NVIDIA founder Jensen Huang is a familiar face who regularly appears at the annual Computex in Taipei. For more than a decade, his trademark black leather jacket remained the same, while his hair turned from black to gray.

This “uncle next door,” familiar to Taiwanese journalists, has become an unexpected superstar in his sixties.

Nvidia stock quadrupled in one year following the launch of ChatGPT, exceeding a market value of over two trillion USD to make it the world’s third-largest company.

Over 10,000 attendees flooded into the recently-concluded Nvidia GTC annual conference in Silicon Valley, surpassing the fervor of Apple's events. Huang, the longest-tenured CEO in the worldwide tech industry, has been dubbed the "new Steve Jobs'' by American media.

The magical journey of the “New Steve Jobs”

To understand this journey, we can begin with a keynote address delivered by Kevin Zhang, senior vice president of business development at Taiwan Semiconductor (TSMC), at March’s International Solid-State Circuits Conference (ISSCC).

In a report by semiconductor analyst Zhang, it was predicted that by 2030, high-performance computing server semiconductors will surpass terminal devices such as smartphones and laptops. These server semiconductors are expected to reach a 40 percent global market share, exceeding the 30 percent share currently held by terminal devices. 

“A few years ago this was unimaginable, when the majority of silicon chips were consumed by end devices,” observed Zhang. However, “thanks to the explosion of AI, the situation has changed dramatically.”

Since ChatGPT emerged on the scene in late 2022, the scope and rapidity of the explosive hardware demand fueled by generative AI over the past year is unlike anything this generation of tech journalists has ever seen. Even more unbelievable is that the buying frenzy has been centered around Nvidia alone.

This has led to the following startling statistic: the annual revenue of Nvidia’s data center for 2022 was just about the same as that of the entire MediaTek company, around US$12.5 billion.

As a result, in the latest quarter ending in January, the revenue of this division exceeded US$15.6 billion, representing 409-percent year on year growth. 

For a company of this scale, such a growth rate is unprecedented. Even in 2021, when Tesla ramped up production of its affordable Model 3 electric vehicle, rewriting automotive history, its growth was a mere 74 percent.

Is this speed of growth sustainable for its scale?

Cloud companies still buying up Nvidia products

Doubts are rising as many worry whether Microsoft and Amazon have already filled demand for the coming years. This reflects the  boom and bust cycles common to the semiconductor industry.

Following the panic buying that took place during the pandemic, it took two years for the industry to digest its inventory. This serves as a cautionary tale.

However, at least for the time being, there is a healthy appetite for buying.

The big three players in cloud services, Amazon, Microsoft, and Google, are still significantly expanding their cloud infrastructure - Microsoft's capital expenditure for the latest fiscal year reached a staggering $46 billion USD, surpassing even TSMC. Moreover, each of them has expressed plans for upping the ante with further investment in the coming years.

This will be strategically spent money - buying up Nvidia GPU worth millions of NT dollars.

To this end, Amazon has even extended the depreciation period of its servers from four to six years, so as to allow for freeing up more funds.

At the Nvidia conference, a proud Jensen Huang declared that CPUs are passe, no longer requiring regular updates or upgrades. And that the prevailing industry trend “has already shifted from general computing (represented by CPU) to accelerated computing” (represented by GPU).

Is this about seizing Intel’s market share? Posed this question by shareholders, Jensen Huang uttered his golden phrase with a big smile, “Nvidia is a market maker, not share taker.”

Nvidia’s profit margins reached 76.3 percent in the latest quarter, not only unprecedented profits for the semiconductor industry, but exceeding even Microsoft’s 68.4 percent.

While it is said that the reward for innovators is that winners take all, in the eyes of its clients Nvidia seems to be taking too much!

Edge Starts Strategizing Counter Strike

At the CommonWealth Economic Forum early this year, distinguished computer scientist with Google DeepMind, Ed H. Chi, said that all the top players in the cloud are looking to apply AI computing power to the “edge,” in reference to the AI phones and AI PCs that have had the industry abuzz of late, This would thereby eliminate the need for NVIDIA GPUs, "because computing in the cloud is too expensive."

Explaining the logic behind this, he said “At first, any new technological development is about what’s possible. Then, when the new technology’s potential has been thoroughly exploited and enters maturity, the next question is ‘what’s the lowest cost?’”

Has generative AI entered the cost-effectiveness pursuit stage? At least in some areas, not yet.

The next, highly anticipated wave in generative AI, covers multimodal foundational models encompassing visual, linguistic, and audio features. This will take more computing power than right now, requiring a more powerful GPU.

One Nvidia manager told me, AI innovation is burgeoning, “like the pioneers moving west, we still cannot see the boundaries.”

From the looks of it “Cloud King” Jensen Huang will be able to keep smiling for quite some time to come.

Yet the cloud camp is strategizing its counterattack.

According to a Bloomberg report, Apple has already joined forces with Google to integrate Google’s large-scale language model, Gemini, into the iPhone. If all goes to plan, the iPhone 16 introduced at the end of 2024 will become the most powerful AI smartphone of all time.

---

Have you read?

• A look inside NVIDIA and Jensen Huang’s ambition
• Behind AMAX, one of NVIDIA’s earliest partners
• Why is Foxconn building "AI factories" with Nvidia

Translated by David Toman
Uploaded by Ian Huang