Many analysts — both on the sell-side and the buy-side — are focusing on DeepSeek’s cost competitiveness and superior performance over other LLMs, particularly in coding, math and logical deduction.
Within a month, China’s major AI companies, such as Alibaba, Tencent, ByteDance, Huawei, and Baidu, have embraced DeepSeek’s AI models in their chatbots and AI systems. Many local governments and public sector institutions, as well as companies, have also done the same. The Guangzhou, Nanjing, Suzhou, Zhengzhou, and Hohhot city governments said their computing networks are connected to DeepSeek-R1, while the Futian district government in Shenzhen said it created 70 AI “staff members” with DeepSeek to handle documents and help its staff.
Within 26 days of R1’s release, the number of Chinese daily active users surpassed 40 million. In the US, more than 30 million downloads took place in Apple’s App Store and Google’s Play Store alone in less than two weeks. Already, some analysts have opined that the capital invested in developing most LLMs in China and elsewhere could well have gone to waste. It might also have put at risk the six little AI tigers in China.
We think the DeepSeek Sputnik moment goes far beyond that. It will reinvigorate the nation in two transcendental, transformative ways.
See also: Xi signals China may finally move to end deflationary price wars
First, inspiration — DeepSeek’s shaking the AI industry to its very foundations will inspire a whole new generation of youthful, energetic, hard-tech entrepreneurs and science geniuses to take the plunge into “0 to 1” innovation, unprecedented since Deng Xiaoping’s reopening and reform in 1978, rather than “1 to n”.”
Next, spillover — We think DeepSeek’s model will spill over into other industries, ranging from biotechnology, GPU and CPU chips, intelligent and autonomous driving, humanoid robots, factory automation, cybersecurity, and the military, to aerospace engineering, to mention a few.
Global conventional wisdom pigeonholes Chinese entrepreneurs and engineers as excellent in “1 to n” rather than “0 to 1” innovation, somewhat like Japan in the 1970s and 1980s, which is not totally incorrect, while the latter remains the preserve of the West. Silicon Valley's Peter Thiel bifurcates innovation as such:
See also: What’s worth remembering?
“When we think about the future, we hope for a future of progress. That progress can take one of two forms. Horizontal or extensive progress means copying things that work — going from 1 to n. Horizontal progress is easy to imagine because we already know what it looks like. Vertical or intensive progress means doing new things — going from 0 to 1. Vertical progress is harder to imagine because it requires doing something nobody else has ever done. If you take one typewriter and build 100, you have made horizontal progress. If you have a typewriter and build a word processor, you have made vertical progress.”
We went out on a limb in our December 2024 APS China Monthly (ACM) essay to call for a decade-long bull run in the stock prices of China’s high-tech companies helmed by well-educated, intelligent scientists and engineers full of irrepressible entrepreneurial drive. These companies are not to be confused with e-commerce firms thin on both technology as well as investments, tangible assets, and R&D spending.
These high-tech companies and the wide application of LLMs will reboot the economy in a way never seen before, with effects that will cascade to most sectors.
Liang Wenfeng, the 40-year-old hedge fund billionaire founder of start-up DeepSeek with a degree in AI from Zhejiang University’s Department of Electrical Engineering, has answered some key questions for us.
Confidence
On the point about innovation and confidence, Liang himself says it best: “For many years, Chinese companies are used to others doing technological innovation, while we focused on application monetisation — but this isn’t inevitable…China should gradually become a contributor instead of freeriding.
In the past 30 plus years of the IT wave, we basically didn’t participate in real technological innovation. We’re used to Moore’s Law falling out of the sky, lying at home waiting 18 months for better hardware and software to emerge… But in fact, this is something that has been created through the tireless efforts of generations of Western-led tech communities… The cost of innovation is definitely not low, and past tendencies toward indiscriminate borrowing were also related to China’s previous conditions. But now you see, whether it’s China’s economic scale or the profits of giants like ByteDance and Tencent — none of it is low by global standards. What we lack in innovation is definitely not capital, but a lack of confidence and knowledge of how to organise high-density talent for effective innovation.”
For more stories about where money flows, click here for Capital Section
Liang says that DeepSeek has no wizards. Its team consists mostly of fresh graduates from top universities, PhD candidates in their fourth or fifth year, and some young people who graduated just a few years ago. No one returns to China from overseas; they are all local.
Just 18 months ago, Liang failed to fundraise from some Chinese tech giants and venture funds, who judged his start-up as a zero that will end as a zero. Most Chinese tech giants have focused on the quick monetising of horizontal “1 to n” applications, eschewing arduous, risky “0 to 1” innovation. So Liang had to go it alone, using his personal wealth earned from his hedge fund business to fund the relatively modest outlay of hiring 150-odd unknown young guns for their passion, a “posse sans egos”.
To the surprise of the big guns in the global AI landscape, DeepSeek’s “out-of-the-box” state-of-the-art approach to LLMs has shaken the AI world. This is what a meld of brilliance, passion, and hunger looks like. Most critically, Liang’s trailblazing team will inspire a generation of “0 to 1” innovators.
Innovation and industrial advancement
We have believed for some time that China will successfully transform into a technologically advanced economy by 2035, a somewhat unorthodox, incredulous view of the ongoing economic slowdown and a bear market. The central pillar for this effort is the semiconductor industry, which is itself critical to all the other pillars like telecommunications, infotech, drones, electric vehicles (EVs), robotics, solar energy, biotech, AI, quantum computing, smartphones, low altitude aviation, deep space technology, and many more fields.
China has poured tremendous amounts of engineering talent and capital into this area in recent years to rapidly close the gap with America and its satellites as part of its plan for GDP per capita to reach that of mid-level developed countries by 2035.
The five key factors behind our belief in this timeline are:
First, an abundance and critical mass of engineering talent. China produces about five million graduates a year, almost four times the US, UK, France, Germany, and Japan combined. Today, it has about 50 million of them are aged 35 or younger! As the saying goes, quantity has a quality all of its own;
Second, access to abundant capital and savings from a nation of 1.4 billion with the world’s second-largest GDP — more than US$3 trillion ($3.9 trillion) of foreign reserves and US$19 trillion of household savings;
Next, a large domestic market;
Also, a whole-of-government approach and leadership continuity that is not subject to electoral cycles as well as pork barrel politics. The Chinese government has steered the nation and committed its financial and human resources to this paramount goal;
Last but not least, China has a strong entrepreneurial spirit. The 2024 Shanghai exhibition, which I attended, attracted over 1,000 semiconductor companies. There must be tens of thousands of companies in China’s semiconductor ecosystem today. If there is money to be made, China will not lack risk-taking entrepreneurs and engineers.
While it has achieved remarkable results in EVs, drones, telecommunications, solar panels, robotics, quantum computing, automation, and bioscience in recent years, it still lags in aircraft, semiconductors, medical devices, and AI, though the DeepSeek watershed moment represents a quantum leap in China’s bid to close the AI gap.
According to the Australian Strategic Policy Institute’s Critical Technology Tracker, China’s global lead extends to 37 out of 44 technologies that ASPI is now tracking, covering a range of crucial technology fields spanning defence, space, robotics, energy, the environment, biotechnology, AI, advanced materials and key quantum technology areas. The chart above from ASPI is just for the area of AI, computing and communications. China’s “quiet success” continues in many other such charts in their report. With or without DeepSeek, these successes must contribute to China’s future in one way or another.
We now think that China will complete its transformation into a technologically advanced economy by 2030, five years ahead of our previous estimate, as the trickle of “0 to 1” innovation that started with DeepSeek 20 months ago swells into a river that will rapidly enrich a fertile river delta in the coming months and years.
Semiconductors are an area in which China has poured enormous resources. It produced 7-10 DUV machines in 2024 and will likely grow that to 12-15 units in 2025 and 25-30 in 2026, according to a Western-trained founder of a successful Chinese semicon company. I learnt recently that China is making significant progress on its first EUV lithography equipment prototype — likely the mother of all breakthroughs due in three to five years. This is probably the last frontier where China is still trailing. If that feat is accomplished, undoubtedly, China would enter a golden age of technological excellence as well as economic success and prosperity that should last for a decade or more.
The forgotten fundamentals
In October 2024’s ACM essay The Forgotten Fundamentals, I wrote about how China has built an immensely productive and efficient manufacturing powerhouse that produces one-third of the world’s manufactured products. I added: “I do not see this changing in the next 20 or even 50 years, despite calls by the West to decouple. Can India, Vietnam or Japan be the next manufacturing complex of the world? I do not think so.
"Building a manufacturing powerhouse is not just about building factories, but also constructing infrastructure such as airports, seaports, power plants, and roads, building an entire supply chain — including materials, components, logistics, nurturing and training millions of high-quality, industrious and resourceful engineers and technicians, a penchant for meticulousness, and so on. This colossal asset, which China has painstakingly and exactingly built, will serve the country well in the decades ahead.”
China’s bull decade
In December’s ACM essay China’s Bull Decade?, I wrote about how the China sceptics among global and domestic investors may have overlooked China’s human assets, including the large pool of STEM graduates. I emphatically said, “Chinese workers still share the same work ethos, mindset and mentality that propel China’s economic miracle. They are still hardworking; they still want to make more money. The new generation is better educated, resourceful, disciplined, and more skilled. Which countries’ students frequently win the International Mathematical Olympiad competitions?
"Future entrepreneurs would make even better businessmen because they will be more educated and have a global outlook; some will have wealthy parents and, therefore, have access to capital from day one, and they also have a massive domestic market. It is this powerful combination of its powerhouse of a manufacturing ecosystem as well as its human talent and enterprise, which cannot be easily duplicated elsewhere, that will contribute to China’s continued (technological advancement and) economic success and prosperity.”
The animated blockbuster Ne Zha 2 was made with AI-enabled rendering with China’s own supercomputing support / CMC Pictures
The DeepSeek effect
The DeepSeek-R1 success will not be solitary, in my view. China’s animated blockbuster Ne Zha 2 boasts ticket sales of over US$1.7 billion and counting, just barely three weeks after its Jan 29 debut, becoming the highest-grossing animated film of all time worldwide, surpassing previous record holder Inside Out 2 from the US and is now also the eighth highest-grossing among movies of all types globally.
Ne Zha 2’s huge success is helping introduce characters from Chinese mythology to new audiences around the world. This “cultural innovation” is itself powered by AI-enabled rendering support from Guizhou’s Gui’an Supercomputing Center in Southwest China, which offers 1,000 high-performance graphics processors and over 700 servers with a combined computing power of 15 quadrillion calculations per second.
While China is indeed facing some daunting headwinds like its massive, moribund real estate sector, which was well-known and hence largely discounted in our view, its other fundamental strengths were not trumpeted and hence not discounted. This was a primary reason for our bullish view last year.
In the post-Jan 20 fortnight, I understand that Nvidia AI chips were changing hands for 40% less in China’s secondary market. Crucially, DeepSeek’s innovation and significantly lower consumption of computing power mean that Huawei’s Ascend 910B and 910C AI chips and CPU chips will do brisk business in China and perhaps even abroad, as they are now cheap, viable, and good enough to run the models, develop applications and do referencing tasks, that were previously the domain of Nvidia chips.
Nvidia’s CUDA software framework is also now not necessary. The upshot is that Ascend 910B and 910C chips — which found almost no customers as recent as six weeks ago, despite costing a fraction of Nvidia’s chips — are now seeing explosive demand. SMIC is most likely the foundry making most of the Ascend and advanced-node CPU chips.
Importantly, as Huawei’s chips are increasingly adopted by more companies, in larger quantities, and across a wider range of industries, Huawei’s engineers will benefit from a larger pool of data and practical experience. This additional information and real-world usage will, in turn, stimulate further advancements, such as better chip designs and software architecture.
In technology, widespread use creates a positive feedback loop: more usage leads to more insights, which then leads to faster and more significant advancements. This will be valuable in areas like Huawei’s CANN (Compute Architecture for Neural Networks), which is designed to accelerate AI computations on Huawei’s Ascend chips, serving a similar role to CUDA. In short, the gap between Huawei’s and Nvidia’s GPU chips will narrow from here.
The companies in this tech-driven bull market now have access to an affordable ecosystem for developing bespoke AI applications.
The bull run in tech stocks has led to a rerating, which propelled China’s stock market to gain US$1.7 trillion in market cap during the post-DeepSeek fortnight. Is the bull run over? We think not; we believe it has just begun.
Already, the Big Four companies Alibaba, Tencent, Baidu and ByteDance will increase their AI investments to US$52 billion this year, a 127% increase. This will not be enough to meet the demand for computing power as many local governments, public institutions, and private sector companies have announced in recent weeks that they will integrate the DeepSeek-R1 model in their chatbots, IT systems, and websites.
As the Chinese are experienced, adept and skilful in monetising the application of technologies, we can expect massive adoption of AI in almost every manufacturing, service, and government sector in the coming decade, which is China’s forte. Let a thousand flowers bloom. In the process, the productivity in these sectors should keep on increasing, thereby making everything more efficient. The total economic value that will be created will be humongous. Shenzhen-based Huawei said its Huawei Cloud has connected with DeepSeek-R1.
Chinese automakers BYD and Geely also said their electric vehicles integrate with DeepSeek-R1, while Tencent’s Weixin messaging app users can now use DeepSeek for searches, with Tencent exploring the integration of multiple products with DeepSeek. Coupled with the emergence of many more innovative companies, we believe our prediction last year that China has entered a multi-year bull market is now reinforced, and it is unfolding before our eyes.
We do not think it is far-fetched to assert that China’s tech stocks, using golf parlance, are probably at the third tee box of a golf game, whilst the US tech stocks — which were thought to be the only serious players in town and had seen a massive rerating — could be on the 18th tee box.
If this view reflects the new reality, where should investors place their bets in 2025? Finally, DeepSeek has probably torpedoed the once-popular pair trades of “Long India, Short China,” “Long Japan, Short China,” or even “Long US, Short China.”
Wong Kok Hoi is the founder and co-CIO of APS Asset Management
