The Nvidia headquarters is located in Santa Clara, about a forty-minute drive from San Francisco Airport.
Silicon Valley The sunshine in June is bright and dry,
The palm trees on both sides of the road remained motionless in the heat wave,
There is California's unique eucalyptus smell in the air.
Lin Shicheng rented a black Cadillac Escalade in advance,
The air conditioner was on full blast, but he still loosened his tie.
Before flying over from Manhattan, he went to get a haircut and put on a new shirt.
At this moment, I am sitting in the passenger seat and repeatedly checking the meeting process on my phone.
That taut Legal Director-like dignity looked like an overly formal coat under the California hot sun, but he didn't care.
Huang Renxun has been studying this person in the SEC's 13D filing documents for a long time--
Paranoid, direct, and lacking basic patience for capital who do not understand technology,
Those who are willing to sit down and study the data are willing to share their knowledge.
In today's meeting, every detail of the first impression cannot go wrong.
Zhao Chen is sitting in the back row, with Kraft Paper File Bag on his lap.
Soon, the triangular glass building of Nvidia headquarters emerged from the horizon.
He had seen this building countless times in Bloomberg Terminal's news photos--
All-glass curtain wall, sharp geometric corners,
There is a certain self-evident echo with the heat dissipation fins on the Nvidia GPU chip.
Before, when he was sitting in the Manhattan hotel studying Nvidia's Financial Report,
This building is just a background image on the screen that has been zoomed out so that the details cannot be seen clearly.
Now it truly stands under the scorching sun of California, with every piece of glass reflecting the dazzling white light.
The car is parked in front of the visitor center.
Lin Shicheng went to the registration office to go through visitor procedures, and Zhao Chen stood waiting in the lobby.
There is a huge black and white photo hanging on the wall of the hall -
A group photo of the three founders of Nvidia signing company registration documents at Denny's restaurant in 1993.
In the photo, Huang Renxun still has thick black hair and is wearing the denim shirt that was standard for Silicon Valley engineers in that era.
He is holding an ordinary ballpoint pen in his hand, and his expression is not celebrating.
It was as if he was staring at some coordinates that only he could see.
Zhao Chen stood in front of this photo for a long time.
Later he talked to Huang Renxun at the Silicon Valley dinner about the afternoon at Denny's--
With only two hundred US Dollar in their pockets, the three of them huddled in the booth, eating pancakes and drawing sketches of the NV1 architecture.
The restaurant waiters thought they were writing term papers - but that's another story.
At this moment he is just standing in front of this black and white photo,
The moment I watched the young immigrant sign his name with a ballpoint pen, he had already decided the path he would take for the next twenty years or so.
Lin Shicheng came over after completing the formalities, holding a visitor badge in his hand.
The two were led to a private room on the top floor Meeting Room,
There was no PR team in the corridor when the elevator door opened,
There are no people from the IR department shuttling back and forth with agendas.
There are no souvenirs or gift bags with the company logo printed on them.
Only an administrative assistant in a dark gray suit led them to the door of Meeting Room.
He opened the door gently, then turned and left.
Huang Renxun stood in Meeting Room waiting for them.
He was wearing a black leather jacket, exactly the same as when he took the stage to speak at the GTC conference.
Huang Renxun strode over and shook hands with Zhao Chen.
His handshake was very strong, not the touch-and-go kind of social etiquette.
This is the habitual force method of an engineer who has personally tightened countless screws in the Fab workshop.
"Zhao Chen. Lawyer Lin told me that you flew here for more than ten hours.
Are you tired? If you are, go and have a cup of coffee first and then chat. If you're not tired--"
He pointed to the whiteboard,
"I just drew something on it, please help me take a look."
Zhao Chen glanced at the densely packed markings on the whiteboard.
Data Center revenue growth rate, CUDA developer number growth curve, computing power demand elasticity coefficient of AI training tasks--
The actual data for the latest quarter are marked next to each variable.
Several sets of core data are circled with red pen, and the arrows point to the deduction interval on the right.
Zhao Chen put Kraft Paper File Bag in his hand on the conference table,
I walked to the whiteboard and stopped looking at those sets of data for a few seconds.
"Is this the latest Q1 actual data?
Data Center's revenue growth rate is much higher than the guidance given in the last quarter's conference call."
"After channel inventories bottomed out, demand rebounded stronger than expected.
Capital expenditures by cloud service providers have accelerated since the beginning of the year.
The elasticity coefficient of computing power demand for AI training tasks is 0.3 percentage points higher than the upper limit estimated by Model."
Huang Renxun speaks very fast,
Every technical term pops up as precisely as if it had been processed by Tensor Core,
But his eyes never left Zhao Chen's face. That kind of looking at each other was not polite eye contact in a business setting.
It's one engineer testing whether another engineer can keep up with his own deduction pace.
"The deviation of zero point three is within the normal range.
Our independent Model also captured this upward trend in elasticity coefficient, about two months late -
Because we use public data, cloud service provider capital expenditure details are updated more slowly than internal data.
Have you compressed the Model tracking window to the weekly level?"
Zhao Chen walked to the whiteboard and tapped the line of numbers on the elastic coefficient with his finger.
"Yes. It used to be quarterly tracking, but in the past six months it has been adjusted to weekly level.
The changes in AI are too fast and we can't keep up quarterly. Waiting for Financial Report to come out before adjusting assumptions will be three months too late."
Huang Renxun picked up a red marker pen from the table and re-corrected the number next to the elastic coefficient.
"Are you doing Model yourself or using external data?"
Zhao Chen picked up the blue marker and drew a parallel line next to the elastic coefficient.
"Made by myself. Henglong Research Institute independently developed valuation Model,
All input variables come from public data--
Nvidia Financial Report, conference call minutes, industry research reports, and cloud service providers' publicly disclosed capital expenditure plans.
We do not cite any internal information or conduct Model cross-verification with your IR department."
Huang Renxun stopped writing, turned around, and moved his eyes from the numbers on the whiteboard to Zhao Chen's face.
"Wait a minute. You said your Model only relied on public data. How long have you been following it?"
"It borrows the original data of Era Vanguard's first purchase of Nvidia stock."
"That's more than just a few quarters.
There were a lot of fluctuations in the process--
Liquidity crisis, global Supply Chain interruption, and channel inventory storm after the earlier wave of mining tide receded.
You told your legal team last time that you continued to buy during the most difficult times not because you knew in advance that it would rise back.
Is it relying on this Model? At that time, your Model had not gone through so many iterations, why do you believe it?"
Zhao Chen put down the marker and walked back to the conference table.
Take out the Henglong Research Institute Independent Tracking Manual from Kraft Paper File Bag,
Turn to the pages with previous quarterly review records and place them in front of Huang Renxun.
Those records started a long time ago,
Updated every quarter,
Each update includes Model output conclusion, actual data backtest comparison and Model adjustment instructions--
Data Center revenue growth rate, CUDA developer growth curve, AI Computing Power demand elasticity coefficient,
The actual trends of the three core variables in previous quarterly reviews and the comparison curves predicted by Model are all compressed in the appendix on the same page.
"When we first started this group of Model, few people in the market believed that Nvidia's Data Center business could surpass the game business.
After the cryptocurrency bubble burst, channel inventory piled up in warehouses,
Wall Street said that the growth rate of your game business has peaked, Data Center is not big enough, and the price of the new generation Turing Architecture is too high and there is no market.
I was sitting alone in the Manhattan hotel, and your stock price on Bloomberg Terminal was still falling.
I took apart all the public data you can get."