A hedge fund owner founded DeepSeek, showcasing the potential of capitalism

DeepSeek has called into question the capital-intensive model of developing AI.

Is any government likely to incentivise entrepreneurs to set up hedge funds or use artificial intelligence in algo trading? To ask the question is to answer it. So-called speculative activities — practiced by the likes of quant hedge funds — are frowned upon almost universally by governments, regardless of their ideological flavour.

Instead, the sectors likely to be favoured are futuristic, such as green hydrogen, semiconductors, solar panels, electric vehicles, batteries, and so on. These industries, for instance, are sought to be promoted under India’s Production Linked Incentive (PLI) scheme.

Ideology can make a difference when it comes to the choice of favoured sectors.

In Donald Trump’s America, in contrast, solar and wind industries are disfavoured, while fossil fuels are back in fashion.

Chinese leader Xi Jinping wants investments in strategic industries such as AI, quantum computing, and the advanced parts of semiconductor manufacturing, rather than consumer internet companies. Hedge funds and other high-flying parts of finance are frowned upon, with Chinese authorities clamping down on high salaries.

Yet, the deeper lesson of the recent, history-changing days — the unveiling in late January of DeepSeek R1, a large-language model from the eponymous Chinese artificial intelligence company whose feats have shocked its American competitors — is surely that these policies are, at the very least, questionable.

What is on display is the transformative and unpredictable nature of capitalism, its unmatched ability for creativity and innovation, which can come from the most unexpected quarters, even those not favoured by the ruling dispensation. DeepSeek has called into question the capital-intensive model of developing AI, exemplified by Stargate, a $500 bn investment in new AI infrastructure announced by President Trump just after his inauguration and a few days before DeepSeek LLM was unveiled.

Decentralised and diversified economies, where sources of capital and entrepreneurship are dispersed, are likely to generate innovations, and hence productivity and growth.

It also illustrates a basic economic principle — excess profits, of the kind Nvidia is making, will attract competition in a well-functioning market economy. It can also be read as a vindication, to some extent, of one of the insights of the controversial Austrian School of economics, which is that individuals, not states or classes, have agency.

In India, we can perhaps stop fretting about the fact that most startups are in the consumer internet space. Perhaps capitalism will work its magic, and Zomato or Zerodha will come up with an LLM.

DeepSeek, as everyone knows by now, is the offspring of a quant hedge fund called High-Flyer, the sort of name which might attract scrutiny from India’s financial regulators.

Liang Wenfeng, the founder of the fund, had hired many PhDs in mathematics for his core stock trading business, which relied on deep learning. This proved to be a strength when he forayed into the world of AI and large language models. Further, he had been prescient in his accumulation of older AI Nvidia GPUs to circumvent US technology restrictions.

It turns out that seemingly unrelated fields — capital markets and the development of LLMs — can become related. In this case, the interlinkage developed because High-Flyer’s trading strategy was powered by AI, which in turn encouraged the owner to embark upon an unrelated diversification: the development of LLM.

DeepSeek, founded in mid-2023 to foray into artificial general intelligence (AGI), appears to have been almost a hobby, though going forward the AI business may well eclipse the stock-trading one.

High-Flyer, which has $7 bn of assets under management, has been hit by the crackdown on high-flying finance types by the Chinese government, donating in mid-2022 to the cause of Common Prosperity, a signature Xi project.

In 2021, it even apologised for losses. Liang Wenfeng is now a national hero.

Details of the DeepSeek origin story — particularly its claims of spending just $6 million to train its LLM model — have been disputed, notably by Sam Altman, the founder of OpenAI. What is jaw-dropping is that, 75 years after the foundation of the Chinese Communist Party, a hedge fund has incubated a company in frontier technology areas which is threatening to upend the leadership of American companies. It’s also ironic that a company originating from what is officially a communist country should have illustrated the workings of Schumpeter’s principle of creative destruction.