Blog Yixin Lin

The fundamental underpricing of the Chinese economy

TLDR: The Chinese economy is big, but the average Chinese person is poor, and this will not change soon even as China catches up to the US technologically. This may mean that the Chinese economy is fundamentally underpriced.


Three basic facts

  1. The Chinese economy, in aggregate, is large. China had a nominal GDP just shy of 19 trillion USD in 2024, second in the world compared to the US at just below 29 trillion; adjusted for PPP, China already overtook the US in 2014.

  2. Chinese living conditions have drastically improved. Compare those present-day numbers to the year 1978, when reform and opening up began: the GDP was around 0.2 trillion USD and the GDP per capita was around $200.

  3. However, the Chinese people are still poor. At a GDP per capita of less than 14 thousand USD, the economic activity of the average Chinese person is more comparable to a Mexican or an Argentinian than an American.

To understand modern China, it’s important to hold these three contrasting facts in mind simultaneously: China is massively strong, and the average Chinese has experienced drastically improved living conditions in their lifetime, yet that same person still has much less wealth than an American.

From low value to high technology

There are a few advantages to not being wealthy. China’s role as the world’s factory, churning out cheap copycat goods, is deeply embedded in the American psyche, and having a massive, lowly-paid labor force willing was the critical ingredient to get there.

There are also certain traits that scale well with total GDP: examples include a massive domestic market, enormous sway as most countries’ largest trading partner, and, perhaps most importantly, the capital to invest: in infrastructure, military, and technology.

These obvious facts nonetheless combine to a very interesting situation: due to the massive domestic market and large absolute investments in critical technology, China’s economy is transitioning from low-value manufacturing towards dominant positions in a variety of critical technologies, while the population remains poor.

Today, Chinese companies dominate industries both in the traditional stronghold of physical goods and hardware (5G telecommunications, TVs, fast fashion, solar, electric vehicles, batteries, drones, and robots), but also increasingly beyond (mobile payments, short-form video, drug discovery). Areas of weakness with strategic importance have shown signs of rapid catchup; examples include the fiercely competitive smartphone market, but of course particularly salient in recent days are semiconductors and AI. Here, Huawei, SMIC, and especially DeepSeek have recently shaken the world (with peers like Xiaomi, Alibaba, and Moonshot not far behind).

It may feel like a sudden phase transition, but the road from low-value manufacturing to state-of-the-art technology across all of these domains is actually quite natural:

  1. First, the domestic industry catches up to modern business practices, often through partnerships with foreign multinational corporations eager to capitalize on cheap labor.
  2. Then, once the companies have the basic knowledge to produce these products, the know-how diffuses across the domestic industry.
  3. The resulting know-how spawns purely-domestic companies which begin fiercely competing, especially on price but also on product features and technology, improving upon the foreign companies’ products along at least one of those dimensions.
  4. The winners of these brutal domestic competitions become successful globally.

It’s a story that’s reminiscent with the tales of other Asian tigers like Taiwan, South Korea, and (to a lesser extent) Japan.

China is different

What’s different with Japan, South Korea, and Taiwan is that in those countries, reaching international technological competitiveness eventually led to high wages and broad prosperity. The cheap-labor-to-high-technology arbitrage trade was successfully executed, and now labor is now no longer cheap. Toyota, Samsung, and TSMC are world-leading giants in their industries, but their products are not inconceivably cheap compared to the competition, owing at least in part to comparable-to-American labor costs.

Of course, Chinese wages are not immune to the same effect, and as noted in the second basic fact, wages have gone up historically; if left to their natural course, China’s situation may converge to a similar equilibrium in the long run. However, two factors throw a wrench in the standard progression: (1) China is absolutely massive, and making a country of 1.4 billion people rich takes a tad longer than one with sixty times less people, (2) the Chinese government is willing – indeed, openly choosing – to tolerate economic pain to drive towards long-term political goals.

An authoritarian government is capable of pushing their population much further away from their natural inclinations than a democratic one. It’s clear through actions like the kneecapping of the Didi and Ant Financial IPOs, the embarrassingly public disciplining of the gaming, education, and hedge fund industries, and the meager stimulus packages that prosperity is no longer the primary goal. The culture of intense work, built up from the pressure of poverty and intense growth, have no escape valve of prosperity anytime soon: the pressure will simply keep increasing.

A different league

There’s a conceit in the science fiction series Dune which could be roughly summarized as the famous meme’s first line, “hard times create strong men”: one army is considered the best because they’re trained in a literal hellscape of volcanic ash, and another army that ends up beating them was only able to do so because they’re raised in even worse conditions, in the eponymous “dune”.

While this example was fictional, it rhymes with reality. Examples from history include the horse nomads of the Eurasian steppe, raised in an inhospital environment which forced the development of near-superhuman horse-archery skill, repeatedly rampaged across the supercontinent; the most successful of them, the Mongols, were the largest land empire in history. Another is the fragmented and ultra-competitive Europe in the 15th through 20th centuries, subject to geopolitical pressures that bred the technological and military superiority which eventually conquered the world. The same pattern can be seen in its most natural form when an invasive species is introduced to a pristine ecology and rips through a formerly unoccupied niche, devastating the native fauna.

In each of these examples, a population is forced by its environment to engage in brutal competition to a level unmatched anywhere else. When they leave the crucible, however, their adaptations yield an insurmountable advantage over those under less crushing pressure. It’s akin to an professional sportsperson, used to competing against the world’s best, dropping down to the minor leagues for a game.

I argue that, for many industries, the Chinese domestic market is simply a different league of competition. Here’s someone who knows something about high-tech manufacturing, talking about competing with Chinese companies:

Electric car sales in China are gigantic … the Chinese car companies are extremely competitive… By far Tesla’s competition is in China. There’s a lot of people out there who think that the top 10 car companies are going to be Tesla followed by nine Chinese car companies. I think they might not be wrong… China is super good at manufacturing, and the work ethic is incredible. If we consider different leagues of competitiveness at Tesla, we consider the Chinese league to be the most competitive. –Elon Musk at the NYT DealBook Summit, November 29, 2023

Less than a year later:

Chinese EV maker BYD’s quarterly sales overtook Tesla’s for the first time. October 31, 2024

The contradiction

The meme’s next sentences are, of course, “strong men create good times; good times create weak men.” The Mongols, having conquered the known world, started enjoying the fruits of civilization and stopped sleeping in the saddle. Faced with the choice between continuing to drink horse’s blood for sustenance in the freezing steppe versus dining in a newly-conquered palace in silk robes, who’d fault them?

A century ago, a worker in America might have accepted a thousand dollars a month to work six days a week, but today, that’s so small a sum that it’s literally illegal. An engineer in San Francisco very rarely chooses to work 996, and certainly not for the compensation of one in Shenzhen. To be clear, higher wages and better work hours in a more technologically advanced society is not new. What is new is that that engineer in Shenzhen getting paid less may actually be outputting higher quality work with more advanced technology, and the Chinese company is putting out a product that may be better, not just on price, but overall.

If this is the case, then standard economic theory dictates that the Chinese company should raise prices and increase profit, leading to higher wages and increased prosperity – but here we diverge from the natural course, because Beijing has signaled quite clearly that business will be subservient to politics. The continued artificial devaluation of the yuan, the unabashed crackdown on business leaders, and the paltry domestic stimulus point to a world where economic scarcity will force Chinese workers to work incredibly cheaply and with unmatched intensity for the foreseeable future, constantly in fear of another person willing to do it for even cheaper to step in.

We are now in a situation where the Chinese economy and technological prowess rivals that of the US, yet the Chinese domestic competition will remain brutally competitive and cost-conscious. The natural progression to higher profit is short-circuited by dozens of domestic competitors willing to undercut any apparent market leader and by direct and indirect government policies, and the ecosystem’s brutality rewards accelerated iteration cycles which rapidly advances technical know-how unavailable anywhere else.

The underpriced economy

I therefore claim that the Chinese economy is underpriced and underestimated.

Of course, the reflexive response to any claim of underpricing should be, “why ain’cha rich?” If there’s truly a market inefficiency, shouldn’t one be able to exploit it?

Perhaps the trade doesn’t exist. After all, there are many reasons to avoid investing in China. The government freely puts its fingers on the stock market, so the entry point is not even clear. (Compare a Chinese stock index with its GDP growth over the past few decades, and it becomes apparent that Chinese equities are far less reflective of economic strength than US equities.) Foreign investment terms are onerous and property rights are enforced selectively, and the courts are greased with red envelopes stuffed with cash. The past few years of government intervention in the economy, not to mention capricious raiding of foreign company offices, has rightfully spooked foreign capital, leading to a flood of outflows. The projected demographics are clearly catastrophic, and the soaring growth of a billion youthful workers is already aging into an albatross around the Chinese economy’s neck.

Perhaps more fundamental is that the prices may in fact be correct: fierce domestic competition could imply a lack of durable profits, and the standard razor-thin margins implies that companies go broke constantly. Even if this analysis is right, the discounted future cash flows may be low!

On the other hand, it seems obvious to me that China has reached technological supremacy across several key industries, and a rapid-enough rate of innovation to put it within spitting distance of many more. Perhaps in reality, China’s position is closer to that of a would-be mononpolist willing to play the long game, undercutting competitors to build market share and technical understanding until the market is cornered. It may be simply that running a direct financial model on “possible control of powerful new technology” is too difficult – at least, until it’s too late.

Conclusion

In China, hard times are creating strong men, but strong men are not creating good times. In fact, the times will be hard for the foreseeable future, and (if the meme is anything to go by) the spigot of strong men won’t turn off anytime soon.

I would not want to work in China, but I would not underestimate the Chinese nation. There will be many more DeepSeeks in the future.

You have been warned.

Thoughts on DeepSeek

Summary

DeepSeek, a small but talented lab spun out of a Chinese hedge fund, managed to approximately replicate OpenAI’s o1 (publicly released September 12) and put it out for free as an open-weight model r1, which is very close to the absolute frontier of model capabilities and earlier than relevant Western competitors Anthropic, Meta, and x.AI. (The current publicly-known frontier of model capability is OpenAI o3, announced Dec 20; Google DeepMind on December 20 announced Gemini 2.0 Flash Thinking, of which the January 21 version achieves comparable performance to o1).

Implications

  1. It’s free to use, cheap to run, and extremely performant; this severely undercuts closed-source model serving businesses (OpenAI, Anthropic; to a lesser extent GDM, x.AI).
  2. More importantly for the future training compute landscape (i.e. NVIDIA stock price), DeepSeek managed to train the model with much fewer compute resources than anyone expected.
    1. A small clarification: the number everyone is bandying about is that it cost $5.5M to train – this is purely an estimate of the cost of doing a single base model training run by renting the GPUs on the open market, and the actual cost to DeepSeek is significantly higher (multiple training run attempts, salaries, fixed costs of setting up a cluster, the actual reinforcement learning training step to produce r1). Still, the total cost is at least an order of magnitude lower than a comparable Western lab.
    2. The technical approach (reinforcement learning on chain-of-thought) is very similar to what is publicly known about o1. Though it requires a good pre-trained model (the costliest step in terms of compute), it can quickly increase model capabilities with much less compute. It’s also simpler than expected (they explain what other, more complex methods they tried), and show that training small models on the outputs of r1 gives them strong capabilities (“model distillation”), further reducing inference cost.
  3. What this implies about China’s position:
    1. DeepSeek was forced to be inventive and highly efficient because of significant chip export controls; for example, they were able to extract more compute than expected from the China-specific chips that NVIDIA were allowed to sell them.
    2. However, that does not mean chip export controls have failed: they have significantly impeded progress by DeepSeek and other Chinese AI labs, and the DeepSeek CEO explicitly highlighted access to advanced chips to be the primary bottleneck over e.g. additional funding.
    3. DeepSeek’s employees are primarily (extremely strong) Chinese students educated within China, not those who did PhDs outside of China, highlighting the depth of Chinese domestic technical talent.
    4. There’s a lot of speculation on whether DeepSeek somehow “cheated”, e.g. (1) by secretly having access to large amounts of banned chips provided to them by the Chinese government as most publicly suggested by Alexandr Wang, (2) by conducting espionage within Western labs like OpenAI to steal the secrets, (3) by training on the outputs of o1, which is against the OpenAI terms of use. These reflect a sentiment of absolute shock that such a strong model could be made with so few resources. My personal opinion is that, (A) given DeepSeek was not one of the main national AI champions (e.g. 01, Alibaba, Bytedance, Huawei; in fact, the Chinese government was not happy with DeepSeek’s parent hedge fund pursuing quantitative trading), (B) the r1 results have been replicated immediately on a small scale by grad students, (C) o1 chain-of-thought is hidden and therefore distillation could have only helped the base v3 model rather than lead to the innovations in r1 (the model that actually surprised the markets), it is unlikely they are the front of a CCP psyop up to now and this is purely cope when confronted with an unusually efficient and innovative company. (Of course, the DeepSeek CEO was seen this week meeting with senior government officials, which means this almost certainly will change going forward.)