Pangyo, Bundang, Seongnam—Nave headquarters.
The executives gathered in one place were cautiously watching one man’s mood.
Last week, Nave had proudly taken first place in the Sovereign Project and become the nation’s representative AI contender, but now was hardly the time for celebration.
“If everyone’s here, let’s begin the meeting. Executive Director Park?”
At Chairman Kim Seokhun’s instruction, Executive Director Park, the CTO, sprang to his feet.
“Yes. The results for the last quarter have been fully compiled. To begin with revenue...”
As Executive Director Park’s report began, the already cold conference room grew closer to freezing.
This quarter’s results were a disaster as well. Nave, which had once controlled 60% of the domestic search engine market, was steadily retreating as it lost market share to AI-based Gogle.
“Lastly, operating profit is...”
“Enough.”
Chairman Kim stopped the presentation and trembled with his fists clenched.
“Today is quite the historic day, isn’t it? At last, our Nave has achieved a 49% share of the domestic search engine market. From 60%, that is.”
The executives shook like aspens.
Today was indeed a historic day for Nave. The domestic search engine market share they had barely managed to maintain had finally fallen below the majority.
Meanwhile, Gogle’s market share, which had been 5%, had soared to 30% before anyone knew it, and among young users—a barometer of future indicators—its usage rate had surpassed 70%.
“I think I need to hear the reason. Executive Director Park?”
“Yes, Chairman...”
“It’s not as if the company hasn’t spent money, and it’s not as if every engineer in our country emigrated. So why does our market share look like this?”
Executive Director Park reported with the face of a man who had committed a grave sin.
“...The primary cause is a lack of technical capability. Recently, Gogle has been rapidly increasing its market share by providing AI-based search services, and it is true that our technology falls short of theirs.”
“So did we not spend money on R&D? We created Clova to stop Gogle’s Gemini, didn’t we? Then why do we keep getting pushed back?”
Kim Seokhun flew into a rage.
It wasn’t as if they had sat on their hands. They had tried everything they could, and this was still the result.
As soon as Gogle launched Gemini, Nave introduced Clova and fought a desperate battle for market share. Thanks to their proactive response, they had initially been able to defend the domestic market with a higher share than Gemini.
But accumulated technical capability and American capital gradually built a wall that could not be overcome.
At the beginning, the two AIs had been similar, but as time passed, the gap widened in accuracy, hallucinations, and speed. That gap was now appearing directly in search market share.
“With all due respect, Chairman, our Clova has far too many structural limitations...”
“What?”
“...To reduce AI hallucinations and improve search accuracy, we ultimately need an enormous database and high-compute chips capable of processing it. But we only have about two thousand H100 chips. With things as they are, how can we possibly beat Gogle, which possesses tens of thousands of H100s?”
Executive Director Park continued as if he had steeled himself.
“The fact that we’ve managed to hold out even this much is the result of grinding down the few engineers we have. This isn’t an excuse. We truly need H100s—and a great many of them. Right now, our engineers are in a situation where they can’t study because they don’t even have pencils.”
Executive Director Park even shed a few tears as he appealed to their difficulties.
At present, Korea ranked third in AI technological indicators and was not a country that lagged far behind in the global rankings. Considering that they had only around two thousand H100 chips, it was almost astonishing that they had managed to rank third globally.
But they were gradually feeling the limits of their stamina.
Big Tech companies, backed by the U.S. authorities, were receiving hundreds of thousands of high-compute chips and correcting AI’s problems one by one. On top of that, they were building data centers all over the world, effectively outsourcing their power supply.
Now, three years later, Big Tech’s AI hallucination symptoms had been markedly reduced, while search accuracy, computation speed, and other factors had advanced dazzlingly.
“To be honest, A100s are no longer of any use. It’s like a motor revolution has taken place in automobiles, but we’re still pulling horse carriages.”
The only reason they had managed to hold on until now was because the AI business had still been in its early stages.
After all, when automobiles were first released, their speed had been similar to horse carriages. But in the end, as time passed, they developed at a speed that horse carriages could never follow, and horse carriages eventually vanished from the market.
The AI market was currently following the same trend.
“So what’s your conclusion?”
“High-compute chips. What we need now are high-spec high-compute chips, not mass-market supplies. We can no longer survive with A100s.”
“Stop asking for things outside my authority, and tell me something I can actually solve. Even the government is trying to secure H100s, but supply isn’t coming through, is it?”
“...We still need them.”
“What do you expect me to do about something that can’t be done?”
“...We have to make the impossible possible.”
As the frustrating conversation continued, Kim Seokhun’s anger rose to the top of his head. But he did not do something foolish like shouting.
If the company CTO was asking him to make the impossible possible, it meant he had already tried everything. In the end, his conclusion was that they could do nothing without high-compute chips.
“What do the other executives think? Is there truly no way besides high-compute chips?”
“...”
“No, China created DeepSync with A100s. Are we really unable to do anything?”
He asked a second and third time, but the answer he received was the same.
As silence settled over the conference room, the vice chairman cautiously raised his hand.
“Chairman, may I say something?”
“Go ahead, Vice Chairman.”
“To be honest, H100s are not cutting-edge models either... MS, Teslan, and GPT are already using H200s. H100s will probably become old models soon as well. Just like A100s.”
The atmosphere in the conference room sank once more.
This was also why Nave, despite taking first place in the recent Sovereign Project, could not smile. The government had promised to provide a considerable number of H100 chips to the final selectee of the Sovereign Project, but Envidi had already created another next-generation model with the H200.
In other words, even if Nave struggled to secure H100s, their competitors would be armed with even better models.
“So Executive Director Park’s demand for H100s is the absolute bare minimum condition required for technological development, not the whole of it. If we can’t even secure that, we might as well shut down the business.”
Kim Seokhun looked at the executives.
“Haa...”
As Chairman Kim sighed, his secretary knocked on the conference room door.
—Chairman. I apologize for interrupting the meeting, but you have a visitor.
“This is an important meeting. If it’s someone without an appointment, tell them I can’t meet.”
—But they seem to be from the government...?
“Who? A government official?”
—Yes. They seem to be from the pension fund, and it looked like they came to discuss the Sovereign Project. What should I do?
Chairman Kim adjusted his tie with a weary expression.
*
Pangyo, Korea’s Silicon Valley.
As befitted its nickname as the IT capital, the atmosphere in the offices here was free-spirited.
There was no sign of the suits that filled Yeouido, and the company welfare facilities were practically reminiscent of Tower Palace.
—James, please have this material ready by today.
—Understood, Catherine.
—Daniel, how about a glass of somaek after work today?
—Sounds good. It’s on me today, Matthew.
With such lavish welfare facilities and salaries at the highest level, the employees’ faces were bright as well.
Nave had recently acquired and merged with Kakao, which had gone through bankruptcy proceedings, and had taken in much of its good corporate culture. In order to improve its vertical workplace culture, the company had employees use English names internally. Perhaps because of that, the corporate culture looked far more relaxed than that of ordinary companies with strict hierarchies.
“...”
But amid this happy atmosphere, I saw hell.
Ten years... if my memory was correct, ten years. Nave, the absolute ruler of the domestic search market, would be pushed back in market share by AI-based Gogle and eventually lose its title as the country’s largest portal site.
With operating losses piling up, it would ultimately carry out the ultra-intense restructuring known as voluntary retirement... and after that, as I recalled, Korea’s search engine market became Gogle’s one-man stage.
“...”
It was truly ironic, in fact.
Wasn’t the first model of AI Nave’s “Knowledge iN”? A platform service that utilized collective intelligence had many similarities with the AI currently being commercialized, so it was deeply regrettable that we had missed such an opportunity.
If, twenty years ago, someone had properly recognized the potential of collective intelligence and big data, and discovered that combining them would become “deep learning,” where computers could learn on their own... by now, even if Korea couldn’t surpass the United States, it would certainly have been able to get ahead of China.
“...”
What on earth had our STEM students been doing? This simple principle was something even a humanities major could think of...
“Good day, Team Leader. I’m Chairman Kim Seokhun, CEO of Nave.”
While I was lost in idle thoughts, Nave CEO Kim Seokhun arrived.
I smiled warmly and shook his hand.
“Yes, it’s a pleasure to meet you. I’m Team Leader Lee Sejun from the Pension Fund Risk Team. I visited without notice, so thank you for making time for me.”
“Not at all. Rather, I’m sure you came because you took time out of your own busy schedule, Team Leader. But may I ask what brings you here...?”
“I wanted to offer my congratulations. I believed in you—that Nave would make it. Sincerely, congratulations on becoming the final selectee of the Sovereign Project.”
“Ah, yes. Thank you. Haha.”
Unlike his smiling face, his eyes were already crying.
I did not need to ask why.
Recently, the KOSPI had entered a roaring bull market, an era where anything you bought went up, but Nave was the exception.
Of course, it was not as if it had not risen at all. It had climbed nearly 50% from its low point, showing the strength of the top domestic platform. The problem was that even after that rise, its stock price was still half of what it had been four years ago...
It was so bad that even after becoming the final winner of the Sovereign Project, which could be called a massive boon, its stock price remained sluggish. Recently, Kakao had gone bankrupt, and Nave had acquired and merged with it to become the only domestic search engine site... but people still seemed to harbor doubts about Nave.
“How do you feel? The media is already expressing expectations.”
“Haha... Yes. But truthfully, right now I feel more in need of consolation than congratulations. Now that we’ve become the nation’s representative AI contender, I understand that our responsibility is tremendous. We will be sure to repay those expectations.”
Looking at his desolate face, I asked,
“Chairman. Is the reason you want that consolation, by any chance, because of high-compute chips?”