In 2016, Jensen Huang, the co-founder of NVIDIA, showcased the groundbreaking DGX-1, the company's first deep learning supercomputerThis machine was donated to a startup that would later take the world by storm—OpenAI, which developed ChatGPTOpenAI's journey was significantly backed by Elon Musk, who co-founded the organizationThe DGX-1 was valued at $129,000 at that time, a hefty investment that would pay dividends that no one could have foreseen.

Fast forward seven years, and NVIDIA's A100 chips have become a hot commodity, with supply barely meeting the soaring demandIn late May 2023, NVIDIA's market capitalization surged past an impressive $1 trillion, solidifying its position as the fifth largest company in the U.SmarketTo put it in perspective, NVIDIA's market value outstrips that of renowned tech giants like Intel and AMD, with figures unparalleled in recent history within the tech sector.

It's fascinating to look at the transformation NVIDIA has undergone over the years

Just a few years earlier, NVIDIA was primarily recognized for its graphics cards targeted towards gaming and its computational power in the cryptocurrency mining sectorThe company's evolution from a gaming tech supplier to a pivotal player in AI, serving as a cornerstone of technological development under U.Strade regulations, raises an engaging question: what led to this dramatic shift?

This article will delve into the story of NVIDIA, examining how it ascended to the title of 'King of Chips.' The narrative highlights its transition from graphics processing units (GPUs) to gaming graphics cards, its pivot from cryptocurrency mining to AI capabilities, and the factors behind Huang’s success in building a chip empireWhy are GPUs the darlings of the AI era? Can NVIDIA remain irreplaceable in the global tech landscape, and what does its ascent symbolize for the future?

The journey begins in 1993, the year China officially connected to the international internet, signaling the dawn of the internet age in the nation.

That same year, Nokia made its foray into the Chinese market

• Why Did Bitcoin Suddenly Plummet?
• Record Losses for the Federal Reserve!
• Huawei Chip Breakthrough Threatens ASML
• Domestic Production of Solid-State Lithium Batteries
• Bitcoin Could Reach $120,000

It was also a time when the Central China Television launched a new program called "Oriental Horizon." In the U.S., cable television and MTV didn't yet exist, IBM and Apple's Macintosh were initiating a personal computer revolution, while Intel and AMD were basking in the glory of their revolutionary Pentium CPUs.

Jensen Huang, who had been in the semiconductor industry for a decade by that time, was just turning 30 and had worked in chip design for AMD, along with various roles that combined technical management and sales at LSILogicWith varied experiences came rising ambition, inevitably shaping his path forward.

Yet, 1993 held a special significance for HuangBorn in Taipei in 1963, he was sent to live with an uncle in Washington State at the age of nineHis economically challenged uncle enrolled him in a rural boarding school, so it wasn't just an academic journey—it was almost a quest for self-discovery.

Huang embodied the quintessential characteristics of hard-working students, and he eventually secured a place at Oregon State University, majoring in electronics

While working in a laboratory, he made a proud promise to his first girlfriend, Lori, declaring that he would have his own company by the time he turned 30. As fate would have it, 1993 became that milestone yearHuang subsequently recruited two of his former technical partners and co-founded NVIDIA, taking on the role of CEO while deliberately choosing to start on his birthdayWith a keen focus on graphics processing technology, they were eager to tap into the burgeoning market for video games and multimedia driven by PCsHowever, their initial endeavor taught them a harsh lesson.

In 1995, NVIDIA introduced its first multimedia accelerator aimed at gaming consoles, the NV1, capturing the interest of the Japanese gaming giant SegaSega signed a $7 million contract to fund further development of the NV2 for its upcoming gaming console.

However, as Microsoft unveiled Windows 95 and Direct3D, the gaming landscape shifted towards PCs, placing NVIDIA in a precarious position as their heavily invested NV2 wasn’t compatible with Windows standards

The company faced a dilemma: failing to deliver the contract could lead to bankruptcy, but pushing on felt futile, akin to an employee realizing their project specifications had changed overnight.

Caught between a rock and a hard place, Huang boldly decided to recognize the misstep in direction with Sega and promptly halted work on NV2.

NVIDIA managed to rebound strategically; in 1997, they launched Riva128, a new graphics accelerator aimed specifically at the personal computer marketWith an unbeatable price-performance ratio, they sold over a million units within just four months.

From that point on, their trajectory changed, setting themselves against 3DFX, the then-dominant player with an astonishing 80%-85% market share in 3D graphics accelerators.

Huang implemented an ambitious three-team, two-quarter R&D model, allowing NVIDIA to release new products every six months, keeping the company ahead of competition by one or two development cycles and thus creating their own market demand.

Huang coined the “Huang’s Law,” directly correlating with Moore's Law, asserting that NVIDIA could double its performance every six months

In 1998, they launched the even more powerful Riva TNT, putting them on a level to rival 3DFX.

Moreover, NVIDIA made two critical strategic decisionsFirstly, they closely aligned with Microsoft during a chaotic time in the graphics processor market, where everyone had competing standards and technologiesBrands like 3DFX flaunted exclusive APIs while NVIDIA embraced Microsoft's direct API standards from the beginning.

Secondly, they partnered with TSMC, a powerhouse that focused on production, which allowed them to manufacture chips with better performance and lower costs compared to 3DFX.

In a pivotal meeting, Huang confidently expressed to TSMC’s CFO that NVIDIA would soon become their largest customer.

Looking back today, Huang's assertion didn’t seem like mere bravado; less than a year later, in 1999, NVIDIA witnessed explosive growth.

Their successful IPO introduced the world to the concept of GPUs

The GeForce256 GPU dramatically improved performance, leaving competitors scrambling and establishing graphics cards as a central component of PCs.

In the same year, NVIDIA's market share surpassed that of 3DFX for the first time, and by the following year, they absorbed 3DFX, solidifying their leadership in the graphics card arenaThey then braced for another uphill battle against the famed ATI, a well-established graphics company.

Following the acquisition of 3DFX, NVIDIA's focus narrowed to a single rival—ATI, founded by the Guangdong-born Henry WongAt that time, ATI responded to NVIDIA's GeForce256 with their own Radion 256 graphics card, which, depending on how one measured performance, was quite competitive.

By 2003, China overtook Japan to become the world's second-largest computer market.

That October, when ATI launched their Radeon 9800XT in China, a wave of excitement erupted among gamers

Huang was quick to respond with NVIDIA's latest GeForce FX5900 just four days laterThis back-and-forth rivalry heightened passionate engagements among consumers.

During this intense competition, financial scandals sidelined ATI’s founder Wong, leading to a decline in ATI's fortunesOver time, through successive product launches, NVIDIA consistently outperforming ATI also drove prices down, particularly seizing a significant market share in the mid-range demographic.

By 2006, after AMD acquired ATI, NVIDIA's dominance became apparent, clinching more than 80% of the desktop discrete graphics market share, while ATI struggled to recover from increasing debt.

At this point, NVIDIA resembled a technological giant like Nokia, while the only real forces that could potentially challenge it were changing market demandsIt became clear that merely selling GPUs to gaming enthusiasts would not sustain this success forever—the company needed to identify a new avenue for growth.

In an interview, Huang revealed that they believed focusing solely on graphics processing wasn't sustainable for the long term

The problem was simple: GPUs were generally too specializedWhile CPUs could be utilized across various computer applications, GPUs mainly catered to 3D rendering, often resulting in idle capacity—an inefficient use of resources.

In 2006, NVIDIA's Chief Scientist David Kirk recommended diversifying the GPU technology beyond just 3D rendering into General-Purpose computing on Graphics Processing Units (GPGPU) to unlock its vast potential.

Despite initial resistance from the board, Huang convinced them of the necessity of building this technology, asserting, "If we don’t build it, no one else will."

NVIDIA Financial Projections

NVIDIA proceeded to develop the famed CUDA platform to enable all its GPUs to support programmingThey did incur significant expenses during this transition, from $500 million yearly in R&D costs to increased chip sizes, greater heat output, and quality control challenges.

Consumers were initially reluctant to pay a premium for chips packed with saturated features that many wouldn’t use, such as panoramic sunroofs and gesture controls—akin to what NVIDIA offered with CUDA.

During this challenging era, NVIDIA faced major profit declines and a five-year downturn; people can view this as one of the company's riskiest gambles.

In 2008, as smartphones gained traction, NVIDIA attempted to penetrate the mobile market with its Tegra chips

However, due to Qualcomm’s dominance in modem technology and difficulties in meshing processor and modem, NVIDIA ultimately retreated from the smartphone frontier.

Fortunately, NVIDIA returned to form upon realizing CUDA's monumental potential.

In 2012, deep learning pioneers Geoffrey Hinton and student Alex implemented AlexNet—a revolutionary deep neural network—utilizing two GTX 580 cards over six days, causing a major stir at the ImageNet Large Scale Visual Recognition Challenge.

The GTX 580 was a robust gaming card, sold in China for roughly 4000 RMB, but its capabilities far exceeded initial expectations, convincingly outclassing competitors in the AI arena.

Hinton himself noted that without NVIDIA, AlexNet wouldn’t have succeeded, and similarly, NVIDIA’s embrace of the AI era was accelerated by AlexNet’s success.

ChatGPT, another monumental AI innovation, owes its existence to thousands of NVIDIA A100 chips powering its training, highlighting that every major AI firm today scrambles to acquire NVIDIA GPUs

This situation can be traced back to the inherent advantages of GPUs over CPUs in AI processing.

Although both are types of chips, GPUs are highly specialized for handling massive parallel processing tasks, such as those required in AI and graphics rendering, outperforming CPUs in specific scenarios.

As CPUs are built for general multi-task processing, they manage complex logic and therefore possess fewer coresIn contrast, GPUs evolved primarily for rendering images, allowing thousands of simpler cores to tackle tasks simultaneously—like thousands of elementary school students racing to solve basic math problems far faster than a few professors tackling advanced calculus.

NVIDIA Thriving Amid AI Boom

AI relies significantly on data processing, and while AlphaGo won through a sophisticated statistical analysis of board configurations rather than leveraging specialized knowledge of the game itself, it falls within the domain of GPU capabilities—especially when dealing with extensive datasets.

The creation of CUDA was crucial, for without it, GPUs remained mere gaming cards unable to unlock their full potential

CUDA democratized access to GPU computing for engineers, scientists, artists, and game developers, facilitating swift resolutions to complex computational problems.

As AI expert Andrew Ng remarked, the number of individuals globally who could program GPUs before CUDA was negligiblePost-CUDA, leveraging GPUs became remarkably straightforward, enabling NVIDIA to establish a robust ecosystemWithin its boundaries, users’ productivity surged, while branching out threatened to stifle performance.

Competing initiatives later emerged, like AMD's RCM platformHowever, lagging behind CUDA by a whole decade, NVIDIA solidified its dominant stance.

For instance, when small enterprises began investing in cloud computing, NVIDIA's Tesla V100 cards—despite conveying vehicular aspirations—were pivotal for data centers

As autonomous driving gained traction around 2015, the company launched its DRIVE series tailored for this emerging market, further aligning its offerings with AI growth.

The cryptocurrency mining boom in 2017 served as another key inflection point for NVIDIA, with the company increasing its offering of mining-specific cards, aptly taking advantage of this unexpected surge.

Currently, NVIDIA's extensive profile spans from AI and gaming to aerospace and bio-science simulation, creating an interface through which computation-related fields are inevitably intertwined with its ecosystem.

It's difficult to label NVIDIA's ascent strictly as a flight of fortune; instead, the company emerges as an essential utility in the digital ageHuang's prophetic decision to invest in CUDA back in 2006 reshaped the company’s identity, transitioning it from a gaming hardware provider to a formidable AI processing leader

Was this shift driven more by foresight than mere luck?

Gaming Graphics Cards

In 2020, Google Brain researcher Sara Hooker articulated that NVIDIA's rise mirrored a lottery win, emphasizing the alignment between advancements in hardware and modeling.

Revisiting Huang’s disclosure to Forbes in 2016 showcases how he always believed in the potential for graphics chips to extend beyond gaming, yet the swift pivot to deep learning caught him by surpriseIn this competitive landscape, the fortunes of the participants can fluctuate entirely based on timing and readiness, making it clear that luck can hold significant influence.

The question remains: could AI have developed in China? Can Huang’s success be replicated in a different environment? While no definitive answers exist, history tells us only a bold technical milestone can invigorate a market, often passing through the eye of a storm filled with countless failures.

Pragmatic entrepreneurs typically do not gamble solely on fortune

Those striving for transformational technology must dare to dream beyond immediate demands or profits, ultimately shaping the future by embracing innovation even with significant uncertainty.

As of November 17, 2023, NVIDIA's 4090 graphics cards will no longer be exported to mainland ChinaThis follows the U.SDepartment of Commerce's restrictions laid down in July 2022 concerning the export of the A100 and H100 AI chips to China.

Presently, vendors in China grapple with insufficient computational powerThis situation resonates with Huawei’s long-standing awareness of the technological challenges they faceThe inability of China to produce high-end semiconductors in-house adds complexities not only in the communication industry or 5G technology but also—and perhaps more critically—in the realm of AI.

In the grand scheme, NVIDIA could become the energy source of the digital era, yet the call for China to develop its NVIDIA is undeniable.