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

In the early 2000s, GPUs were still pure graphics pipelines, with separate vertex and pixel shaders, making programming difficult and only allowing the use of graphics application programming interfaces.

Although some people in the financial industry had already started using GPUs for general computing, such as using pixel shaders to calculate fluid dynamics, it required bypassing the graphics application programming interface, which was extremely troublesome.

In 2003, Ian Buck, a doctoral student at Stanford University, published a paper and presented it at the annual conference on computer graphics, titled "Brook for GPUs".

This is the first paper to demonstrate that a compiler can abstract a GPU from graphics hardware into a general-purpose stream processor.

Jensen Huang astutely realized that this was the future of Nvidia.

He met Ian Buck with a check for 500,000 and a top-of-the-line GPU prototype. The following year, Ian Buck joined Nvidia and was given a great deal of freedom and power by Leather Jacket.

Ian Buck, along with his hardware and software engineers, began creating a general-purpose graphics processing unit (GPGPU) model, which became CUDA.

Two years later, in 2006, NVIDIA officially launched the CUDA platform, the world's first general-purpose computing solution on GPUs.

From then on, using GPUs for general-purpose computing no longer required disguising matrix multiplication as texture rendering and using graphics application programming interfaces to trick the graphics card.

Leather Jacket Huang clearly knew that this was Nvidia's future, and Nvidia's greatest opportunity to go from being a game company to reaching for the stars.

So he began to wholeheartedly support CUDA, and from 2006 to 2012, Nvidia invested more than one billion US dollars every year.

Wall Street shareholders collectively opposed the move, and the internal graphics team also complained that the space occupied by CUDA logic circuits caused a 10% drop in game performance, but none of this made him compromise.

The most dangerous incident occurred in 2008 when the thermal expansion coefficients of the bonding materials between the GPU and the PCB board in the 65nm process packaging were mismatched. During packaging, the bonding materials between the chip and the PCB board had defects, which could easily lead to open solder joints at high temperatures.

NVIDIA's G84/G86 core graphics cards, such as the GF8400 and GF8600 series, have hardware defects that can easily cause laptops to overheat, display screens, or fail to boot.

This not only forced Nvidia to spend huge sums on recycling and repairs, but also greatly dissatisfied a number of its previous partners.

In addition, the subprime mortgage crisis hit, causing Nvidia's annual profit to plummet by 74% and its stock price to drop by more than 80% from its peak, resulting in Huang Renxun's personal wealth evaporating by 90%.

Even at this point, Leather Jacket Huang did not stop funding CUDA.

He launched the CUDA Center of Excellence and CUDA Teaching Center programs, directly donating GPU hardware, textbooks, and software licenses to various universities.

By 2012, more than 350 universities worldwide had offered CUDA courses, with a cumulative total of over 100,000 active developers.

The annual NVIDIA GTC Developer Conference, organized by NVIDIA, has grown from a small developer exchange meeting with 300 people into a large-scale event with nearly 3,000 people, more than 50 exhibitors, and more than 200 technical presentations.

Chen Yiran's appearance further convinced Huang, the man in the leather jacket, that Nvidia was about to reach its peak.

The DeepFlow framework has made various deep learning technologies essential to the industry, and the launch of AI has led to an exaggerated shortage of Nvidia's graphics cards.

However, this favorable situation only lasted for a year. AMD made its move, shamelessly giving away shares and seats to pull Chen Yiran into OpenCL.

With sufficient technical support, AMD finally has a cross-platform parallel programming standard that can rival CUDA.

In just one year, AMD began to exert its might and frantically eroded Nvidia's market share.

Normally, Nvidia could accept its fate and focus on competing with AMD. DeepFlow's framework seems too amazing, and Nvidia's engineers unanimously believe that it is impossible to compete with it.

However, Huang was not satisfied. In his vision, regardless of the AI ​​framework, NVIDIA should leverage the more mature CUDA ecosystem to make NVIDIA graphics cards the mainstream in the market.

He believes that once Nvidia reaches that point, it will become a company on par with Apple and Microsoft, and one of the main players in Silicon Valley.

So even though he knew Chen Yiran had a terrible temper and was already eyeing Google, he still secretly partnered with Google.

Engineers from Google and Nvidia will collaborate in hopes of leveraging better hardware optimizations to enable Google's Transformer to achieve performance levels comparable to Deepflow on Nvidia graphics cards.

However, Huang, the leather jacket owner, was also quite shrewd. He signed a contract with Google stipulating that the collaboration would only be announced after the project was successfully completed.

In other words, the performance of Transformer on NVIDIA graphics cards can reach 80% of that of the Deepflow framework on AMD graphics cards of the same level.

Only in this way would Nvidia dare to break with Chen Yiran and confront him head-on.

Huang, the man in the leather jacket, was also feeling bitter. Apple was now almost a trillion-dollar company, and Aurora Group was also a behemoth.

Nvidia is currently just a small company worth 20 billion US dollars, and at least half of that market value comes from its cooperation with Chen Yiran. He is afraid that this immature Chen Yiran will overturn the table and treat Nvidia the same way he treats Google.

However, he had no idea that Nvidia, a company that Chen Yiran had always been very wary of, had been secretly monitored all along.

In response, Chen Yiran began to frantically whip the cattle and horses below.

If the Coffee Project's chip is successful, it can completely break free from its dependence on Nvidia and AMD for chip development.

In the future, we will have control over the entire AI industry chain.

By September, the coffee project team had finally completed the optimization of the previous plan.

Samsung and Wongoyuan were both assigned the task of expanding the 16×16 chip to a full-size fabrication for testing, and reliability verification is expected to be completed by the end of October.

The merger between Apple and Disney is also progressing very smoothly. Disney has reached a $40.5 billion acquisition agreement with Lucasfilm, and the formal handover will take place on October 15.

Apple and Disney will then announce the acquisition at lightning speed, although this process is expected to take more than half a year.

Time Warner is pulling a bit, you could say they're stuck.

They are still very willing to sell themselves, but hope to discuss it again after Man of Steel is released in 2013.

It's obvious they're seeing Marvel raking in money like crazy and their own Batman trilogy doing so well, hoping this Superman movie can launch a successful DC universe.

Chen Yiran didn't care. He knew that DC movies were basically unreliable.

While Man of Steel was decent, it was definitely not a commercial success, and by 2013, DC's price would only be lower.

Therefore, he instructed Cook to maintain contact first, and that an acquisition in 2013 was also possible.

Many things cannot be rushed, and Chen Yiran's own laboratory also encountered problems that required waiting.

After spending two months with Olson, frantically catching up on all sorts of knowledge, he finally formulated a preliminary experimental plan and began placing orders for customized equipment.