A Wall Street Genius's Final Investment Playbook-Chapter 273 : The Invisible Hand (8)

Meanwhile, the Envid board of directors had spent a troubling three weeks.

Ever since Ha Si-heon had foretold the "intensification of the war."

The string of events unfolding in reality had all been unsettling.

<Stark’s “LLM”… Is the Gooble-centered ecosystem shaking?>

When Stark first unveiled its LLM, most of the board members accepted it calmly.

‘Ah, Ha Si-heon must’ve known about this.’

That alone could be written off as high-level intelligence gathering.

<Gooble claims first human victory with AlphaGo… “AI can surpass humanity”>

Even when Gooble pushed AlphaGo to the forefront, it wasn’t all that shocking.

It made sense that Gooble would respond aggressively to Stark’s LLM.

But that was merely the usual market rivalry.

Just the typical push and pull—nothing that could be called a "war."

However—

Right after that, things began to change dramatically.

<Stark partners with AWSS… The war for cloud computing resources begins in earnest>

<Stark signs multi-trillion-won HBM deal with Hynixen… Memory market upheaval looming>

Stark began partnering with Gooble’s rivals and suddenly secured massive amounts of memory.

‘That’s... a little aggressive?’

Just as the board thought that, the situation took an even stranger turn.

<HBM stock depleted… Market price rises 37% in just two weeks>

<Upstream manufacturers adjust production schedules… Memory crisis becoming reality>

A memory market meltdown.

And then—

<Stark and Gooble snatch up all Parser GPUs… Market supply chain collapse>

<Small AI companies forced to halt training… “All cloud GPUs are gone”>

<Used GPUs trading at double retail price. Reseller market overheats>

Like fire spreading through a dry field, the crisis jumped into the GPU market.

When Parser architecture supply was monopolized by Stark and Gooble, the supply-demand balance collapsed instantly.

Prices surged while stock vanished, and the chaos only grew as hoarding and speculation followed.

‘What… is this?’

At first, it seemed like just another flare-up in competition...

But now, it was like someone had dumped gasoline on the flames—it had become a full-blown inferno.

And Envid stood right in the center of it.

Because both Stark and Gooble were pressuring the company to "accelerate the release of the next-gen GPU."

‘Why did things escalate like this so suddenly…?’

The board members were all stunned.

Someone clenched their teeth while recalling the peaceful days of the past.

‘When did the board become like this…!?’

Just a few months ago, the boardroom had been a completely different place.

A formality held once every quarter.

A simple business report from the CEO, followed by one or two light questions—and that was it.

Most board members had kept their seats thanks to their "ability to avoid decisions."

And in a culture that didn’t question that, they had enjoyed a comfortable peace.

But now?

The boardroom had transformed into a full-blown crisis control center.

They were being asked to make major decisions, and whatever choice they made would come with real consequences.

It was no wonder they were overwhelmed.

So whenever they thought of the root cause of this entire crisis, the board members felt a surge of stress—as if their headaches and hair loss were returning.

That root cause being—

Bang!

Just then, the door opened and in walked Ha Si-heon.

Unlike the gloomy board members, a cheerful smile lingered on his lips.

“You’ve convened earlier than expected. It hasn’t even been three full weeks.”

His voice was practically humming with joy.

As if he were genuinely enjoying the current crisis.

“I hear both Stark and Gooble are asking for Bolt-on. In that case, shouldn’t we just give it to them?”

Just a few meetings ago, Ha Si-heon had been the one insisting on an early release of Bolt-on.

So perhaps the current situation was quite welcome to him.

However—

Even considering that, the look on Ha Si-heon’s face was too cheerful.

In that moment, a strange suspicion rose in the minds of the board members.

‘No way… is all of this going exactly the way he planned…?’

Ha Si-heon was the one funding Next AI, so it was entirely possible that he had some influence over Stark’s moves.

If so, could he have orchestrated this whole situation?

But—

They quickly brushed that thought aside.

Who would throw the entire global market into chaos just to launch the next GPU a few months early?

That would be completely irrational…

…wouldn’t it?

At that moment, memories of the incidents Ha Si-heon had previously involved himself in flashed through their minds.

Had any of those incidents unfolded "rationally"?

A heavy silence fell over the room.

Then, the chairman quickly moved on to the agenda.

“Now that everyone’s here, let’s begin. We have only one item today. Stark and Gooble have both requested that we supply Bolt-on within Q3.”

Bolt-on was Envid’s next-generation GPU currently in development, boasting an architecture specialized for deep learning.

The development was nearly complete, and they were ready to start mass production.

But they had been delaying the release for one reason.

The current-generation Parser architecture was still performing well in the market.

If they released Bolt-on now, it would eat into Parser’s sales.

This is what they called “self-cannibalization.”

So Envid had been waiting.

Waiting for Parser to hit its peak revenue and start to decline naturally.

But now—

Stark and Gooble, fully aware of this, were still demanding an early launch.

“They’re calling it a ‘pre-contract request,’ but it’s basically a threat. They’re saying if we don’t supply by the end of the year, they’ll switch to other vendors.”

“Both companies are our largest clients. If they really leave, the damage will be significant.”

“But we can’t just abandon Parser like this.”

“Still, it’s better than losing our customers.”

Many comments were exchanged in a short time, but none of them carried real accountability.

Then—

“I believe we should stick to our original schedule.”

As always, CEO Jackson laid down the final word.

“Even if Stark and Gooble say they’ll find new suppliers, no one else can deliver a Bolt-on level product in that timeframe. Eventually, they’ll have no choice but to go along with our schedule.”

He was prioritizing maximum profit over customer pressure.

“But what if both companies do leave?”

That question wasn’t really a counterargument—it was more like a way to shift blame.

Just something to say later: “I did raise a concern.”

But Jackson stood firm.

“Envid holds a dominant position in the GPU market. No matter how desperate Stark and Gooble are, they’ll eventually come back.”

His voice was filled with confidence.

His tone was resolute, and within that firmness was the heavy burden of responsibility that no one else on the board wanted.

But then—

Ha Si-heon, who had remained silent the entire time, finally spoke up.

“I don’t think that’s a very wise decision.”

Someone had finally voiced genuine opposition to CEO Jackson.

“Under normal circumstances, they might wait. But this is war. They’re doing everything they can to get ahead, and they’re not going to let a bottleneck like Envid hold them back. The side with the money will look for a workaround. And if that happens, they could end up placing massive orders with Envid’s competitors.”

“Our competitors are far behind technically. They wouldn’t be able to handle that kind of demand.”

“With enough funding, they’ll catch up fast.”

“But that would take a massive investment. There’s no evidence that they’d go that far just for a few months’ head start—”

“There’s no evidence. Yet.”

Cutting Jackson off, Ha Si-heon grinned.

“Out of all the things I’ve ever said, how many came with evidence from the start? And yet, they always turned out right.”

Then he declared firmly:

“This war is only going to escalate.”

He was warning them that the already overheated situation was about to get worse.

“If you really need proof, I can show you in another week or two.”

And yet… something was off.

That excessive confidence of his…

It sounded like he was saying he could create that scenario himself if he had to.

Just then, the board members remembered something Ha Si-heon had said during the last meeting—"See you in three weeks."

And what had the last three weeks been like?

The war between Stark and Gooble—and its rapid escalation.

‘No way… could he really have…??’

Could everything that has happened so far have been orchestrated by Ha Si-heon?

As that suspicion flashed through the board members' minds, Ha Si-heon slightly curled the corners of his lips and spoke.

“My advice is simple. If you wait for evidence and respond too late, the situation could worsen. But if you listen to me and take action now, this crisis could actually turn into an opportunity.”

The meaning behind his words was clear to everyone.

‘If you follow my lead, you’ll survive. Ignore me, and you’re doomed.’

In that moment, a warning resurfaced in the board members’ minds.

It was something Pierce had once said.

— Ha Si-heon is like a natural disaster. The more you try to stop him, the more he will escalate things in ways beyond your imagination.

At the very least, that statement had proven undeniably true.

You’re better off giving him what he wants. Naturally, your instincts will reject the idea. But if you suppress that instinct and accept him… you might be greatly rewarded.

It was advice they hadn’t understood at the time…

But now, they could begin to grasp what it meant.

Just then, the chairman moved to conclude the meeting.

“Then let’s proceed with the vote. Those in favor of advancing Bolt-on’s production schedule within Q3, please raise your hand.”

Rustle rustle rustle.

Everyone raised their hand—except Jackson.

“With 11 votes in favor and 1 opposed, the motion is passed.”

***

‘Well… that’s one thing settled.’

I left the boardroom with a light step.

To open Milo’s vault, I need three keys.

GPU, GNN, and Ignus.

Of the three, the GPU issue was the easiest to handle, and it was now resolved.

Once production begins, the first shipments will go to Stark, and from there, the supply will naturally flow to Next AI, Stark’s affiliate.

That takes care of one major hurdle…

Now it’s time to tackle the remaining two.

GNN and Ignus.

Here, GNN serves as the stethoscope needed to open Milo’s vault.

It’s an AI model with a neural network structure, specialized in detecting relationships within complex, entangled data.

On the other hand, Ignus is like a precision tweezer.

It’s a development framework for building and training deep learning models, used to implement the actual system based on the clues captured by the GNN.

The problem was that neither of these technologies had yet been unveiled to the world.

No one even knew the names or whereabouts of the developers.

But to open the vault, I had to find these unknown talents no matter what.

The method I chose was simple.

Push Stark into action.

“To win the resource war, shouldn’t we secure the most critical resource first?”

“What do you mean by critical resource?”

“Talent.”

I proposed an idea to Stark.

“How about holding an AI forum? A gathering where industry professionals can come together to discuss the future.”

Given Stark’s status and the current momentum, anyone in the AI field would be eager to attend.

The talent I was searching for would be among them too.

“That’s a great idea.”

Stark accepted my suggestion without hesitation and quickly organized the forum.

He made the invitation open to “anyone interested in AI” and personally sent out invites to every single employee at Gooble.

It was, in effect, a provocation.

As if to say, “I can poach every one of your people.”

And to top it off…

“We invite you to a place where AI knowledge is shared.”

That phrase was printed right on the invitation.

Clearly, it was a jab at Gooble’s recent “sharing” campaign—a deliberate attempt to provoke.

‘What a character.’

But that wasn’t my concern.

I only needed to find the talent I was after.

I planned the forum’s program in collaboration with Next AI, and I created two special sessions.

They were bait.

<Graph Thinking: Beyond Sequences, Beyond Grids>

This session was bait to attract GNN developers.

<Dynamic Graph Utilization in Deep Learning Frameworks>

This session was aimed at drawing out the developers of Ignus.

Both topics were still abstract concepts that hadn’t even been developed yet, but the developers behind them must have already been forming similar ideas at this point in time.

And if, at an AI event, they came across a concept they believed only they had thought of?

They would be unable to resist showing up.

‘Not many attendees.’

Because both were such unfamiliar topics, fewer than ten people attended each session.

Fortunately, identifying the developers I was looking for wasn’t too difficult.

During the Q&A, while most attendees asked, “What exactly does this concept mean?”, a few sharp questions stood out.

“When multi-hop message propagation is repeated, over-smoothing happens far too quickly. Are there any structural approaches to mitigate this?”

“...With the Define-By-Run method, I have no clue how to implement graph visualization or debugging tools. Are others struggling with this too?”

These questions reflected the kinds of trial-and-error that only come up when you're actually developing and testing these ideas.

‘Those are the ones.’

I approached them and made an offer.

“Would you be interested in working at Next AI?”

“Huh? Did you say Next AI?”

It was a small nonprofit on paper, but their eyes lit up.

In the AI industry, very few people don’t know that Next AI is the real developer behind MindChat.

Right now, it’s practically the beating heart of cutting-edge LLM tech.

“If you’re interested, we’ll hire you on the spot. The conditions are…”

Their eyes widened at what I offered.

The salary I proposed was in the top 0.1% of the industry.

And on top of that—

“We’ll help you bring your ideas to life within six months.”

That promise sealed the deal.

They wrapped up their current jobs within a few days and started at Next AI within a week.

I gathered the new hires and said:

“As I mentioned, I intend to implement the systems you’ve envisioned within six months. Tell me what you need, and I’ll make it happen.”

All they had to do was focus on developing their ideas.

I would provide the necessary resources and funding.

However, they looked puzzled.

“Wasn’t that just an exaggeration? How can we do this in six months...?”

“It’s not an exaggeration. I’m serious. Just tell me what you need.”

“But that’s... impossible with current technology.”

I shook my head with a smile.

“Don’t worry about that. Just tell me what you need.”

They exchanged uncertain glances, but eventually started listing requests.

“We’ll need a large-scale graph dataset. Also, a multi-GPU environment that supports distributed processing... It’s impossible to train this on a single machine, so we’ll need a node cluster of dozens of GPUs.”

“A dynamic graph engine, redesign of tensor operation modules...”

Compute resources, kernel structure, optimization engine, parallelization scheduler, graph compiler…

I listened to it all and nodded.

“Understood.”

“Understood…?”

“You’ll be hearing good news very soon.”

They still looked skeptical, but I simply smiled.

As I left the conference room, I fell into thought.

‘The seeds have been planted…’

They are the seeds.

I’ve just finished sowing them into the ground.

Of course, they’ll eventually grow on their own.

But I don’t have time to wait.

So if I want these seeds to sprout faster?

I need to water them constantly.

And in this case, that “water” is none other than capital.

I must invest heavily into the areas they mentioned.

Where capital flows, paths will open.

Of course, my own money won’t be enough…

But that doesn’t matter.

If I run out of my own money, I’ll just use someone else’s.

‘Now then… is it time for Phase Three?’

Looks like it’s time to move on to the next stage of the war.