How did Nvidia traverse three technological cycles, transforming from a gaming graphics card to a dominant force in AI Computing Power?

Author: Deep Tide TechFlow

Original Title: The History of Nvidia's Rise: From Gaming Giant, Crypto Mining Tycoon to AI Arms Dealer

On October 30, Nvidia's market value surpassed $5 trillion, exceeding the annual GDP totals of developed countries such as Japan and Germany.

From $12 at its IPO in 1999, adjusted for splits, NVIDIA has generated over 8000 times return over 26 years.

NVIDIA, the most enviable aspect is undoubtedly its “freedom from cyclical constraints”; it has always served as a foundational infrastructure, continuously “collecting taxes”. No matter what you do, you cannot do without it.

As the creator of GPUs, Nvidia seized the opportunity of the “PC wave” and entered thousands of households with the explosion of the gaming market.

Then, as the gaming business weakened, the crypto bull market arrived, and NVIDIA graphics cards were widely used for “mining” cryptocurrencies such as Ethereum, making a fortune quietly.

Subsequently, the smart car industry has risen, and its onboard chip business has also developed rapidly.

Finally, ChatGPT emerged, and Nvidia transformed into an AI arms dealer…

Looking back at Nvidia's growth history, it has also hovered on the brink of failure and bankruptcy time and again. Jensen Huang once shouted: My will to survive exceeds almost everyone else's will to kill me.

NVIDIA, the creator of GPUs

The birth of the graphics card (GPU) dates back to the 1990s.

At that time, some people in Silicon Valley proposed an idea: to ease the workload of the central processing unit (CPU) by using specialized chips such as sound cards for processing audio and network cards for handling networking. Similarly, it follows that manufacturing a chip specifically responsible for computer graphics output, that is, a graphics card, is a logical step. For example, the gaming console PlayStation, launched by Sony at the end of 1994, used a graphics card to process images.

However, at that time, there were many technology paths for graphics cards. The breakthrough point found by NVIDIA was to achieve 3D graphics acceleration through parallel computing, especially applied in the gaming field. Parallel computing refers to breaking down a complex task into multiple smaller tasks and processing them simultaneously to improve computational efficiency.

In 1999, NVIDIA launched a graphics card named GeForce. This graphics card was designed for gaming, focusing on “parallel computing,” which significantly enhances 3D graphics processing capabilities, providing a smoother and more realistic gaming experience.

The success of GeForce has enabled NVIDIA to rise rapidly and become a leader in the graphics card field.

At that time, it was not only NVIDIA researching graphics processing units, but NVIDIA successfully deeply associated itself with the label of “GPU inventor.”

NVIDIA's then marketing head Dan Vivoli promoted the concept of “graphics processing unit” (GPU) to market their chips. He believed that by repeatedly emphasizing that NVIDIA was the inventor of the GPU, they could become the industry leader.

Indeed, this was the case later on; Nvidia became synonymous with GPUs, paving a new path for itself by marketing GPUs.

NVIDIA, the big winner in the crypto bull market

NVIDIA's market value increased from $14 billion in 2016 to a peak of $175 billion in 2018, with the more than tenfold increase over two years possibly attributed to the cryptocurrency mining boom.

In 2017, the cryptocurrency market experienced a major bull run, attracting a large number of miners competing for GPUs, turning GPUs into cash printing machines, with global graphics card sales surging and prices skyrocketing.

Taking the NVIDIA GTX 1060 graphics card used by miners as an example, the purchase price before May 2017 was about 1650 yuan per unit, and after June 2017, it rose to around 2900 yuan.

NVIDIA has become the big winner behind the cryptocurrency bull market, receiving a windfall.

Benefiting from the cryptocurrency mining boom, NVIDIA's total revenue for the fiscal year 2018 reached a new high of $9.7 billion. Jensen Huang stated, “Our GPUs support the world's largest distributed supercomputing, which is why they are so popular in the cryptocurrency space.” Additionally, NVIDIA launched the GTX 1060 3GB and professional mining cards P106 and P104 specifically tailored for mining.

In 2020, after two years of a bear market, the cryptocurrency market set sail again, with Bitcoin increasing more than 2 times and Ethereum rising 4 times, while Nvidia once again became a beneficiary of the “crypto boom.”

NVIDIA has swiftly responded and actively participated in the mining market by launching the CMP series of professional mining cards. These cards eliminate graphics processing capabilities and have lower core peak voltage and frequency to enhance mining performance and efficiency.

At the end of 2020, NVIDIA released the RTX 30 series graphics cards, with the entry-level RTX 3060 priced at 2499 yuan and the RTX 3090 priced at 11999 yuan. However, with the surge in cryptocurrency prices, the RTX 3060 was sold for as high as 5499 yuan, while the RTX 3090 skyrocketed to 20000 yuan.

After the release of the Q1 2021 financial report, NVIDIA's Chief Financial Officer Colette Kress revealed that the sales of NVIDIA's cryptocurrency chips reached as high as $155 million, with graphics cards used for “mining” accounting for a quarter of the total sales in the first quarter.

In 2021, NVIDIA's annual revenue reached a record of $26.91 billion, an increase of 61% compared to the previous fiscal year, with its market value once exceeding $800 billion. However, the good times didn't last long. In September 2022, the Ethereum execution layer and proof-of-stake consensus layer completed the merge, transitioning the Ethereum blockchain network mechanism from PoW (Proof of Work) to PoS (Proof of Stake), signaling the gradual end of the GPU mining era.

This has also affected NVIDIA's development to some extent. In the third quarter of 2022, NVIDIA's revenue and net profit both declined, with quarterly revenue of only $5.931 billion, a year-on-year decrease of 17%, and net profit of only $680 million, a year-on-year decline of as much as 72%. On November 23, 2022, NVIDIA's stock price was reported at $165 per share, nearly half of its peak last year.

At that time, both overseas media such as “Financial Failure” and domestic technology media were pessimistic about Nvidia.

In extreme difficulty, unexpectedly, the tide turned, and the winds of AI and large models began to blow, with Nvidia once again standing at the forefront.

NVIDIA, AI Arms Dealer

In March 2016, AlphaGO defeated Lee Sedol, which was shocking and sparked a heated discussion about AI.

A month later, Jensen Huang officially announced at the GTC China conference that NVIDIA is no longer a semiconductor company but rather an artificial intelligence computing company.

In August 2016, a historic moment was born when NVIDIA donated its first AI supercomputer, the DGX-1, to the newly established OpenAI. Jensen Huang personally delivered this computer to OpenAI's office, where the chairman at the time, Elon Musk, opened the package with a box cutter.

Jensen Huang left a message: “To compute for the future of humanity, I donated the world's first DGX-1.”

Later, OpenAI trained the globally popular ChatGPT using NVIDIA's supercomputers, and NVIDIA's subsequently updated hardware product DGX H100 was in high demand, with supply unable to meet the demand.

Rome wasn't built in a day, and Nvidia's dominance in the AI industry began with earlier accumulation.

David Kirk, the former chief scientist of Nvidia, has long dreamed of generalizing the 3D rendering computing power of GPUs, not limited to the gaming field.

Under the leadership of David Kirk and Jensen Huang, NVIDIA launched the revolutionary GPU unified computing platform CUDA in 2007, unleashing massive computing power resources.

But at that time, CUDA did not impress investors at all; instead, due to the massive investment in creating a “supercomputing” system ahead of its time, Nvidia's profits were significantly reduced, and Wall Street booed as a result.

Ben Gilbert, the host of the popular podcast “Acquired” that is all the rage in Silicon Valley, commented: “They weren't targeting a large market at the time, but rather an obscure corner of academia and scientific computing, and they spent billions of dollars to do so.”

The voices from the outside world did not affect Jensen Huang, as his consistent investment in CUDA over the past decade has led NVIDIA to its current position.

Jensen Huang views computing power as core. Whether it is AI, autonomous driving, the metaverse, robotics, or cryptocurrencies, NVIDIA is leveraging massive computing power to seek new opportunities.

Computing power, Nvidia's eternal weapon.

Three failures

In 2023, Jensen Huang delivered a speech at the graduation ceremony of National Taiwan University, where he shared three stories of failure and imparted the secrets of Nvidia's success to the students.

Survived on the brink of bankruptcy after the first failure.

In 1994, Nvidia's first customer was the Japanese gaming company SEGA, which designed graphics cards for its game consoles.

However, in the second year, Microsoft released the graphical interface Direct3D for the Windows platform, which caused Nvidia to feel very flustered due to conflicts with their design.

Ultimately, Nvidia chose to terminate its contract with SEGA and instead develop GPUs for the Windows platform. This is a risky decision, as SEGA was their only customer, yet they abandoned it. Nvidia's funding can only support them for 6 months, and if they cannot launch a new product within this time, they will face the risk of bankruptcy.

Fortunately, just one month before bankruptcy, as funds were about to run out, Nvidia designed the Riva 128 chip and achieved success. By the end of 1997, the shipment of Riva 128 exceeded 1 million units, allowing Nvidia to survive.

The second failure, giving up short-term profits, has achieved great things in the future.

In 2007, NVIDIA launched the CUDA GPU acceleration computing program, aiming to make CUDA a programming model that can enhance a variety of applications, from scientific computing and physical simulations to image processing.

Creating a new computing model is very difficult; the CPU computing model has existed as an industry standard for nearly 60 years since the launch of the IBM System 360.

CUDA requires developers to rewrite applications to showcase the benefits of GPUs; however, to develop such programs, there needs to be a large user base and significant demand to drive developers to create.

To solve the “which came first, the chicken or the egg” problem, Nvidia leveraged their large user base of gamers using GForce graphics cards. However, the additional costs of CUDA were very high, leading to a significant decline in Nvidia's profits over the years, and their market value has been fluctuating around the billion-dollar mark.

NVIDIA's years of lackluster performance have led shareholders to be skeptical about CUDA. Shareholders are more eager for the company to focus on improving profitability, but NVIDIA has persisted, believing that the opportunity for accelerated computing will come.

Jensen Huang founded a conference called GTC, tirelessly promoting CUDA worldwide. In the end, hard work pays off, and a host of applications truly emerged, including CT reconstruction, molecular dynamics, particle physics, fluid dynamics, and image processing.

Until 2012, AI researchers discovered the potential of CUDA. Renowned AI expert Alex Krizhevsky trained AlexNet on the GForce GTX 580, triggering an explosion in artificial intelligence.

The third failure, NVIDIA exits the mobile chip market.

Do you still remember the stage sharing between Lei Jun and Jensen Huang?

In 2013, at the invitation of Lei Jun, Huang Renxun attended the launch event of the Xiaomi Phone 3.

In his youth, Huang Renxun went to the United States and was asked by Lei Jun to speak Chinese. He wasn't very fluent, but confidently exclaimed in Chinese: “NVIDIA's GPU is the best in the world.”

At that time, the Xiaomi 3 flagship version was equipped with the mobile version of the Tegra4 processor launched by Nvidia, which was also the swan song of this series.

At that time, the mobile phone market was on the rise, and Nvidia also entered the mobile chip market. Although the entire phone market was very large and Nvidia could have fought for market share, they made a tough decision: to abandon this market.

Jensen Huang stated that Nvidia's mission is to build computers that ordinary computers cannot, and they should be committed to realizing this vision and making unique contributions. Nvidia's strategic retreat has been rewarded.

Life advice: Experience suffering, lower expectations

In 2024, Jensen Huang returned to his alma mater, Stanford University, and gave a speech at the business school, sharing some life experiences.

When the host asked Jensen Huang if he had any advice for Stanford students about success, he replied, “I hope you all have the opportunity to experience a lot of pain and hardship.”

He mentioned that one of his greatest strengths is “my expectations are very low.”

Jensen Huang stated that most Stanford graduates have very high expectations of themselves, but they absolutely deserve those high expectations because they come from one of the best universities on Earth, surrounded by equally incredible peers, and having high expectations is a very natural thing.

“People who have very high expectations of themselves often have low resilience,” said Huang Renxun. “Unfortunately, resilience is critical to achieving success.”

Jensen Huang emphasized, “Success does not come from intelligence, but from character, and character is shaped by enduring hardships.”