Wang Huai of Linear: God has only opened two doors leading to the future of AI, one in the United States and the other in China.

More than a decade ago, as one of the most influential early Chinese engineers at Facebook, Harry Wang returned to China with the romantic spirit of a tech enthusiast and founded Linear Capital. Since day one, it has been continuously focused on early-stage technology investments.

When he dissected the Silicon Valley methodology in the best-selling book Building Facebook, AI was still a sci-fi term that most people regarded as something only "crazy people" would talk about. But today, whether you believe in AI or not, it truly concerns the survival of every individual and race in the future.

Looking back at 2025, Linear Capital consistently maintained a stable investment frequency in the AI field, investing over $100 million.

Under the non-linear acceleration of AI, how should we rethink the logic of investment? How should we view the technological competition between China and the United States? When thinking can be calculated and judgment can be predicted, where should the value of human beings retreat to? Below are Harry's answers:

The Sino-US AI War with 100% and 50% "Chinese Content"

We started talking about AI ten years ago. At that time, if you said that AI would become the monopolistic core power in the future, people would think you were crazy because there was no real way to prove it either technically or commercially.

But today, whether you believe in AI or not, it truly concerns the survival of our future individuals and races. The speed of change in this world always exceeds the expectations of the most optimistic people.

I have an engineering background and studied machine learning during my Ph.D. However, the manifestation of this new wave of AI capabilities, from the launch of ChatGPT 3.5 until now, has constantly refreshed my expectations of its upper limit. What's truly astonishing is not always the speed itself, but the acceleration.

I believe that in the face of this superpower of AI, God has only opened two doors on earth, one in the United States and the other in China.

The first door is in the United States. I must admit that Americans have a pioneering spirit in this area. Sometimes it can be quite wild, but they can truly achieve something from scratch. They have some out-of-the-box ideas, daring to think what no one else dares to think and do what no one else dares to attempt. Sometimes this kind of attitude may make people uncomfortable, but you can't deny its effectiveness.

The second door is in China. The Chinese have their unique advantages - our ability to scale from 1 to 100 is unparalleled in the world. The Chinese are patient and have strong visions. More importantly, we have a perseverance to "do things to the extreme." Historically, the Chinese have had a great pioneering spirit, but after reaching a certain level, we were content with the "Greater China" and didn't want to expand further - of course, this was a good thing for other regions at that time.

But now, the times have changed.

In my opinion, these two doors are the doors to future survival. How should one choose between these two doors? As an individual, you may only be able to choose one door to go deep into; but as an organization or an institution like ours, you must understand the rules behind both doors.

For this reason, we did something within Linear Capital: We require all our team members to subscribe to and use the best AI models from both China and the United States, and the company will fully reimburse the expenses. This amount of money may not be much and may not be considered a great benefit, but it is a survival training in the AI era.

By letting the team use ChatGPT, Gemini, Claude, as well as DeepSeek, Kimi, etc. at the same time, you will find an interesting phenomenon: The US models do have an advantage in data openness and creative divergence, but they often fall into a certain level of "self-censorship"; the Chinese models are actually more open in aspects such as DEI. So, depending on what you want to do, if you want a relatively complete perspective, you should refer to the best models from both China and the United States.

In the AI era, the model you use determines your perspective of seeing the world. A single perspective is fatal.

Speaking from a business perspective, we haven't invested in basic large models. The main reason is that as an early-stage fund, a lot of the money invested in this area may just be wasted. In the end, although it may bring some good reputation, the returns may not be that significant. The best basic large models in China have taken a more cost-effective and inclusive AI route. I'm prepared to be wrong about this view. If I really am proven wrong, I'll actually be glad.

In the process of the Sino-US AI competition, a long-neglected variable is emerging. I refer to it as CBV - "Chinese by volume" by analogy with the alcohol content in drinks. I've done some data research, which may not be very accurate. Among the top AI researchers in the United States, the proportion of Chinese is close to 50%. At OpenAI, the proportion of Chinese among AI researchers is also quite high. This means that at the talent level, the current AI competition is essentially a competition between the United States with 50% "Chinese content" and China with 100% "Chinese content." When I was working as an engineer at Facebook more than a decade ago, there was a similar situation, but it was 25% vs 100% back then.

The underlying logical basis for this is the actual distribution of talent and engineering capabilities. Behind AI lies mathematics and engineering, which involve boring matrix operations and parameter tuning. The Chinese have unique advantages in terms of patience, discipline, and scale in these two aspects.

I have a half-joking statement: The competition of AI models is a competition between the Chinese in the United States and the Chinese in China. But the key difference here is that the top Chinese talents in the United States are very strong in innovation and breakthroughs, while the local talents in China may have more advantages in engineering, scaling, and continuous optimization. These are two different capabilities, both of which will be crucial in the future.

Re-examining the Value of Human Beings

Let me talk about my view of the future. It doesn't make sense to talk too far into the future, so I'll just talk about our lifetimes. I have a judgment that may sound harsh but that I firmly believe in: In the next 1 - 3 decades, AI will definitely achieve its dominant position; by 2030, 30% of the current human jobs will be replaced by AI; by 2040, another 40% will be replaced; by 2050, the current human work system will be basically reconstructed, and there won't be much left for humans to do.

Many people think this timeline is too radical. But to be honest, I may even be being conservative here. Why? Because the concept of acceleration is very important here, as well as a paradigm shift or leap. It's very difficult for us to understand such changes with linear logic - although our institution is called Linear Capital, we've always encouraged non-linear thinking to understand the changes in the world.

Let me give a common example: Mobile phones were commercially launched in the 1980s, and their penetration rate was still less than 1% in the early 1990s. But by 2000, the penetration rate began to soar. And now, except for some remote areas in the world, the penetration rate of mobile phones has exceeded 100%. The penetration curve of AI will only be steeper - ten times, or even a hundred times steeper.

Of course, what's most disturbing is not the speed of replacement, but the logic behind it. Many people think their jobs are full of creativity and unique, but in fact, they may not be, or at least most of them aren't. But the cruel truth may be that your "eureka moment" may have occurred many times in the long history of humanity.

The power of AI lies in its ability to learn from similar "eureka moments" recorded by someone somewhere in the fields of mathematics, physics, chemistry, etc. 100 or 200 years ago, and finally incorporate these "eureka moments" into your content.

When creativity becomes a "commodity" that can be mass-produced, where is our value anchor? I'm not promoting pessimism. On the contrary, I think this gives us an opportunity to rethink the "value of human beings." If repetitive creative work can be replaced, then what is the truly unique and irreplaceable value of human beings?

Of course, as we all know, current large models have a problem - they tend to "talk nonsense seriously." This is the original sin of probabilistic models - the choice of the next token is often based on extremely small probability differences. The best answer and the second-best or third-best answers may only differ by 0.0001% in probability.

But I think that to a large extent, AI is already "taking our jobs." This imperfection reveals a more cruel truth: AI doesn't need to be a superman and get 100 points in every subject to replace you. It only needs to exceed the expectations for your position in the comprehensive dimension.

Many people are still waiting for the ultimate form of AGI (Artificial General Intelligence) or ASI (Artificial Superintelligence) and don't believe that day will come soon. I still think that most people's jobs will be replaced. So, within our organization, we've proposed a concept, which is "Everything can be AI-enabled." This is not just a slogan but our daily code of conduct. That is when you encounter a new problem, first mindlessly think about "Can AI help us with this? Can I complete it with the help of AI?" This has completely changed our workflow.

This may sound simple, but it's extremely difficult to implement in practice. For most of us, all our experiences, intuitions, and workflows formed in the past two or three decades are based on the inertial logic of the "Pre-AI era." Whether you can enter the next era depends on whether you can break these inertia with your own hands and put AI at the premise of all your thinking.

I have a large amount of reading needs every day. Now, I use Notebook LM to help me digest books and generate podcast manuscripts. I walk two or three kilometers to work in the morning and back home in the evening, and this time is perfect for listening to these AI-generated podcasts. Artificial intelligence may not necessarily make me understand problems more deeply, but it can greatly improve my efficiency. Then I can decide which content is worth in-depth reading. Now, if you read a new piece of content without the help of AI, it's not that you can't read it well, but it will take you much longer to reach the same level of understanding as others. In this era where speed decides everything, the efficiency gap is the gap between life and death.

This has even brought about a new type of "social screening": Previously, we may have focused on a person's values and professional background. Now, we may also need to look at how they use AI and how they view AI. We've put forward a rather straightforward view, which is "Those who don't embrace AI will die." - This "death" doesn't mean physical death, but the demise of your competitiveness and the value of your existence.

In my opinion, no matter how smart people were in previous generations, if they can't learn and recognize the concepts of "AI-lize everything" and "AI first," they will be eliminated. Many well-known and successful entrepreneurs and investors may have no problem maintaining their current lifestyles. But if they want to continue to be successful in the new era, they must become "AI-first" people and cooperate with more younger generations to understand the ways of doing things and thinking of these young people.

The end of an old era and the rise of a new era is a long process, not an overnight change. If you can't embrace these new changes, you're not losing to the machines, but to the people who can use the machines.

Beer foam and soap bubbles are different kinds of bubbles

Many people ask me: Is the current boom in China's AI field and the hot investment and financing a bubble?

I answered this question seven or eight years ago. My answer is: There are two kinds of bubbles. One is like a soap bubble. Once it bursts, it's gone, leaving only a mist and nothing else. The other is like beer foam. Beer with foam tastes much better than beer without foam. Under the foam is the real fragrant beer.

In the computing power battlefield, NVIDIA is still the undisputed king. Jensen Huang is not only a great engineer but also a great salesman - he knows what to say and do at the right time. In my opinion, NVIDIA's biggest pressure is the rise of China in terms of computing power, but this will take time.

Maybe there is a bubble in AI now, but we still believe that in many fields, it may be like beer foam. For example, in the field of robot embodiment, the "ChatGPT 3.5" era has not arrived yet. Once it does, I think this may be the biggest revolution since the Industrial Revolution, and the anxiety in the United States will reach a new level. So, although we've always thought there is a bubble, we've still invested a lot of money without hesitation.

For example, the company White Rhino that we've invested in. Although we've invested for several years, this year is the first time they've truly entered a stage of practical significance in terms of scaling up. Their business growth rate has exceeded 20 times that of the same period last year. They've taken the lead in completing the mass production preparation of vehicle-grade products for driverless logistics vehicles and have also raised more than $100 million this year.

Another example is Tash Robotics, which has attracted attention in terms of financing this year. They raised $120 million in the first round and $122 million in the second round. We participated in both rounds and are their third-largest investor. Their initial application is in the wire harness scenario, which requires a large amount of labor and has been difficult to be replaced by machines in the past. The other day, they just demonstrated the world's first robot capable of autonomous embroidery, becoming the world's first company to solve the wire harness manufacturing problem in the field of robotics.

In technology investment, you must respect the current situation and also be able to help them think about the future. By the time things are truly implemented, their applications will show a huge J-shaped curve. So, in the early stage, it requires investors to maintain a stable mindset. No matter how turbulent the short-term market is, in the long run, this direction is the inevitable path for the digitalization of the physical world.

When the capabilities of the ChatGPT 3.5 level are injected into robots, it will be the biggest change since the Industrial Revolution - because it will no longer only process digital signals but directly manipulate atoms.

What will the future factories look like? I can describe a scenario: There will be all kinds of robots in the factory. They have the same basic architecture, but the end-effectors can be quickly replaced. They can interact with each other. This interaction may be in a language we understand or may not need to be. As long as they are given a clear goal, they can operate autonomously. Which role each machine plays mainly depends on reconfiguring the perception module. These robots may be producing mobile phones today and may switch to producing drones tomorrow - All it takes is to send a new model from the cloud and adjust the end-effectors, and the whole conversion process may only take a few minutes.

I think in China, there are only two places where we can achieve such a vision more efficiently and at a lower cost: the Yangtze River Delta and the Pearl River Delta. Behind this is the spillover effect of the electric vehicle supply chain in these regions of China. The future of robots is essentially the second explosion of China's electric vehicle industry chain. These supply chains highly overlap with those of the automotive industry, especially the electric vehicle industry. In a sense, the future of robots is an extension of the EV supply chain.

In recent years, China's rise in the consumer electronics field is also obvious to all. We are the earliest investors in two of this year's "Seven Rising Stars in Hangzhou." We invested in Rokid ten years ago. They've always been fully committed to the smart hardware field. Recently, they broke the global crowdfunding record for the smart glasses category on Kickstarter, and the first-round pre - sale has been sold out and delivered.

For products like DJI, Insta360, and AfterShokz, their global performance has its reasons and history. In the past, China has produced or done OEM work for too many electronic devices for the world, which has created a strong supply chain capability in this area. In the process, it has also cultivated many talents who not only know about technology, products, and supply chains but also more and more enterprises know how to do distribution and marketing.

This evolution of capabilities is giving rise to a new group of entrepreneurs. They not only understand technology but also understand human nature. So, we believe that in the future, there