Strategy, strategy, strategy.
So guys, after the SK Hynix Q1 call, I revised the strategic ranking of the three HBM suppliers and rebuilt the scoring framework as we got new information.
A few things to say about strategy first, because I think it’s the most underrated variable in how people read this market and how they navigate it, especially with this everyday news flow in HBM, which does not give you a single clear picture of what actually is going on.
We integrate this in our full HBM research, and this is just a part of it.
So why strategy matters more than most people think.
Most semiconductor investors focus on capacity, technology, and pricing, which are all important, but it does not mean anything without the strategic framework the company executes against. What tells you who wins in the long term is strategic position.
A strategic position is a compound of outputs related to choices that a company makes under pressure over years that also produce the culture and capabilities the company actually has today.
So we frame it through customer preference versus strategic positioning.
When you look at the supplier, there are really two questions that are related to this. The first one is whether the supplier delivers what customers are asking for today, and the second is whether the supplier holds a structural position that has enough strength through multiple generations.
A supplier can win the first question and lose the second. That is crucial to understand when you analyze a cycle like this.
And customer preference does not tell you whether the supplier still has a defensible position in, let’s say, 2028 or 2029. In these timelines, you get HBM4 mass production and HBM5, which has a custom base die, which makes full-stack integration a deciding variable.
So we build two separate matrices, one for each question.
In Matrix C3, you see the combined aggregate and why we weighted 40–60 in advantage of strategic positioning. The 40–60 split is a judgment call, and I want to be transparent about it. You could argue 50–50 or even 70–30 depending on your time horizon.
Our reasoning is that the 2026 to 2030 window that we are analyzing is defined by structural scarcity and contracted multi-year architecture. This is backed by the HBM engine I built to track certain constraints.
I have to put a side note on this, though, because cumulative evidence currently weighs a 64% scenario that it will lean into structural scarcity and contracted multi-year architecture, but it is not concluded yet, as we are waiting for multiple T01 signals to acknowledge it.
So we are still monitoring the situation upfront.
I wrote a full article on SS for free with an explanation of how we come to certain weightings and why certain weighted scenarios matter more than the other ones in our eyes. The link is in the comments, and you can fully read it for free.
This is just a snippet of our total HBM research that we are building, so more soon.
$005930.KS $000660.KS
$MU $SNDK $NVDA $AMD