I am not so much into these things to offer a technical perspective on issues I am highly ignorant of. However, I can discuss from the economics, geopolitics and business aspects of the unfolding development.
First thing that crossed my mind was how easy it is to misjudge issues especially when one is backed by the arrogant assumption of having all the information.
I still remember what happened middle of this month, when the outgoing Biden administration added the Chinese tech start-up Zhipu to its trade-restrictions blacklist as part of efforts to slow down China’s race to develop an alternative to OpenAI’s ChatGPT. Zhipu’s major backer is Tencent Holdings and Alibaba.
America’s export prohibitions on advanced AI chips from Nvidia to China affected Zhipu’s access to the most advanced AI development hardware. But they didn’t know that what would hit them is coming from a “small ant.”
Most cases, being small means being underrated, being invisible, thus giving you freedom to be you. And that was what happened. Ochu nwa Okuko new ada….
Liang Wenfeng, DeepSeek’s founder ran into murky waters in 2022 when investors lost huge sums through his Zhejiang High-Flyer Asset Management, which used artificial intelligence to pick stocks. His AI faltered.
It was a disappointment that became a blessing for him. His assets shrunk from more than $12-billion to about $4 billion, forcing him to start thinking as if behind the scenes Liang was laying the groundwork for a new AI start-up, DeepSeek.
DeepSeek is just 20 months old. The founder, Liang was born in 1985. Let that sink. Wiping out close to $1 trillion from giants such as Nvidia and co is more that disruption. It is a challenge.
It could be recalled that on 6th of January, Nvidia unveiled its most advanced Blackwell chips which can process trillion-parameter large language models at up to 25 times lower cost and energy consumption than its predecessor. It shares soared and closed at a record high of $149.43.
Same shares tumbled 17% to $118.58 on investor concerns that DeepSeek had matched rivals such as OpenAI using far fewer Nvidia chips than U.S. firms. The yet to be answered question by the investment community is whether AI requires hundreds of billions of dollars in capital expenditure to come up with the latest innovations and vanguard AI models — and whether export controls can hold off Chinese competition.
If DeepSeek achieved this feat with lower capability Nvidia’s A100 chips which the US government considers not advanced enough for export restrictions, would it not push other companies to move away from its premium highly advanced chips that are responsible for its soaring wings?
To save face and salvage the situation, Nvidia said today that DeepSeek’s advances show the usefulness of its chips for the Chinese market and that more of its chips will be needed in the future to meet demand for DeepSeek’s services.
Another interesting aspect which I think had the greatest impact was that he achieved this inspite of the constraints from access to the latest semiconductor technologies and limited resources. This is a pointer of how possible it is for underdogs with the right talent and motivation to hit above their grades.
In July last year, Liang was quoted as saying that “OpenAI is not a god and cannot always be at the forefront.” He said that the problem with China was lack of confidence, and how to organize high-density talents to achieve effective innovation. Liang recruited engineering talent almost exclusively from China. Many were fresh out of top universities, interns in their final leg of doctoral studies and Olympiad medal holders.
To the geopolitics of it all. Why did DeepSeek release its R1 advanced AI reasoning model on 20 January, the same day Donald Trump was sworn in as America’s 47th president. Is there any cryptic message they used that date to convey? President Trump said yesterday that DeepSeek is a wakeup call.
DeepSeek has made China proud. Yesterday, Liang was invited to a closed-door business symposium in Beijing hosted by Chinese Premier Li Qiang. The event had China’s leading experts in technology, science, education and other fields and they discussed on drafting a government work report on AI.
If you ask DeepSeek what the outcome of that meeting was, be sure that it would inform you that that is beyond its pay grade. As a proper Chinese raised child, DeepSeek seem to avoid topics that the CCCP may frown at.
As I wrote somewhere earlier today, technology reflects the biases of its inventors.
Contrary to claims that some wanted to sabotage their platform yesterday, I believe what happened was unpreparedness to handle the huge surge from global traffic. That is why concerns over its infrastructure have been raised.
If DeepSeek achieved this with Nvidia’s 10 000 A100 Nvidia chips, there is the possibility that they’d need close to 50 000 of Nvidia’s far-more powerful chips, the H100s to remain effective which they are not going to get.
What this development teaches us all, and most especially those of you who are techies is a reiteration of what computer scientist Kai-Fu Lee said earlier this year that “The US is great at research and innovation and especially breakthrough, but China is better at engineering,” but “in this day and age, when you have limited compute power and money, you learn how to build things very efficiently.”
Analysts believe that inspite of this huge feat by DeepSeek, the US still has the ace because Mag 7 and US tech is focused on AGI endgame with infrastcuture that China and DeepSeek cannot come close to.
But in this time, and age, you can never say never, especially with minds like Liang who from available evidence is not focused on making money. Remember Ceaser’s admonition that “Cassius has a lean and hungry look; he thinks too much, such men are dangerous.” Liang Wenfeng shares similar thin and sharp appearance.
