Reborn in 2010, the AI ​​in my mind is too powerful.

The enjoyable holiday ended, and Hannah went back to school.

The afternoon after returning from Palm Springs, Chen Yiran also went to the company to meet the new employees.

Zhou Feng sat in a corner by the window, three monitors on his desk displaying dense OPC simulation data. Jin Chenghao had already set up his development environment next to him. Matthew's workstation was nearby, and he was also busy.

I greeted the three new employees and casually made some promises.

As usual, he then gathered the employees together, gave them a pep talk, and promised that he would start looking for a new company address immediately.

I didn't expect Jasmine's framework to be so strong, to develop so quickly. In just two months, this space is already a bit insufficient.

However, Chen Yiran was also conflicted about what size to rent. He could only tell Emily to find several different sizes for him to see, so that he could choose one that he liked in the end.

Emily was very dissatisfied with this irresponsible behavior, since Catherine was now relying on Chen Yiran for her livelihood.

Catherine had gotten used to Chen Yiran's occasional strange ideas and considered them as flaws that God set for geniuses, so that people wouldn't be overly envious of geniuses.

The rest of August was indeed a harvest season. Chen Yiran basically had TED help her review the contracts and then signed them herself.

The most important of these is, of course, the PDK certification that was passed so quickly.

Without PDK certification, even if a design company really likes your software, they simply won't sign a contract. This is because the final product always requires the manufacturer's approval before it can be used; otherwise, the design is useless.

So after returning from the exhibition, Gao Dewei, despite his busy schedule, continued to go through the verification and certification processes with the three OEM factories.

GlobalFoundries, the least busy of the three, was the first to finish the test. They said at the exhibition that it would take a month, but they actually completed it in three weeks.

This success was thanks to the company's internal emphasis and Chen Yiran's efforts; his employees were incredibly proactive. They diligently checked every detail, including the minimum spacing of the metal layers, hole size tolerances, and stress effect compensation parameters, ensuring complete format compatibility.

Martin of GlobalFoundries had been closely monitoring the situation and notified Chen Yiran on the same day he received the self-inspection report. He then promptly updated the relevant information on GlobalFoundries' ecosystem partner page, adding a gray Grey logo icon under the "Certified EDA Partners" section.

Samsung's PDK was only one day later than GlobalFoundries', which is an incredibly fast pace for a company based in South Korea and of enormous size, showing none of the bloated sluggishness of a large corporation.

Although the contract drafted by Park Jae-sun's legal department was long, the technical certification process was indeed very fast.

This is likely also due to Samsung's own eagerness for OPC and its unwillingness to waste any of Chen Yiran's time on PDK.

In response, Chen Yiran could only silently curse TSMC in his heart, thinking, "Look at others, look at you, look at Li Ka-shing next door. We gave you a chance, but you didn't use it. Just watch, you'll regret it to death in the end."

The last one is SMIC. Engineer Zhou said it would take a month and a half, and it really did take a month and a half.

After the three PDK certifications were in place, Chen Yiran notified TED to contact several design companies to discuss the formal contract.

Chen Yiran priced Grey at $6 per person per year, with a minimum of five seats available. Of course, he was open to negotiating preferential policies with individual companies he favored.

This price is quite reasonable, but it's not exactly cheap either.

After all, the annual fee for the three major services is roughly between $12 and $15.

For a single-point tool, the price of 6 for Grey is just right.

Moreover, Chen Yiran wasn't expecting to make too much money from this, because relatively speaking, Grey's physical design requires the fewest chip slots. Chen Yiran initially chose this as the entry point for chips purely because of Jasmine's suggestion that Deepflow was more suitable for this approach.

To truly make money, one must try to establish a presence across the entire industry chain.

Right now, all I can say is that I'm just making a little money through hard work. After all, even Intel might only have a hundred people needing grey, while for smaller companies, a basic starting seat is usually enough.

For example, Hisilicon, the first company to sign up this time, gave very positive feedback from their internal trial of Grey. So they didn't even wait for the 30-day free trial to expire before hastily sending over a twenty-page annual license agreement.

In the end, they only asked for 10 seats, which meant paying FlowEDA $60 a year.

This can be considered FlowEDA's first revenue.

Next was Rockchip. After completing the 30-day trial, they were also satisfied with the results. After confirming that SMIC had PDK certification, they quickly placed an order.

Compared to HiSilicon, they were a bit short of funds and ultimately set a starting price of $30 for five seats.

SiTime's CTO confirmed the order via email after GlobalFoundries' certification. They said that GlobalFoundries' mixed-signal dedicated line had no problem with their MEMS clock chips. After the placement was moved to Grey, the area was reduced by more than three points. They decided to buy five at a time, and another 30 yuan came into their account.

Broadcom was the last to sign the contract. After Han Yu survived the three days of macrocell congestion, the other party's technical review was delayed for a while. The internal process was only completed after the certification and trial data were in place.

They were much more generous, paying for 30 seats in one go, bringing in $180 million.

Finally, there's the tragically pitiful Adapteva.

Olofsson sent an email stating that Grey had reduced the area of ​​their second-generation test chip by nearly 7%, resulting in a lower starting point and greater potential for improvement. The email included a screenshot of their internal design tool interface, with Grey's logo already embedded in the toolbar. So far, everything seemed to be going smoothly, which made Chen Yiran a little uneasy.

This guy can't succeed in America; he has to be executed. Otherwise, how could we have tricked him into coming here?

Chen Yiran made up his mind to develop CUDA at a frantic pace, so that Olofson's grand vision could not be realized.

If it's not commercially successful, then there's no point in talking about it theoretically anymore.

NVIDIA's trial has also begun. A director from the GPU physics design team came directly to handle the process, and the data handover and acceptance standards were all laid out on the first day after the contract was signed.

Chen Yiran was confident that it wouldn't be long before Nvidia would come and pay. Although their team wasn't as impressive as Intel's, it was still possible for them to have forty or fifty seats.

At the end of the month, the Hot Chips technology conference began.

Chen Yiran had carefully considered whether to take this opportunity to promote his company. After all, the location was in Stanford, which he was familiar with.

However, the OPC acceleration prototype for Samsung is still running, and although the speedup has reached eight times, it's just one step away from the acceptance standard of ten times. But it's still not perfect, and it's a bit of a shame to release it now.

After careful consideration, Chen Yiran decided to hold back and wait until the NVIDIA GTC conference in September to make a big move.

Therefore, Chen Yiran attended the conference entirely from a learning perspective. The conference was indeed very exciting, with each party presenting significant news.

AMD took a big gamble, releasing a new chip model that pioneered a modular design with a shared front-end and independent integer arithmetic units, attempting to achieve nearly double the multi-threaded performance at the cost of a very small chip area.

IBM, the "Big Blue," demonstrated its raw power with the unveiling of the Z196 mainframe processor, boasting an astonishing clock speed of 5.2GHz—the world's fastest processor at the time.

Huawei also demonstrated its innovativeness, boldly employing an unconventional solution: dividing the 32MB of embedded memory into 16 blocks, with each block directly connected to a dedicated processing engine via an internal high-speed mesh. This directly increased the network packet processing bandwidth by 16 times, showcasing its design capabilities.

Of course, none of this compares to a young researcher's paper presentation, in which he coined a very cool new term, a new concept: dark silicon.