At the blunt question, he looked momentarily flustered.
But it wasn’t something that could be hidden just because one tried to hide it. Wasn’t it something everyone in the world already knew?
“Yes, to be honest... while our competitors are building their databases with H100s, our main computing chips are still A100s, so I’m worried about the future.”
“So isn’t the government supposed to supply H100 chips in bulk through this Sovereign Project?”
“...Technology has advanced further in the meantime. The big tech companies are currently using H200s, and the difference in specs between the two computing chips is nearly double.”
Scholars had said it.
That the development of civilization after the Industrial Revolution was far faster than the development from the Paleolithic Age to the Industrial Revolution... and that in the Information Age, it would grow even faster.
And the world was proceeding exactly as they had predicted.
Roughly five years ago, when Envidi announced the A100, it had seemed so astonishing that there could never be a better computing chip. But in just two years, the H100, with roughly three times better performance, was announced. And then, after just one more year, the H200, with specs twice as good as that, ended up being released.
If you compared the processing speeds of these models, the difference in specifications was as vast as a cultivator, a car, and an airplane.
“...”
I clenched my fist tightly.
Back in the early days of the Sovereign Project, I had met Envidi’s Jackson Hwang and secured a promise for 100,000 H100 units. At the time, the H100 was a chip every country in the world was lining up to get, so we had no choice but to pay a markup twenty times over.
But in the meantime, Envidi had announced an even higher-spec chip and was supplying all of that inventory to big tech.
For reference, other countries were still desperate because they couldn’t even get H100s. A year had already passed since the latest model was announced, and they were still scrambling because they couldn’t even get the older model.
That was why Envidi’s stock price had risen tenfold in five years.
“Chairman. Still, the government has done what it can. It’s not as if there are no H200s at all in the country, and I understand that Seoul National University’s research team has 10,000 units. If we combine that with the 10,000 H100 units the government is supporting us with, shouldn’t we at least be able to put up a fight?”
Kim Seokhun spoke with a dazed expression.
“Yes... Korea has roughly 20,000 units combined of H100s and H200s. But MS alone uses 500,000.”
“...Pardon?”
“Facebook uses 350,000, and Teslan uses a total of 200,000.”
“...”
“And Gogle, which has fewer than them and is our main competitor, is still estimated to possess at least between 100,000 and 200,000 units.”
As I listened to the explanation, I ended up just as dazed as Chairman Kim Seokhun looked.
There was no way American engineers were less intelligent than Korean engineers. The most renowned geniuses from all over the world were arming themselves with high-performance, high-compute chips and continuing to widen the technological gap.
Was there truly no hope for our country’s platform industry? Just as WebTube had completely monopolized the global streaming market, did we have to lose our search engine market to Gogle as well?
“...”
Absolutely not.
As we had seen with in-app payment commissions, Gogle was a company that was only kind at the beginning. Like a giving tree, they provided everything for free without holding back, but once they finished taking over the market, they immediately turned into thugs.
And if my memory was correct, Gogle’s tyranny over fees began once they had finished seizing the Korean market.
But we couldn’t respond the way we had with in-app payments. Because by then, it had already solidified into a monopoly...
“Chairman. Then please tell me what Neive needs most. Personally, I don’t want the Sovereign Project to end as a show. I truly want Korea to be reborn as a country with AI sovereignty. What can I help with?”
After gauging my mood for a while, he said to me,
“In truth, we’ve also obtained a great deal of internal data while launching our Clova service. We finished identifying the problems and the methods to improve them two years ago.”
“Is it the chips?”
“Yes... Since we knew it would be difficult to secure inventory, we tried every method we could think of to improve things without chips. But the more we looked into possible methods, the more clearly we realized the importance of chips.”
“How many do you need?”
“At the very least, tens of thousands. Our current main computing chip, the A100, now runs at the speed of a cultivator, so we need to scrap all of them and replace them with H100s. Furthermore, we need to possess at least 30,000 H200s, the kind big tech companies are using by the hundreds of thousands, if we are to compete with Gogle.”
I asked,
“Anything else? Are there no issues with recruiting talent?”
“No, there are no major problems there. We at Neive have done our best to match the going rates for master’s and doctoral talents so we don’t lose them. The problem is that we can’t give those talents notebooks and pencils...”
I nodded.
The one fortunate thing was that Korea, too, was a country that ground down its talents no less fiercely. Of course, they said the recent AI market was meaningless outside of the G2, but even so, Korea’s AI competitiveness ranked third. The fact that we were maintaining that ranking with A100s was proof that we were grinding our researchers down fairly well, wasn’t it?
“...But would that truly be possible? The U.S. authorities control H100 inventory as well, but they seem to treat H200s almost like military supplies.”
“Isn’t that because of technology leakage to China?”
“To be honest, those of us in the industry see that as an excuse. Just like when the U.S. used fentanyl as the excuse for every tariff they imposed on China. So our internal forecast is that if we secure a meaningful quantity of H200s, the U.S. authorities will definitely take issue with it.”
Inwardly, I was impressed by Chairman Kim Seokhun’s insight.
His prediction was flawless. The U.S. authorities, both openly and behind the scenes, tightly controlled H200 chips, as well as the next-generation computing chips that would come out in the future.
And this was an issue that had no exceptions, not only for China but even for allied countries. Whenever they released the latest chips to their allies, it was always after the next-generation chip had already been developed.
It seemed Neive’s strategy office had already seen through that.
“Then in the end, the task is to break through America’s stringent guard and secure the inventory... I’ll give it a try.”
“Pardon? B-but, Team Leader.”
“Chairman, it will probably be possible. The government is quite motivated about this Sovereign Project. Replacing all of our current main computing chips, the A100s, with H100s should be possible within two years. Since our pension fund received an inventory promise from Envidi even while paying a twentyfold markup, we’ll definitely receive priority supply for that.”
I stood up and said,
“And as for the H200s you mentioned... and further, the next-generation computing chips that will keep improving... we’ll somehow secure the inventory through a public-private partnership. But there is one condition.”
“P-please tell me. If we can secure the chips, anything...”
I handed him a document.
“Please invest a significant portion of the profits into autonomous driving according to this paper.”
“A-autonomous driving? But we’re a platform company...”
“That is precisely why I’m saying this. If Neive truly considers Gogle a competitor, then please challenge businesses like Waymo. Now that the AI era has arrived, distinctions between industries no longer hold much meaning. A company with AI technology can now become a portal company, an autonomous driving company, or an automation and process-optimization robotics company.”
“...”
“Neive is now, in the strictest sense, Korea’s national representative AI company. Petty conglomerate expansion, fighting over neighborhood markets and rice bowls, no longer suits you. Reduce corporate R&D focused on service platforms and use all of it for AI technology development. If you do, then I will believe in your sincerity as well, Chairman.”
Chairman Kim Seokhun turned pale.
In truth, Neive’s main revenue came from service platform commissions, and the reason its stock price was still holding out was the expectation that Neive would at least not lose the domestic market. But now I was telling him to reduce the R&D that would be invested there and invest it into AI. Wasn’t this far too much of a gamble with the company’s fate on the line?
“T-that is... first, we’ll need to hold a meeting...”
I said to the hesitant man,
“Chairman. Do you not need high-compute chips?”
*
“Why do you always go against the grain?”
The next day.
After reading my buy report, Manager Oh snapped at me bluntly.
“...This seems like the right direction.”
“From what angle? When the KOSPI doubled, Neive only went up 50%. Even after rising that much, its stock price is still half of what it was three years ago. So why are you buying Neive shares?”
“But in this Sovereign Project...”
“Sovereign is just an annual event the political circles hold as a matter of routine. So what if they come in first? Can they secure the high-compute chips that MS uses 500,000 of, Facebook 300,000, and Teslan 200,000? They’ll give meaningless support, say they did everything they could, and that’s the end. Am I wrong?”
I smiled bitterly.
It certainly seemed there were many investors who thought like Manager Oh.
After a fierce competition, Neive had been selected as our representative AI player, but as if to render such good news meaningless, Neive’s stock price was continuing to fall day after day.
“Team Leader Lee, why don’t you sit this one out? The AI bubble theory hitting the Nasdaq these days is already looking serious. To be honest, it’s gone up too much. Five years ago, if you didn’t buy real estate, you became suddenly poor; these days, if you don’t buy stocks, you become suddenly poor, don’t you?”
“...”
“I think the stock market, like real estate, is about to face one big cold wave. Given the current situation, even I can’t accept an investment in a platform company.”
I raised my voice.
“Manager, aren’t stocks originally supposed to rise in an N-shaped pattern? This is a healthy correction. The fact that AI will become the center of high-tech industry hasn’t changed in the slightest.”
“I know. It hasn’t changed. But if you know that, why are you telling me to buy Neive shares?”
“Pardon?”
“The fact that AI will become the center of high-tech industry means Envidi’s importance will grow. It means companies receiving GPU supplies from Envidi will become richer. Then it’s obvious Neive’s search share will be pushed out by Gogle, so why Neive?”
Damn it. I had nothing to say.
“...I’m saying we’ll try to change that.”
“How, exactly? Are you going to kidnap Jackson Hwang and bring him to Korea?”
“It can be done. The government didn’t launch Sovereign just to put on a show. They must have planned out the inventory, and they’re probably already aiming to secure next-generation models as well.”
“That’s just your wishful thinking...”
“No, it isn’t. I heard it through a tip from someone well-connected in political circles. The government has secured supply not only for H100s, but also for 200s and next-generation models.”
Manager Oh’s eyes widened.
“Where did you hear that from now? Is it true?”
It was a lie. If my memory was correct, the government had never signed such an MOA.
“Yes. It’s true. It’ll be announced in a week.”
“Who the hell are you hanging around with to hear news like that...”
“Manager, that’s not what matters right now. Neive has ample growth potential, and the government will make efforts that match it. Right now, instead of worrying about a bubble theory, we should be investing aggressively in domestic AI.”
Manager Oh sank into thought for a while before finally handing over the documents.
“You’re sure, right? There really is big news coming from the government in a week, right?”