Pre-call note: SK hynix looks priced for the end of the boom

Since July 1, SK hynix shares have lost nearly half their value, falling roughly 47% by October 27's close.

Yet over that same period, the spot price of the 16Gb DDR5 chip we track in Tessara has risen 31%.

Analysts haven't cut their earnings expectations nearly as much. Q3 revenue forecasts are down about 5% from our July snapshot, while Q4 estimates have fallen just 2%.

That's a striking disconnect.

Our read is the market isn’t betting on a weak quarter. It is betting that the memory boom won’t last.

That makes SK’s October 28 earnings particularly interesting. It’s going to be about 1 critical question: Is there any evidence that the boom is coming to an end?

So far, rising memory prices, strong customer commitments and tight supply suggest otherwise.

We've been following SK hynix closely, and we should also own our last miss.

Five days before its July results, we published a view in Tessara assigning a 65% probability to operating profit of ₩65 trillion or more. SK delivered ₩60.5 trillion.

What went wrong? We assumed industry DRAM price increases would flow through to SK's reported selling prices at roughly the same rate as in Q1. They didn't.

That lesson matters for tomorrow's results.

Can SK clear that bar?

The industry backdrop looks favorable. TrendForce forecast conventional DRAM contract prices rising 13–18% in Q3, while SK guided to roughly 10% more DRAM bit shipments.

Higher prices and more volume. That's usually a good combination.

But here's what July taught us: Industry price increases don't translate directly into SK's revenue.

Much of its output is sold under longer-term contracts, so the benefit from rising market prices can arrive more slowly. And the Street is already expecting 24% sequential revenue growth in won.

Conventional DRAM alone may not get SK there. HBM, NAND and product mix will have to contribute.

Korea's customs data offers some encouragement, with DRAM and HBM-class exports reportedly rising 34% in dollar terms in Q3. But those are industry-wide exports, not SK's own sales.

The question is how much of that strength SK actually captured.

Our base case is an in-line result or modest beat. We're more interested in whether the underlying profitability holds up, and how much of that strength can carry into 2027.

The Q3 numbers matter, but what management says about the next few quarters could matter far more for the stock.

The Street expects approximately ₩110.7 trillion in Q4 revenue, another 13% increase from Q3 consensus.

TrendForce forecasts conventional DRAM contract prices rising another 10–15% in Q4, although longer-term contracts could delay how quickly suppliers benefit.

We'll be listening for two things:

For a useful comparison, look at Micron's September 30 earnings.

Micron guided next-quarter revenue to $61.5 billion and said agreements already covered the vast majority of its 2027 HBM bits at significantly higher prices.

That's encouraging for the industry. But how much of SK's own 2027 HBM volume and pricing is locked in?

That may be the most valuable thing management can tell us tomorrow.

There's another lesson from Micron worth paying attention to.

Its cloud-memory gross margin was flat sequentially despite higher pricing, because a greater share of HBM sales offset those gains.

More HBM doesn't automatically mean better margins.

For SK, the important question isn't just how many HBM chips it can sell, but how much profit it earns on the expensive manufacturing capacity those chips consume.

We've had our full SK hynix pre-call research page live in Tessara since October 26.

It sets out three tests with clear thresholds, written before the results.

For example, a Q3 operating margin of 78% or more would support the view that HBM4 and NAND mix are sustaining profitability. A fall below Q2's 76% would put that view in doubt.

We've also identified five questions for the call, including how much capex goes toward upgrading existing production versus adding new capacity, and whether share buybacks accelerate.

It's easy to declare that memory has reached "peak cycle." Much harder to identify the evidence that would actually prove it.

We're watching four signals across the supply chain.

For now, we don't see convincing evidence that the boom is about to end.

HBM pricing remains firm. Advanced packaging demand still exceeds supply. Equipment orders are growing.

But it's not all bullish.

Server-module prices have stopped rising, at least for now, and consumer demand is under pressure.

The distinction matters: Memory supply can remain tight even as earnings growth begins to slow.

That's why we want to watch the entire supply chain, rather than rely on one price chart or one company's quarterly results.

There is a credible bear argument, and it has less to do with tomorrow's earnings than with who ultimately captures the profits from HBM.

Vikram Sekar argues that custom base dies, the logic chips sitting beneath HBM memory stacks, could shift more bargaining power toward the companies designing AI accelerators.

Samsung gaining HBM market share is another risk.

If customers gain more leverage as competition and capacity increase, SK could keep shipping record volumes without maintaining today's extraordinary margins.

A booming industry doesn't guarantee that every supplier keeps its share of the profits.

That's the longer-term risk we take seriously, even if the near-term supply picture remains tight.

Here's how we'll judge the results.

If SK confirms stronger 2027 contracts alongside disciplined investment, it would strengthen the case that today's tight memory market can support earnings for longer than investors expect.

That would challenge the market's apparent bet that today's profits are about to fade.

A strong quarter without better visibility into 2027, however, wouldn't settle the debate.

We want our research to be something readers can check against what actually happens. Read our SK Hynix earnings brief.

Try this question in Tessara: What do SK hynix's customer commitments and Micron's latest agreements tell us about 2027 HBM pricing and supply?