"Iron" for trillion

Why the cost of nvidia took off to heaven and what will follow

At the end of May, a significant event occurred in the world industry in the production industry. For the first time, the company of this industry entered the Trillionniks Club - a business giant with a market value of more than $ 1 trillion - which already includes Apple, Microsoft, Saudi Aramco, Alphabet (Google) and Amazon. In the context of the beginnings of artificial intelligence, NVIDIA claims to be the main producer of iron for AI.

In May, NVIDIA shares have risen by 29%, and over the past six months - by 138%. And, it seems, this is only the beginning of an unprecedented take -off of the company.

“It seems that a new gold fever is approaching us, and Kirk and shovel can only be bought from NVIDIA,” CHISTOPHEHANNA Financial Group, Christopher Rolland, wrote to customers, increasing the target price of NVIDIA shares from 350 to 450 dollars. “We believe that this is only the beginning of the wave of generative AI, and so far it seems that NVIDIA has captured almost the entire economy,” Blein Curtis predicts Barclays, raising the target price of the company's shares from 275 to 500 dollars.

Analyst Rosenblatt Securities Hans Moshesmann in a note for his customers and completely raised the target price from 320 to 600 dollars. What happened?

It all started with the appearance of the so -called large (basic) models of learning on big data in the fall of 2021. In simple words, these are machine learning models adapted to applications that study on huge arrays of unprocessed data. For successful training, the productivity of “iron” and software (Software) (Software) is extremely important. Breakthrough results in this area launched the AI ​​revolution, often called the ChatGPT revolution (by the name of the most famous dialogue AIBTGPT). In turn, the AI ​​revolution led to explosive demand for "iron" and to the unprecedented rise of shares of the main manufacturer of this "iron" NVIDIA .

The development of “deep training of large models on big data” turned computing power into a key factor in progress of almost all intellectually capacious industries: from the development of new drugs to new types of weapons. And where AI is involved, computing power generally decides everything. The formula of superiority is extremely simple:

• collect as much data as possible;

• Create as complex as possible (according to the number of parameters) model;

• Teach the model as quickly as possible.

The one who succeeds “more - more - faster” has the highest chances to win in the technological race. And here everything rests on the computing power of the “iron” (HW) and the quality of algorithms (SW).

With all due respect to algorithms, in this pair, their role is secondary.

For to invent an algorithm, copy it or even stealing it is still easier than making an “iron”: it either is, or it is not - we took place in the USSR.

Can NVIDIA really be considered a monopolist on kirs and shovels of the Golden Factory AI? A short answer - yes. Now Nvidia owns approximately 95% of the market of powerful graphic processors for deep learning servers (GPU). At the same time, they are constantly not enough: Openai alone is already consuming almost 20% of the world's high -performance GPU. And almost all of them are NVIDIA production.

How long can this monopoly last? According to experts, there are no signs that the lack of graphic processors that is observed today will decrease in the near future. The NVIDIA separation from competitors is colossal: 20-100 times. And if, as experts suggest, the demand for servers with GPU for teaching large models will increase by three orders of magnitude by 2030, NVIDIA has something to answer. The company has just announced its new supercomputer platform for AI called DGX GH200.

So the main issue of NVIDIA success is how long the “gold fever” of AI can last in general and where it can lead. Since the beginning of the ChatGPT revolution, the development of II technologies has so accelerated that any predictions in this area have become impossible. The only thing that can be confidently fixed so far is the further growth in the development rate. Now, for the decline of the “gold fever”, the AI ​​is not visible the slightest reason. But the further development of the industry in the next five to ten years depends on whose hands AI will concentrate. The possibilities of large models depend on the financial capabilities of their owners: after all, it is necessary to spend money on their training, and this is costly.

Most likely, only a few largest corporations and the authorities of the most technologically advanced states will be capable of this.

In any case, the scale of the new breakthrough potential is such that in 5-10 years, the systems will learn to do what they have not dreamed about before. They can become a new information gadget for people, replacing all the previous ones: from desktopes and laptopes to smartphones and smart hours. Personal II assistants will be enormously surpassed Siri and Alexa and will become widespread, being connected to their “companion” 24x7. Education will be transformed by relying on AI-Tutors, and mass medicine-on AI-Personal health consultants.

And after 5 years, truly slaughter AIs will arise, which we can imagine today with great difficulty. For example, a new entertainment industry based on controlled dreams. The transition to the selection of partners is exclusively on the recommendations of AI. Total centaurization (synergistic synthesis of the musculs in people and AI) of intellectual and creative professions, all science and all types of arts.