American AI Now Worries About Power

A friend I studied with at the University of Michigan came to Korea for the first time in 15 years. They're Korean American, born in the US and raised there their whole life, so they were never an international student like me. They worked at a financial firm in New York and are now a PM at a publicly traded cloud company. They just got married and picked Korea for the honeymoon. I was grateful they made time for me at all in the middle of a honeymoon, and before I had to leave for a Bloom event that evening, we ended up talking far more deeply than I expected.

Seoul, seen by someone who grew up in America and hadn't been back in years, looked different from the Seoul I see every day. And the American AI industry they work in is already worried about a problem one step ahead of what we talk about here. I'm writing down what I heard before I forget it.

Seoul moved fast and left people behind

The first thing my friend said about Seoul was that it had changed so much. The second was that the income gap was visible and it made them a little uncomfortable. Seongsu-dong (a trendy Seoul neighborhood full of cafes and pop-up stores) was packed with young people, but almost everyone who served them while they were out and about was older.

They want to move fast and they left behind the older generation.

I've had the same thought. With so many people getting rich off cutting-edge tech and pop-up stores opening on every block, I've wondered what the older folks walking past think when they see all this, and how they make a living. My friend said that 15 years ago Seoul still had a neighborhood feel, a kind of "village life," and now even the back alleys have turned into those kinds of shops. They described it as "citizens are getting gentrified." The people who lived there are being pushed out.

The shape of the city felt strange to them too. In New York there's one consistent texture wherever you go. In Seoul, one neighborhood feels like this and another feels like that, and one alley over you suddenly hit a street lined with lamb skewer restaurants. They said it seemed like there was no urban planning, but that they didn't dislike it. "I like it, but it leaves me unsettled." That was exactly right.

We also talked about how all the jobs are in Seoul, so everyone crowds into Seoul, while the rest of the country loses people and works hard to attract foreigners. What's funny is that a lot more foreigners live in Daejeon or Busan than you'd think. Ask them why, and they say they went because it was the second biggest city. They figure Busan will be to Seoul what LA is to New York, and then they're surprised by how big the gap is. For us, there's only one metropolis: Seoul.

Three parts coffee, one part milk

I asked my friend when they feel Korean. They called themselves "three quarter American, one quarter Korean." Something like three parts coffee mixed with one part milk. They said it isn't about food. Their American wife actually likes Korean food more than they do.

It's more in the way they think. There was a recent story in the Korean news about an older farmer in the countryside who hands fresh vegetables to the driver of the bus they ride every day, and my friend said scenes like that have built up inside them too. When they first met, their wife asked, "Why do you give so much detail?" meaning why they looked after every little thing. As we talked, they decided that's where it came from. I laughed when they told me their relatives teased them for calling elementary school "gukmin hakgyo", the old name Koreans stopped using decades ago.

What my friend carries is the old Korean sensibility they inherited from their parents' generation. So I asked what Korean identity is today, and neither of us could answer right away. My friend said they understand how older Koreans think but don't really get how the younger generation thinks, and I could only nod.

People who pull each other up

In the American tech industry where my friend works, the network of Indian engineers is famous. There are a lot of them, enough that 71% of approved H-1B work visas in the US go to people from India. But what my friend was talking about was less the numbers and more the way they pull each other up. People who grew up in the same towns and went to the same schools settle in all corners of big companies, and when it comes to hiring or making big decisions, they back each other. It makes sense that people who left home and had to make it in an unfamiliar place would do whatever they could to look after family and friends.

I looked it up, and this culture even has a named organization. TiE (The Indus Entrepreneurs) was founded in Silicon Valley in 1992 by Indian founders. It's a nonprofit that started so that experienced founders could mentor and connect younger people who wanted to start companies but had neither the know-how nor the connections.

My friend has also spent years around Jewish neighbors and coworkers, and said that community is similar. They may disagree and argue among themselves, but they see the bigger problems as always being outside, and toward the outside they move together.

As for Koreans, my friend laughed and said that growing up in America, if there was one other Korean kid in class, they ignored everyone else and thought, I have to beat that kid first. I wasn't much different. After competing for so long in such a small country, memories like messing with other people's studying before an exam are still in our bodies, and even in America we see other Koreans as competitors first. Things have loosened up a lot, but that habit doesn't go away easily.

Korean people watch out for Korean people.

Koreans should look out for Koreans. That line of my friend's connects to an idea I keep holding on to while running Bloom.

We look up to America too much

We talked about how Koreans see America. What I feel lately is that we see the American market and American people as much bigger than they really are. We think of going to America like going to heaven, so even once we get there, we don't blend in naturally. I told my friend that someone like them, working in America, could come to Seoul and become a speaker at a COEX conference (Seoul's big convention center) just for working in America. My friend said, "That's crazy."

Put kindly, it's a desire to learn. Put less kindly, it's a way of keeping your distance. Once you draw the line first, you're American and I'm Korean, it's hard to get close.

It's all about your branding and positioning. It's just all smoke and mirrors.

Whether you go to a big company or a small one, people are pretty much the same everywhere, and in the end it's a question of how you come across. My friend said that these days Americans actually want to learn from Korean design and taste. "We have respect for you guys too." Their diagnosis was that we think we can't build something like ChatGPT or Claude because we sell ourselves short.

That's why I think we need people who blaze the trail first. When Park Jin-young (the singer and founder of JYP Entertainment) tried to break into the US market long ago, he took a lot of heat, but because a few people like him pushed through first, we have K-pop as it is today. My friend named Blind (the anonymous workplace app) as the most successful Korean startup in America. Everyone working in American tech uses it, they said. If you can't make news in Korea, build it in America and take it global from there. That's what the internet is for, my friend said.

The freedom to experiment on the government's dime

What surprised me was that my friend envied the Korean government's support for startups. These days Korea has a huge number of government events that gather people to teach them AI, plus startup competitions. I said it's a shame that so many of them just hold the event and stop there. But my friend said Korea actually needs that. To do something new, you need room to experiment freely, and in America the cost of living is so high that not just anyone can try starting a company. For someone barely getting through the month on a paycheck, a startup is somebody else's story. In the end, only people with money or exceptionally talented people take the shot.

The Korean setup has its gaps, of course. From what I hear on the VC side, a big share of the investors putting money into funds are government bodies, so going around to local governments to raise money is part of the daily routine. My friend said that even for programs that bring in foreign VCs or accelerators, pairing them with Korean VCs would keep the know-how in Korea. They also said it would be great if big companies stopped just sitting on their technology and played a role like YC, investing in startups.

Apps for bringing people together

These days I'm growing Mixer, a curated gathering that connects people, and my friend said they'd looked into a lot of similar services. We were both watching 222. It's an app where you fill out an application and it puts together dinners for people who are a good match, and it's backed by Y Combinator. My friend's phone was full of apps like this.

They said that in America people use Partiful more than Luma. It's an invitation app backed by Andreessen Horowitz, but according to my friend it's for hanging out with friends you already know and doesn't work for meeting new people, and once you've gotten the invite there's no reason to open the app again. Apple also launched an invitation app called Invites early last year. People clearly want this kind of thing. The problem is money, my friend said. If you set out to make money from the start, people push back, so it's hard.

When I talked about Bloom, my friend told me to look at branding before money.

You're associated with really hot things. And like you're a big name there.

Being mentioned alongside things that are taking off is itself a kind of power, and those pieces add up and fall into place all at once when the opportunity comes. They asked how I'm so good at connecting people, and I said I just ask. I ask whether someone knows a person, and if they don't, I ask someone else. The world is smaller than you think. My friend said, "That's the first step, but no one does that."

The money flows below the model

When we moved on to AI, my friend said to think of AI like the computer. It'll end up in everything, but it won't look like a chat window such as ChatGPT. People would rather have AI come into the work they already do than learn a new tool. That's why American VCs are looking at service businesses these days, they said. First you help businesses add AI to their work, you get data in the process, and you learn what's really hard and turn that into your moat. There's also a lot of talk about a small private equity style approach: buying small shops or service companies outright, putting AI into them, and taking the data and the way they work. This is also why I'm trying to widen my AI conversations beyond tech, toward shops and service businesses.

Still, my friend said that when people in America talk about AI, what they're really watching is the infrastructure underneath the models. They showed me a ranking of GPU cloud companies. It's called ClusterMAX, put out by the semiconductor research firm SemiAnalysis, which actually uses the GPU clouds and grades them. In the latest edition, from September 2026, the top tier, Platinum, is CoreWeave and Nebius. Nebius split off from Yandex, once called "the Google of Russia," and it's also a company Bloom co-hosted a hackathon with in Seoul.

My friend said AWS got big because it made Netflix succeed, and being part of Amazon wasn't the reason. The bigger Netflix got, the more cloud it needed, and the two grew together. Cloud and GPU companies are pouring money into OpenAI and Anthropic these days for the same reason: those companies have to succeed for them to succeed.

It's just going round and round until more people can bundle in. Creating the bigger flywheel.

Now it's electricity

What my friend stressed most was that the bottleneck keeps moving. First there weren't enough GPUs. Then it was memory. Memory is already something everyone knows about, since SK hynix shot up. Next is electricity, and after that come the networks and cables that link data centers.

Electricity isn't solved by just building power plants, they said. Getting the power connected to data centers and through regulation takes 18 months at the short end and 10 years at the long end. Companies that need AI data centers right now can't wait 10 years, so the companies that solve this problem are taking off. The example my friend gave happened to share our name: Bloom Energy. It's a fuel cell company that generates power right next to the data center. Last year it signed a $5 billion AI infrastructure partnership with Brookfield, and it supplies fuel cells to Oracle's data centers.

As I listened, Korea and the US looked like they were standing in different places. In Korea we're still talking about how many GPUs we can secure and what to do about physical AI, while America is already worried about the next thing, energy. American big tech is already going nuclear. Microsoft agreed to restart the Three Mile Island plant and buy its power for 20 years, Google and Amazon teamed up with small modular reactor companies, and Meta also signed a nuclear power deal. The AI race is turning into a fight over who can lock up more electricity, faster.

Where Korea fits in

If America is stuck on electricity, making and transmitting power is something Korea is good at. Doosan Enerbility and Hyundai Engineering & Construction are already in on a giant AI energy campus being built in Texas, handling the main equipment and design for four nuclear reactors. Thanks to demand from the US power grid and AI data centers, Korean transformer makers like Hyosung Heavy Industries, HD Hyundai Electric, and LS Electric have years of orders stacked up. Bloom Energy, the company my friend mentioned, has also worked with SK ecoplant for a long time, building fuel cells together in Gumi, in North Gyeongsang Province.

Something my friend added while talking about Blind came back to me here. Think of America as a strategic place to invest, and think about who you can help there. The way AWS grew alongside Netflix by helping it, Korea can grow with American AI by helping solve its electricity problem. We've already done it once with memory.

I think AI is an opportunity for countries like Korea to compete with bigger countries.

My friend said Korea is good at math and computer science but still doesn't get proper recognition in the world. Everyone is tied up worrying about getting by day to day, they said, but this time we have to step out of that and look further ahead. With that, we each headed off to our evening plans. I went straight to the Bloom event.

Originally written in Korean on Brunch, October 4, 2026.