China Just Broke America’s AI Monopoly | Warwick Powell & Carl Zha

追加: 2026-05-12

[00:00:00]Speaking of uh AI, you know, you know, there's just recently the US Congress has came out basically siding with um with American AI company by accusing Chinese AI companies like Deepseek of quote unquote stealing American AI the AI intellectual property by uh distillation. You know what they call distillation is know they will they accuse the Chinese AI company of posing questions to the American AI by getting by get by sort of data mining and then and then so they didn't have to spend the time and energy to train their own uh the data for data set for their own AI. Now this is a common accusation by American uh AI company level at Chinese but this is the first time I saw the American government has publicly take a stance uh combi in in combination with you know the the Treasury Department's uh sanctioning of of the the five Chinese refiners in relationship to Iranian oil purchase. Why do you think all these happened just before Trump visit? Do you think that's related? Is is this part of a you know just Trump's hard negotiation tactic?

[00:01:25]>> Well, Trump's style always is to create things that he can then negotiate around, right? So, you know, create a problem so that you can take it away and and claim that as a as as some kind of a of a decisive, you know, victory or an act of benevolence or what have you. Um the issue of of AI and IP theft of course um has been around uh forever and a day. The United States has got it in its head. And I mean by this, of course, the folk in inside the beltway in Washington, that Chinese people and Chinese enterprises have zero capacity to innovate. And therefore, anything that remotely appears to be technically innovative must be the result of theft.

[00:02:11]And this is so deeply ingrained in that sort of foundational layer of world views that uh it's an easy one to trot out to in a sense marshall the troops to um to show the American public that uh that the government and the politicians are standing up for you know what's good and right. Um but in the end it's largely politicking these days and you'll notice that the Chinese side largely shrugs its shoulders and just moves on. you know, uh, it it it no longer takes these accusations particularly seriously in large part because over the course of the last 10 to 20 years, there have been concerted efforts actually made by uh, you know, governments in China to um, improve this whole arena of intellectual property rights um, the protection of them to a point where now, you know, uh, Chinese researchers are the most significant beneficiaries of the global intellectual rights regime, intellectual property rights regime um than anyone else. You know, if you look at patents submitted um publications made, clearly China has become a significant beneficiary of this entire regime and so has an interest in the the integrity of the regime. But I think this AI question has some slightly broader ramifications. KL um firstly almost all of America's uh last 12 months or so of so-called economic growth GDP growth has been a result of uh capital expenditure from the AI industry mainly building out data centers and without that um you'll recall earlier this year there were some reports that uh that showed that without uh AI data centers uh American GDP would be um anemic it would be less than half a percent. So, uh, keeping the AI engine going is actually quite an important part of the of of the story about the American economy, particularly leading up to the midterms. It's also incredibly important because the AI industry is actually facing some serious headwinds.

[00:04:27]Not because people aren't clever, not because they're not able to um you know develop uh you know more powerful models, but they're facing headwinds in uh in I think a number of very specific ways. Firstly, there is a physical material constraints that is manifesting around the problems of electricity supply.

[00:04:50]AI is really just electricity, right?

[00:04:54]and um and without being able to resolve this question of electricity supply uh the ability to keep punching forward with models um becomes problematic. The electricity supply problems in the United States go to the shortages of firm capacity which is generating capacity estimated by the energy secretary to be somewhere in the order of 100 gawatt um over the next few years. Um, but not only do you need to generate electricity, you also need to be able to distribute electricity from where it gets generated to where it gets used. And that requires poles and wires.

[00:05:28]It requires substations, converters, transmitters, and those sorts of tools, you know, equipment. And in all of those areas, there are now supply constraints.

[00:05:38]There are also constraints in terms of available labor force, people with the right skills to do the highowered work, meaning the um the high voltage related work. and uh and so the ability to expand the American electricity supply industry is actually quite severely constrained and that's led to a number of scheduled data center projects to be postponed or to be delayed and some of them have actually been cancelled in certain locations in part because there's been public reactions and they've been forced to look elsewhere.

[00:06:12]the public reactions emerge because data centers actually also create inadvertent or unintended consequences which is to increase the cost of electricity um for other users and that of course means other industries and households. So that's the second part I guess of this data center or AI question. The third issue, Kyle, is that given all of these issues, the American AI industry and the business models that underpinned its valuations and the money capital that it raised over the course of the last two years, which it is spending of course with gay abandon um is now also under stress because of um the expanded roll out of high performance open-source solutions from China and that has fundamentally undermined the basic business model upon which American AI was originally premised. The idea that you could have um motored general models that you could charge lots of money for um was destroyed when Deep Seek came out with its models and subsequently of course um there have been a number of other Chinese developers who've done the same. So the business model is under incredible stress. research suggests that um adoption of American AI remains um well below the levels from a revenue point of view that would justify the valuations. And we also know that um Chinese models such as such as Quen and and Kimmy are finding their way in very very large market share terms into the developer space as well. So the American AI model is under stress and that matters because it feeds back to this issue of the share market valuations, right? Um and I think heading into the midterms, the last thing that the administration uh can afford to see given all of the other challenges that it's facing is the popping of the AI bubble. So it is going to be doing everything it can to boost market confidence that the American government has the back of the American AI industry that it's going to fight Chinese thievery you know from top to bottom uh that it will protect American AI uh that it will provide lucrative contracts through the Pentagon and other agencies to um to to substantiate the business models but ultimately also to ensure that there's sufficient liquidity within the economic system at large to keep pouring um uh you know trading capital into these markets. So these are these are I think where the real issues are. So you touched on this issue of of you know allegations of IP theft but it's actually a symptom of these other problems.

[00:09:02]>> Yeah. This is where I actually don't understand uh because the as you mentioned the the the AI buildout requires the buildout electric grid and they require equipment like a lot of the equipment will have to be imported from China yet Trump administration is putting on tariff on those those Chinese equipments raising the cost and and you like right now the the the cheapest way to generate electricity is also solar um with the advancement in the in the in the solar panel production again mostly coming from China yet Trump has made a conscious effort to not support the the solar roll out because he wants to concentrate on what he consider America's strength with gas and oil jewel baby jewel and and so you know chi so okay fine America don't need solar because America have plenty of gas. But now with the Iran war, the gas price has soared. Even though America do produce gas but because gas is a globally traded commodity the gas you know US people in the US who uh consume the gas including you know farmers who rely on fertilizers which is you know which is use natural gas to produce and uh the this AI company who is building data centers who are building their own gas turbine uh powered grids they all run into this increasing cost issue and and everything is self-inflicted by the Trump administration. I don't Is it just because of lack of expertise that that the Trump administration don't understand what it takes?

[00:10:54]Yeah, look, I I think that there are legitimate and serious questions now around this issue of competence and uh you know, it's one thing to strip the public service for you know, whether it's for budgetary reasons or what have you, but the downside risk of doing these wholesale um you know, attenuations of public sector um knowledge capabilities is that a whole bunch of people with institutional knowledge knowledge and experience um ultimately leave the system and it means that those making decisions ultimately are also u making decisions both on the fly and largely blind to um some of the complexities of the issues involved. The electricity system is a classic case in point because the American electricity system of course is actually a state-based system and it is a patchwork. Some states are more um renewable energy oriented than others.

[00:11:57]Interestingly, uh two of the most renewable energy and solar energy oriented are in a sense at opposite ends of the spectrum. So you've got Texas which is uh rolling out a lot of uh solar power but so did California, right? Um but this gas situation because as the systems became tight, some data centers turned to the idea of having dedicated gas turbines to um to fuel the um to to generate the electricity needed for the data centers. And as you say, as gas prices rise, it drives up the cost of um the uh the the data centers themselves. So they are very self-inflicted. I increasingly think that there is a competence question that um is not just a short-term problem but it's a a longerterm institutional problem that has been exacerbated by the kind of administration that this one clearly is one where it literally uh you know stripped out a whole bunch of capability. Now, that's not to say that there might not have been reasons why there were parts of the public sector that weren't performing, that were um no longer relevant or what have you. Don't need to enter into that argument. Simply have to say that um in the processes of cleaning that out, it would appear that a bunch of people who know about stuff got cleaned out as well. But on top of that, you've also got an administration that um has a has a very uh um how would we say, you know, a very uh fly by the seat of one's pants style. The president clearly has a view about decision-m that revolves around his own particular ability to have insights into things, the instincts that he has, the knowhow that he brings to the table, etc. And therefore is able to make some of these decisions um uh literally on the basis of the gut feelings that he has. And we know this because he's said so about other decisions that he's made.

[00:14:06]And we've seen that in reports around the um you know in the New York Times for example of the meetings in the war room where the um the the 28th of February attack on Iran was discussed and ultimately the decision came down to um the the president's instincts and his gut feelings. So there's a lot of fly by the seat of one's pants uh in the style of this administration.

[00:14:34]There is upside to that. You know, it can have moments of decisiveness. Um but the downside is is on complex things um it often makes, you know, the wrong calls. And I think on infrastructure on the way in which the American AI system can be developed, it has made some uh uh decisions that have for now at least shot itself in the foot. I mean another classic and and this is clearly not in certainly not in the interest of American shareholders but but the the the um restrictions on the export of um Nvidia chips to China um has led to Nvidia having zero footprint in China now. Um, but it's not just that. It actually, as we now know, catalyzed this incredible effort in China to develop a domestic capability. Now, this domestic capability doesn't just mean that Nvidia loses China's market share. As we know once Chinese industries develop a capability that satisfies the quality requirements, the performance requirements and the scale requirements of Chinese industries. Um they rapidly move into global markets. And the last bit of the jigsaw puzzle, if you will, for the pressures that American AI is under is that it's, you know, think four years ago it was the unquestioned um global leader that was going to take over the world, so to speak. and and even to a point, you know, where where Vice President Vance was in Europe what um 18 months or so ago pushing the Europeans to basically open the doors for um American AI companies and now across the world um not only in the United States where they're using these models as I've mentioned across the global south they're now uh facing options that only three years ago they didn't have and those options are between a proprietary American system with all the things that um come with that including I think a growing awareness that there is an incredible amount of data surveillance, data weaponization, etc., etc., on the one hand, and then on the other hand, you've got this sort of Chinese infrastructure that enables what I've called in the past a kind of digital west failure. you know this idea that uh countries can adopt use of Chinese technologies full stack from the energy level all the way through to applications level but in a way that enhances uh energy and information sovereignty rather than conceds it to um a small number of American big technology companies. So, you know, you you put all of that together and um you know, the United States has been kicking some, you know, very serious own goals.

[00:17:39]Yeah.