SK Hynix blew past expectations in Q3 as a surge in AI-related memory demand turned into record sales and profits.
The South Korean memory giant reported revenue of 104.6 trillion won (about $78 billion) and operating profit of 83.3 trillion won (roughly $65.5 billion) — the first quarter its sales topped 100 trillion won. Year‑over‑year revenue jumped 328% and operating profit surged 632%, producing an extraordinary 80% operating margin (about 80 won of operating profit per 100 won of sales). Bloomberg had forecast operating profit of 77.5 trillion won. Net income was 39.8 trillion won, up 216% versus a year earlier but down 58% from Q2, after the company absorbed significant non‑operating hits.
CFO Kim Woo‑hyun said those hits included 32.6 trillion won in net non‑operating expenses, driven in part by an 11.6 trillion won currency loss and a 22.9 trillion won investment valuation loss. Those items depressed net profit even as core chip sales surged.
The core message to investors: memory demand — especially for AI — is far outstripping supply and is expected to remain tight through 2030. SK Hynix says DRAM and HBM volumes through 2029 are effectively sold out. DRAM is the mainstream memory used in servers, phones and PCs; HBM (high‑bandwidth memory) is the stacked, ultra‑fast memory that sits alongside AI accelerators and feeds models with data.
“Given the complexities involved in implementing leading‑edge processes and the lead times required to construct new production facilities,” Kim said, “a situation where supply cannot keep up with demand is expected to persist for some time, a constraint projected to continue through 2030.” The company noted that some customers are already adjusting HBM specifications — a sign, SK Hynix says, that buyers are trying to stretch limited chip supplies to build more AI accelerators. On the earnings call, SK Hynix said customer volume requests are “more than double” what it can supply today.
To lock in future revenue and allocate scarce wafers, SK Hynix has signed long‑term supply agreements with 14 key customers, including major cloud providers. These contracts — typically five years — already reserve more than half of the company’s planned production for 2027–2028, feature price bands and include safeguards such as deposits.
Looking ahead, SK Hynix expects a major investment uptick in 2027 as the first phase of its Yongin complex ramps to full‑scale operation.
Market skeptics remain. Citigroup’s Peter Lee asked how SK Hynix can be confident in sustained AI investment when commercial monetization is still uncertain — echoing doubts like those raised by Citadel’s Ken Griffin, who recently warned of an AI bubble. CEO Kwak Noh‑Jung pushed back, pointing to robust cloud indicators: major cloud customers reported year‑over‑year revenue growth of 30%–80% and order backlogs that are two to five times larger. Kwak also said AI spending is broadening beyond hyperscalers into non‑tech firms, sovereign projects and on‑prem deployments, reducing the risk tied to any single buyer slowing down.
Still, financing is getting tighter for infrastructure builders: Elon Musk noted on Oct. 25 that debt markets have largely closed to firms without significant non‑AI revenues. SK Hynix expects demand from PCs and smartphones to soften because of higher memory prices and supply constraints, but says the boom in AI infrastructure should more than offset that decline.
Why crypto readers should care: a sustained shortage of DRAM and HBM and the long lead times for new fabs mean higher and more volatile memory prices for years. That raises the cost of cloud‑based AI services and on‑prem AI deployments — relevant to crypto projects using AI for analytics, trading, or developer tooling — and could constrain the rollout of next‑gen AI accelerators that underpin both web2 and web3 services.
The South Korean memory giant reported revenue of 104.6 trillion won (about $78 billion) and operating profit of 83.3 trillion won (roughly $65.5 billion) — the first quarter its sales topped 100 trillion won. Year‑over‑year revenue jumped 328% and operating profit surged 632%, producing an extraordinary 80% operating margin (about 80 won of operating profit per 100 won of sales). Bloomberg had forecast operating profit of 77.5 trillion won. Net income was 39.8 trillion won, up 216% versus a year earlier but down 58% from Q2, after the company absorbed significant non‑operating hits.
CFO Kim Woo‑hyun said those hits included 32.6 trillion won in net non‑operating expenses, driven in part by an 11.6 trillion won currency loss and a 22.9 trillion won investment valuation loss. Those items depressed net profit even as core chip sales surged.
The core message to investors: memory demand — especially for AI — is far outstripping supply and is expected to remain tight through 2030. SK Hynix says DRAM and HBM volumes through 2029 are effectively sold out. DRAM is the mainstream memory used in servers, phones and PCs; HBM (high‑bandwidth memory) is the stacked, ultra‑fast memory that sits alongside AI accelerators and feeds models with data.
“Given the complexities involved in implementing leading‑edge processes and the lead times required to construct new production facilities,” Kim said, “a situation where supply cannot keep up with demand is expected to persist for some time, a constraint projected to continue through 2030.” The company noted that some customers are already adjusting HBM specifications — a sign, SK Hynix says, that buyers are trying to stretch limited chip supplies to build more AI accelerators. On the earnings call, SK Hynix said customer volume requests are “more than double” what it can supply today.
To lock in future revenue and allocate scarce wafers, SK Hynix has signed long‑term supply agreements with 14 key customers, including major cloud providers. These contracts — typically five years — already reserve more than half of the company’s planned production for 2027–2028, feature price bands and include safeguards such as deposits.
Looking ahead, SK Hynix expects a major investment uptick in 2027 as the first phase of its Yongin complex ramps to full‑scale operation.
Market skeptics remain. Citigroup’s Peter Lee asked how SK Hynix can be confident in sustained AI investment when commercial monetization is still uncertain — echoing doubts like those raised by Citadel’s Ken Griffin, who recently warned of an AI bubble. CEO Kwak Noh‑Jung pushed back, pointing to robust cloud indicators: major cloud customers reported year‑over‑year revenue growth of 30%–80% and order backlogs that are two to five times larger. Kwak also said AI spending is broadening beyond hyperscalers into non‑tech firms, sovereign projects and on‑prem deployments, reducing the risk tied to any single buyer slowing down.
Still, financing is getting tighter for infrastructure builders: Elon Musk noted on Oct. 25 that debt markets have largely closed to firms without significant non‑AI revenues. SK Hynix expects demand from PCs and smartphones to soften because of higher memory prices and supply constraints, but says the boom in AI infrastructure should more than offset that decline.
Why crypto readers should care: a sustained shortage of DRAM and HBM and the long lead times for new fabs mean higher and more volatile memory prices for years. That raises the cost of cloud‑based AI services and on‑prem AI deployments — relevant to crypto projects using AI for analytics, trading, or developer tooling — and could constrain the rollout of next‑gen AI accelerators that underpin both web2 and web3 services.