Rogé Karma: “Donald Trump launched his political career by insisting that free-trade deals had sacrificed the national interest in the pursuit of corporate profits. One wonders what that version of Trump would make of his most recently announced trade policy.
“On Monday, he declared on Truth Social that the United States would lift restrictions on selling highly advanced semiconductors to China. In doing so, the president has effectively chosen to cede the upper hand in developing a technology that could determine the outcome of the military and economic contest between the U.S. and its biggest geopolitical rival.
“The U.S. is currently ahead in the AI race, and it owes that fact to one thing: its monopoly on advanced computer chips. Several experts told me that Chinese companies are even with or slightly ahead of their American counterparts when it comes to crucial AI inputs, including engineering talent, training data, and energy supply. But training a cutting-edge AI model requires an unfathomable number of calculations at incredible speed, a feat that only a few highly specialized chips can handle. Only one company, the U.S.-based Nvidia, is capable of producing them at scale.
“This gives the U.S. not only an economic advantage over China, but a military one. Already, AI systems have revolutionized how armies gather intelligence on enemies, detect troop movements, coordinate drone strikes, conduct cyberattacks, and choose targets; they are currently being used to develop the next generation of autonomous weapons. ‘Over the next decade, basically everything the military and intelligence communities do is going to some extent be enabled by AI,’ Gregory Allen, who worked on the Department of Defense’s AI strategy from 2019 to 2022, told me. This is why, in October 2022, the Biden administration decided to cut off the sale of the most advanced semiconductors to China. The aim of the policy, according to the head of the agency in charge of implementing it, was ‘to protect our national security and prevent sensitive technologies with military applications from being acquired by the People’s Republic of China’s military, intelligence, and security services.’
“The policy seems to have done its job. Chinese AI firms tend to explicitly cite export controls as one of the biggest obstacles to their growth. DeepSeek, the Chinese company that earlier this year introduced an AI model nearly as good as those made by the leading American firms, is the exception that proves the rule. At first, DeepSeek’s progress was taken as evidence that restricting China’s access to advanced chips was a failed project. However, the company turned out to have trained its model on thousands of second-tier Nvidia chips that it had acquired via a loophole that wasn’t closed until late 2023. DeepSeek’s AI model would have been even better if the company had had access to more and better Nvidia chips. ‘Money has never been the problem for us,’ Liang Wenfeng, one of DeepSeek’s founders, told a Chinese media outlet last year. ‘Bans on shipments of advanced chips are the problem.’”
the „rip american tech dominance“ framing assumes china even wants to be dependent on american chips long term, which… they don’t lol
the whole point of their strategy is indigenization. after watching what happened with huawei and the constant threat of export controls shifting with each administration, why would any chinese strategist build critical infrastructure around components that are tied to the whims of the US? that’s just handing the US a kill switch.
plus if you have abundant cheap power (which china is building aggressively), horizontal scaling with less efficient chips can match vertical scaling with bleeding edge ones.
quantity has a quality of its own.
DaySecure7642 on
The EU companies sold their IPs for short term profits and market access to China. Look where they are now, permanently behind. I hope the US doesn’t make the same mistake.
ABlackEngineer on
Definitely a leveled headed and prudent article title and not at all a desperate grab for attention for a dying medium.
This doesn’t even jive with the CCP’s recent moves to curb H200 for domestic Ai growth.
Tennis-Affectionate on
Top 7 US Tech Companies Value: $21.5 Trillion
China Total GDP: $19.8 Trillion
AI Chip Global Market Share
USA: 98%
China: 2%
Cloud Computing Global Market Share
USA: 65%
China: 5%
Total Data Centers (Facilities)
USA: ~5,300
China: ~450
Private AI Investment
USA: $67 Billion
China: $8 Billion
Major AI Model Releases
USA: 42
China: 12
Good luck
Leave A Reply
Du musst angemeldet sein, um einen Kommentar abzugeben.
5 Kommentare
Rogé Karma: “Donald Trump launched his political career by insisting that free-trade deals had sacrificed the national interest in the pursuit of corporate profits. One wonders what that version of Trump would make of his most recently announced trade policy.
“On Monday, he declared on Truth Social that the United States would lift restrictions on selling highly advanced semiconductors to China. In doing so, the president has effectively chosen to cede the upper hand in developing a technology that could determine the outcome of the military and economic contest between the U.S. and its biggest geopolitical rival.
“The U.S. is currently ahead in the AI race, and it owes that fact to one thing: its monopoly on advanced computer chips. Several experts told me that Chinese companies are even with or slightly ahead of their American counterparts when it comes to crucial AI inputs, including engineering talent, training data, and energy supply. But training a cutting-edge AI model requires an unfathomable number of calculations at incredible speed, a feat that only a few highly specialized chips can handle. Only one company, the U.S.-based Nvidia, is capable of producing them at scale.
“This gives the U.S. not only an economic advantage over China, but a military one. Already, AI systems have revolutionized how armies gather intelligence on enemies, detect troop movements, coordinate drone strikes, conduct cyberattacks, and choose targets; they are currently being used to develop the next generation of autonomous weapons. ‘Over the next decade, basically everything the military and intelligence communities do is going to some extent be enabled by AI,’ Gregory Allen, who worked on the Department of Defense’s AI strategy from 2019 to 2022, told me. This is why, in October 2022, the Biden administration decided to cut off the sale of the most advanced semiconductors to China. The aim of the policy, according to the head of the agency in charge of implementing it, was ‘to protect our national security and prevent sensitive technologies with military applications from being acquired by the People’s Republic of China’s military, intelligence, and security services.’
“The policy seems to have done its job. Chinese AI firms tend to explicitly cite export controls as one of the biggest obstacles to their growth. DeepSeek, the Chinese company that earlier this year introduced an AI model nearly as good as those made by the leading American firms, is the exception that proves the rule. At first, DeepSeek’s progress was taken as evidence that restricting China’s access to advanced chips was a failed project. However, the company turned out to have trained its model on thousands of second-tier Nvidia chips that it had acquired via a loophole that wasn’t closed until late 2023. DeepSeek’s AI model would have been even better if the company had had access to more and better Nvidia chips. ‘Money has never been the problem for us,’ Liang Wenfeng, one of DeepSeek’s founders, told a Chinese media outlet last year. ‘Bans on shipments of advanced chips are the problem.’”
Read more: [https://theatln.tc/cNbcXRpD](https://theatln.tc/cNbcXRpD)
the „rip american tech dominance“ framing assumes china even wants to be dependent on american chips long term, which… they don’t lol
the whole point of their strategy is indigenization. after watching what happened with huawei and the constant threat of export controls shifting with each administration, why would any chinese strategist build critical infrastructure around components that are tied to the whims of the US? that’s just handing the US a kill switch.
plus if you have abundant cheap power (which china is building aggressively), horizontal scaling with less efficient chips can match vertical scaling with bleeding edge ones.
quantity has a quality of its own.
The EU companies sold their IPs for short term profits and market access to China. Look where they are now, permanently behind. I hope the US doesn’t make the same mistake.
Definitely a leveled headed and prudent article title and not at all a desperate grab for attention for a dying medium.
This doesn’t even jive with the CCP’s recent moves to curb H200 for domestic Ai growth.
Top 7 US Tech Companies Value: $21.5 Trillion
China Total GDP: $19.8 Trillion
AI Chip Global Market Share
USA: 98%
China: 2%
Cloud Computing Global Market Share
USA: 65%
China: 5%
Total Data Centers (Facilities)
USA: ~5,300
China: ~450
Private AI Investment
USA: $67 Billion
China: $8 Billion
Major AI Model Releases
USA: 42
China: 12
Good luck