Article by: Dao Ge

A large number of articles about AI are emerging almost every day on the internet. Most of these articles, in my opinion, have very low reference value except for attracting attention, driving traffic, and deliberately stirring anxiety. Therefore, I usually just quickly skim through these types of articles.

However, the article I am recommending today, while inevitably inducing a bit of anxiety, closely reflects the situations I and my friends have personally experienced in several aspects. The overall judgments and viewpoints presented are relatively objective and forward-looking, so today I would like to share some points from the article "Everyone is Token-Maxxing, an Arms Race No One Dares to Stop" (for details, see the reference link after the article) that I find very thought-provoking.

- “I was sitting in the audience at the YC W26 batch Demo Day, and when the fifth company presented, I decided to stop taking notes... I realized that what I had noted down might be outdated by next month... Among this batch of over a hundred companies, what they are doing is actually highly concentrated: about 80% are vertical agents, such as helping lawyers organize documents, helping customer service distribute tickets, helping HR sift resumes... These vertical agents, before forming any business barriers, can be replicated by an ordinary engineer, even by myself, over a weekend; they have already lost their investment value... Silicon Valley itself is starting to struggle to keep up with its own pace.”

In my previous articles, I mentioned that for the masses to enjoy the dividends of the AI era, there are three methods: starting a business in the AI field, seeking employment at AI companies, and investing in the AI sector.

Regarding starting a business in the AI field, I am well aware that I am not skilled in this area, so I write very little about it, but I am very supportive of those who have the courage to venture into it.

However, after reading this description, I feel that starting a business in AI (at least at this stage) is likely much more difficult than I originally imagined.

YC is a leading incubation organization in Silicon Valley, and even their incubated projects are constantly facing the risk of being disrupted by large model companies; thus, it is even harder for other projects to emerge successfully.

“Silicon Valley itself is starting to struggle to keep up with its own pace” — when I read this sentence, I felt that if I had not personally experienced that kind of brutal pressure, it would be very hard to articulate such a statement.

What does this indicate?

This indicates that starting a business in the AI era, at least at this stage, carries exceptional risks, so I am now more inclined to be conservative and cautious regarding this option.

- “This sense of anxiety is spreading among the researcher group. Researchers are the top-tier talents; it does not refer broadly to 'research personnel,' but to the group responsible for model training and algorithm innovation in large model companies (such as OpenAI, Anthropic, DeepMind)... And now, even the work of researchers themselves is being automated. This is what the students at DeepMind are doing — using models to train models, which is also the highly discussed AI self-evolution in Silicon Valley this year. This year, engineers are being eliminated, and by the end of the year, researchers will also begin to be replaced.”

I have long been accustomed to hearing about ordinary engineers being replaced and massive layoffs.

But researchers are academic professionals. Generally speaking, many researchers in large companies can match the level of university professors or associate professors. If even these talents are facing unemployment and replacement risks by the end of this year, it is hard for me to imagine what other professions that were previously technology-gated in AI companies could be safe from replacement.

Here, the professions I mentioned that were previously technology-gated refer to many mid-level and even senior technical personnel. If their positions are not secure, what can ordinary people pursue in AI businesses?

Of course, I still believe that top-tier talents are hard to replace, but those positions do not belong to the masses, nor do they belong to general senior talents.

There is also a broader argument that AI will create more new positions while replacing human jobs; I believe this statement, but in the short term, I have not yet seen what new positions are created that can accommodate ordinary people on a large scale. Even if such positions emerged in the future, it may be that for our generation, once we pass a certain age threshold, we might not have the opportunity to participate.

As of now, the only option that I can still see somewhat clearly among the three options mentioned above is investment.

- “The entire Meta is using Claude Code. This is not a startup, not an experimental team, but a company valued at a trillion dollars... Moreover, the token consumption is likely not linearly increasing... Once the reliability of agents goes up another level, token consumption will not just be a 50% increase each year, but could skyrocket overnight. There is a consensus among friends about a prediction that by the end of this year, many companies (including tech giants) only actually need 20% of their workforce... The future may be a scenario where 10 people do the work of 100 people, earning for 20, while 90 people are left unemployed.”

This statement highlights the widespread use of Claude at Meta. In fact, it's not just Meta; almost all major companies in Silicon Valley (except those specifically prohibiting the use of Claude, like Google) are deeply embedding Claude in their operations.

Given the current situation, the large model companies in the U.S. have basically formed a monopoly pattern. I believe that in the end, no more than three will emerge as winners, likely OpenAI, Anthropic, and Google; even XAI is likely to be eliminated.

Among these three, I have the most faith in Anthropic.

Whichever of these three ends up dominating the market, their future market value could reach or even exceed that of Nvidia.

Additionally, what struck me significantly in this statement was the mention of token usage. If this means that once the stability of AI agents is improved, then token consumption could grow non-linearly, or exponentially?

If token usage can increase at an exponential rate, then all industries surrounding tokens, especially those with strict constraints, will still have substantial growth potential in the future, such as graphics cards, electricity, data centers...

Nvidia is particularly interesting in this context.

- “This seemingly distributed world of innovation is, at its core, extremely centralized. This center is Nvidia... I originally thought the scarcity of cards had eased over the past year. But during this visit, I found that scarcity has returned, and it's even more outrageous than last time... The power structure behind it is very clear: those who have cards are powerful, and who has the cards is determined by Nvidia.”

This passage emphasizes Nvidia's monopolistic position.

For a long time, I have been observing whether Nvidia's monopoly position is solid.

Although Google and AMD are also fiercely competing, I intuitively feel that Google seems to lack the genetic makeup for hardware development; its TPUs may find it hard to capture the market on a large scale like Nvidia. And what about AMD? Historical cases of the second-largest player successfully displacing the largest are extremely rare.

So if AI development and technological routes continue as they are now, Nvidia's monopoly position might not be easily shaken.

However, if a new AI development path or an entirely new technological route emerges that requires a complete overhaul of computing chip design, then at that time, Nvidia's position may potentially be threatened.

But no one knows if or when this situation will occur.

Thus, Nvidia's monopolistic advantage might be greater than one imagines.

While Nvidia is a good target, its price point is quite awkward.

I remember that recently when Nvidia was around $180, a reader commented that it was not appropriate to buy Nvidia at that price, and I would not buy it at that level either.

But not long after, its stock price actually exceeded $210.

I still do not dare to buy.

Compared to Nvidia, I would now prefer to buy Anthropic.

- “Silicon Valley is experiencing a deep-seated crisis of security. During this trip to Silicon Valley, I repeatedly heard friends seriously discussing the same topic: buying Bitcoin, building bunkers, and installing bulletproof glass at home; they were not joking.”

This statement piqued my curiosity about their logic for buying Bitcoin.

From this description, I guess they are considering this from the perspective of safety and extreme situations, so in the physical world, they choose to insulate themselves with bunkers and bulletproof glass, while in the virtual world, they choose Bitcoin as their insurance.

Reference link:

https://mp.weixin.qq.com/s/kwErGjX231e2efVWhERzTw

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。