate. It also reported net cash equivalent to about US$23.4 billion. The company attributed the performance to strong AI demand and higher-value products, but a strong starting quarter does not remove the memory industry's pricing cycles, customer concentration or execution risk on a much larger capital programme.

The metrics investors can monitor#

Start with the capital raise itself: total common shares outstanding, the amount and timing of capital expenditure, construction progress at Yongin and P&T7, EUV scanner deliveries, and any change to the uses of proceeds. These measures show whether the new equity is becoming productive capacity on the schedule described in the prospectus.

For the operating business, monitor quarterly revenue, operating margin, cash generated from operations, net cash and the gap between operating cash flow and capital expenditure. DRAM and NAND shipment growth and average selling prices, HBM product ramps, manufacturing yields and the revenue share of the largest customers help explain whether growth comes from volume, pricing or mix. The 2025 baselines—77.1% of revenue from DRAM, about US$35.7 billion of operating cash flow and US$18.4 billion of capital expenditure under the article's presentation rate—make later changes measurable rather than anecdotal.

Use peer disclosures as cross-checks rather than trading signals. Micron's split between selling prices and bit shipments can test whether a move is industry-wide, while Sandisk's price per gigabyte, exabytes and end-market mix provide a separate NAND lens. If SK Hynix diverges, product mix, HBM yields, customer timing, foreign exchange and manufacturing execution should be investigated before treating that divergence as positive or negative.

For the US security, monitor SKHY trading volume and liquidity, the premium or discount to one-tenth of the currency-adjusted Korean common-share price, depositary fees and the treatment of dividends. These are security-structure measures, not evidence that the underlying business improved or weakened.

The most useful interpretation is conditional. If capacity milestones progress while margins and cash generation remain resilient, the financing will have supported a structural expansion. If spending rises faster than demand, pricing or yields can support, the same raise may expose shareholders to the downside of the memory cycle. The listing supplies capital and access; later operating results determine what that capital earns.