While in Seoul for ICML, I spoke with dozens of people about Korea's unique position in AI. Together, Samsung and SK Hynix control 80% of the global high-bandwidth memory market (the "Strait of Hormuz of the AI ecosystem", as someone said). These two companies now dominate Korean stock indices. At the same time, the government has proposed a groundbreaking "citizen dividend" to redistribute revenues from the AI boom. I spoke with economists, legislators, academics, and Samsung workers. Some of my major takeaways: 🧵
Original post unavailable.
Sentiment
Users praise insights on Samsung and SK Hynix controlling 80% of the AI memory market because the data raises useful awareness for AI hubs.
Pos
100.0%
Neg
0.0%
1 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
VIEWS1.6KBOOKMARKS5LIKES18
REPLIES1
2/ Even if Korea does not develop models at the very frontier, it may benefit from an export strategy that leverages their relative geopolitical neutrality. They may look to their defense industry as a guide, which has grown 60% in 2025, with buyers from Europe to the Middle East.
1/ There is a lot of interest in developing sovereign AI models. Why? - concerns about sending important proprietary data (e.g. SK Hynix workers are apparently not allowed to use AI models for this reason) - concerns about vulnerability from foreign dependence an important technology (especially since the US is no longer seen as a reliable partner) - some optimism from the country's history of successfully developing domestic capacities for other technologies (e.g. automotive) - probably also reflects patterns across the world towards the on-shoring and self-sufficiency
6/ Questions about societal impacts often reflected societal issues that are top of mind. E.g. could automating work improve the birth rate and redu the suicide rate, by reducing costs and stresses of education and work, which are both very competitive? Could it increase gender discrimination if only physical labor remains? Work as a source of dignity or purpose, as is often raised in the US, was not much mentioned
5/ A lot of interest in redistribution, though this is not universal. Incidentally, SK Hynix workers have successfully lobbied for redistribution of the company's recent windfalls (e.g. bonuses at ~10x normal compensation). But the general population does not see their actions positively, nor as a first step towards wider edistribution -- it's seen mainly as self-interested, especially given that it was only negotiated for the memory workers (not for SK Hynix workers in general).
1/ There is a lot of interest in developing sovereign AI models. Why? - concerns about sending important proprietary data (e.g. SK Hynix workers are restricted in AI usage for this reason) - concerns about vulnerability from dependence for critical technology (especially since the US is no longer seen as much as a reliable partner) - some optimism from the country's history of successfully developing domestic capacity for other technologies (e.g. automotive) - it also probably reflects patterns across the world towards the on-shoring and self-sufficiency
While in Seoul for ICML, I spoke with dozens of people about Korea's unique position in AI. Together, Samsung and SK Hynix control 80% of the global high-bandwidth memory market (the "Strait of Hormuz of the AI ecosystem", as someone said). These two companies now dominate Korean stock indices. At the same time, the government has proposed a groundbreaking "citizen dividend" to redistribute revenues from the AI boom. I spoke with economists, legislators, academics, and Samsung workers. Some of my major takeaways: 🧵
2/ Even if Korea does not develop models at the very frontier, it may benefit from an export strategy that leverages their relative geopolitical neutrality. They may look to their defense industry as a guide, which has grown 60%in 2025, with buyers from Europe to the Middle East
1/ There is a lot of interest in developing sovereign AI models. Why? - concerns about sending important proprietary data (e.g. SK Hynix workers are restricted in AI usage for this reason) - concerns about vulnerability from dependence for critical technology (especially since the US is no longer seen as much as a reliable partner) - some optimism from the country's history of successfully developing domestic capacity for other technologies (e.g. automotive) - it also probably reflects patterns across the world towards the on-shoring and self-sufficiency
This came from just a week of conversations, and is of course limited. However, I think it's useful for us in AI hubs to be more aware of how the rest of the world is thinking -- please do share any thoughts or additional comments. I did not mention folks by name to maintain anonymity, but thank you so much to everyone for the interesting and thought-provoking conversations (and the new friends along the way)!
6/ Questions about societal impacts often reflected societal issues that are top of mind. E.g. could automating work improve the birth rate and redu the suicide rate, by reducing costs and stresses of education and work, which are both very competitive? Could it increase gender discrimination if only physical labor remains? Work as a source of dignity or purpose, as is often raised in the US, was not much mentioned
5/ A lot of interest in redistribution, though this is not universal. Incidentally, SK Hynix workers have successfully lobbied for redistribution of the company's recent windfalls (e.g. bonuses at ~10x normal compensation). But the general population does not see their actions positively, nor as a first step towards wider edistribution -- it's seen mainly as self-interested, especially given that it was only negotiated for the memory workers (not for SK Hynix workers in general).
3/ There is anxiety that the current dominance within the chip supply will not last, as the US and China could catch up in manufacturing.
3/ There is anxiety that the current dominance within the chip supply chain will not last, as the US and China could catch up in manufacturing.
4/ Interest in forming middle power coalitions. Unclear yet who would be the most natural partners. Some potential interest in Europe and Japan, though of course Japanese collaboration is sensitive
4/ Interest in forming middle power coalitions. Unclear yet who would be the most natural partners. Some potential interest in Europe and Japan, though of course Japanese collaboration is sensitive