AI Lobbyists Use 'China Threat' to Push Deregulation and Gain

Name: bUt wE cAn"T lEt cHinA WiN tHe AI aRmS rAcE!!
Uploaded: 2026-05-14T14:02:44+00:00
Duration: 19 min 26 s
Channel: How Money Works
Description: Summary and key takeaways on bUt wE cAn"T lEt cHinA WiN tHe AI aRmS rAcE!! — Summary, covering The “China Threat” Lobbying Campaign AI firms poured more than

How Money Works

May 14, 2026

•

19 min video

•

2 min read

YouTube video ID: UgDEyQ1h-EA

Source: YouTube video by How Money Works — Watch original video

PDF

AI firms poured more than $100 million into lobbying last year. Nvidia multiplied its lobbying budget fivefold, and OpenAI dramatically expanded its own efforts. The “Build American AI” network channels money to influencers, pushing the claim that AI must stay in the United States or China will win.

The Three Big Ironies

Irony 1 – Companies lobby to sell cutting‑edge chips such as Nvidia’s H200 to China while warning that Chinese AI dominance is a generational threat. A 50 % cap on H200 shipments could still allow up to one million units to flow south, boosting Chinese compute capacity by 2.5 times compared with a domestic‑only supply.

Irony 2 – The industry frames deregulation as a national‑security necessity, yet the unleashed compute power fuels consumer products like image generators and chat tools rather than defense systems.

Irony 3 – Firms avoid “weapons‑grade” regulation—licensing, background checks, end‑use restrictions—that would match the existential threat they proclaim. As one commentator put it, “The harmless version comes out in front of regulators, where the goal is to keep liability frameworks off the table. The species ending version gets saved for investor decks and government grant committees.”

The Reality of Data Center Projects

Kevin O’Leary announced massive data‑center proposals in Utah and Alberta, citing the need to outpace China. The projects secure tax exemptions and fast‑track regulatory approvals, yet little construction has materialized. Investors receive government‑approved packages that can be flipped for profit, illustrating a pattern of personal gain from public incentives.

Comparative Regulation: US vs. China

China enacted mandatory algorithm registration, AI‑generated content labeling, and a court‑ordered rule that companies cannot fire workers solely to replace them with AI without retraining or severance. These measures took effect on January 1 and September 1 respectively. In contrast, the United States lacks comparable federal AI laws, and industry lobbying actively blocks state‑level attempts to impose similar oversight.

Takeaways

AI firms spent over $100 million on lobbying last year, with Nvidia quintupled its spending and OpenAI markedly increasing its efforts.
The industry simultaneously lobbies to sell advanced chips like Nvidia’s H200 to China while warning that Chinese AI dominance is an existential threat.
Companies promote deregulation by framing AI as a national‑security issue, yet the resulting compute power powers consumer products rather than defense applications.
Kevin O’Leary’s announced Utah and Alberta data‑center projects use government incentives to attract investors, but little construction has materialized.
China has enacted mandatory algorithm registration, content‑labeling and labor rules, while U.S. lobbying blocks comparable federal or state regulations.

Frequently Asked Questions

Why do AI companies claim the "China threat" while selling advanced chips to China?

They say that limiting chip sales would let China win the AI race, so they frame the issue as a national‑security threat. This rhetoric fuels lobbying that pushes for looser rules, allowing firms to keep selling high‑end hardware like Nvidia’s H200 to China while publicly warning about the same threat.

How does China’s AI regulatory approach differ from the United States?

China has enacted mandatory algorithm registration, AI‑generated content labeling, and labor protections that prevent firing workers solely for AI replacement, whereas the United States lacks federal AI laws and industry lobbying actively blocks comparable state‑level regulations.

Who is How Money Works on YouTube?

How Money Works is a YouTube channel that publishes videos on a range of topics. Browse more summaries from this channel below.

Does this page include the full transcript of the video?

Yes, the full transcript for this video is available on this page. Click 'Show transcript' in the sidebar to read it.

Helpful resources related to this video

If you want to practice or explore the concepts discussed in the video, these commonly used tools may help.

Books On Political Lobbying And Influence Recommended

Provides deeper context on how corporate lobbying and dark money campaigns shape government policy, mirroring the tactics discussed in the video.

Amazon →

Books On The History Of Technology Policy

Explores the historical relationship between government regulation and technological innovation, helping the reader contextualize the current AI policy debate.

Amazon →

Books On The Geopolitics Of Technology

Analyzes the global competition for technological dominance, providing a broader framework for the 'China vs. US' narrative presented in the brief.

Amazon →

Links may be affiliate links. We only include resources that are genuinely relevant to the topic.

Summarize another video

Full Transcript YouTube

Major AI companies have quickly and quietly become 
some of the biggest lobbyists across America,  
spending over $100 million to influence policy 
in the last year alone. In what might be their  
first major investment to actually generate a 
financial return, OpenAI has massively increased  
its lobbying efforts year-over-year. In addition 
to the usual suspects that have also uped their  
political contributions and Nvidia, who 
has typically been quiet in Washington,  
quintupling its spending to make sure that 
their voice was heard. And well, all of their  
well-funded voices have overwhelmingly been saying 
one thing. I mean, we can't let the Chinese beat  
us in AI. We are going to be the dominant player 
or China's going to be the dominant player. But  
it's nothing compared to what will happen if 
China beats the US on the ultimate goal of AGI,  
artificial general intelligence. Yeah, 
that's right. If we overregulate AI,  
China will win. If we underregulate AI, China 
will win. If the government doesn't support  
and financial backs stop AI development, China 
will win. If we sell cuttingedge chips to China,  
they will win. But if we don't sell cutting 
edge chips to China, they will also win. AI  
lobbying has become some of the most effective 
in Washington because of China. And well,  
you know what? As dumb as it sounds, they might 
actually have a point. Despite all of the talk  
around how indispensable Nvidia is to the current 
AI ecosystem, they themselves have admitted that  
their market share in China is now effectively 
zero. As local suppliers have developed their own  
alternatives and that homegrown Chinese models are 
matching or even exceeding American models while  
using supposedly inferior hardware and using far 
fewer resources, why would they admit to that if  
it wasn't true, right? At the same time, it's 
also becoming increasingly clear that current  
AI models might fall short in a lot of the 
applications they have been pitched on. But  
they are becoming incredibly effective tools for 
mass surveillance, war fighting, cyber warfare,  
and even just simple propaganda. So, if this 
really is an existential threat that we simply  
can't let China gain the upper hand in, it is 
worth at least understanding the hypocrisies  
around how we are dealing with it. Citizens 
will be on their best behavior because we're  
constantly recording and reporting everything. 
For example, active cyber attacks attack a whole  
country. Do it until everybody's dead. And you can 
imagine that scenario. You can also imagine the  
scenario where you say, "I want to kill a million 
people. Show me a biological path to do it." New  
companies cannot comply with can't afford to 
comply with. It's too complicated. It's too  
hard to get to from a standing start. And then 
what happens is that big company now basically  
has a permanent government supported monopoly. 
We believe that excessive regulation of the AI  
sector could kill a transformative industry just 
as it's taking off. Late last week, Kevin Olirri,  
the method acting movie star who occasionally 
plays a businessman on TV, announced that he would  
be building the world's largest data center in 
Utah, a project that is slated to use more energy  
than the rest of the state combined and cost 
tens of billions of dollars to build out. Now,  
despite wearing two very expensive watches 
at the same time, Kevin Olirri does not have  
tens of billions of dollars. So most of his 
work has been dedicated to securing favorable  
government approvals, regulatory fasttracks, and 
stateisssued tax exemptions with the vague hope  
that these can be offered up to a consortium of 
investors who will ultimately fund this project.  
Last week's announcement also came around 16 
months after a similar announcement by Olyri  
that he would be building the world's largest 
data center, but this time in Alberta, albeit  
with a similar pitch around an accommodating local 
government that would cut red tape and hook them  
up to a practically unlimited supply of cheap 
natural gas. Now, as a shock to basically nobody,  
to date, there has been effectively no tangible 
progress on either of these projects. Now,  
we have already made a video about how a large 
share of data center construction sites have been  
quietly cancelled over the last year. And out of 
all of the data centers that will never get built,  
these are surely the least buildiest of all. But 
there is a chance, and I can't believe I'm going  
to say this, that Olyri may not be as silly of 
a sausage as he seems. Now, the only reason why  
these proposals are really worth mentioning in 
the first place is because of the messaging that  
he has been using to make several governments 
bend over backwards to roll out the regulatory  
red carpet for a project. they surely knew was 
never going to see the light of day. Not in this  
case. We're building power from scratch from the 
pipeline. And by the time the pipeline is natural  
gas, natural gas and we're going to burn it with 
turbines clean, but we can also give back to the  
grid. That's good for the community, but for the 
country, we need to compete with China. So yeah,  
it's clearly an effective message and for what 
it's worth, we probably don't want China to take  
a commanding lead on this technology. But there 
are three big dumb ironic problems to how we are  
currently addressing that threat. The first is 
that we are really only fighting this supposedly  
existential arms race when it is profitable to do 
so. The test of any threat narrative is whether  
the people pushing it actually act like the threat 
is real. And as it turns out, they really don't.  
The same companies warning that a Chinese AI 
takeover is a generational emergency are also the  
ones lobbying to sell their best chips into China, 
the ones telling regulators their products are too  
harmless to need a liability framework, and the 
ones actively fighting any regulation that would  
limit how fast they can scale. So, these tools 
exist in some kind of techbro superp position of  
simultaneously being the most existential threat 
to American dominance since the Cold War while  
also being too harmless to require any regulation 
within America itself. It's a dumb message,  
but it's one that the industry is spending an 
ungodly amount of money on amplifying. Beyond  
the nine figures of direct political lobbying 
that require official disclosures and major  
uh podcast acquisitions to promote their talking 
points, certain companies in the space have also  
been exposed for running what are referred to 
as dark money campaigns that sit just outside  
of official regulations. As an example, OpenAI 
founders and early investors have been funneling  
millions of dollars into a group called Build 
American AI, which is in turn redirecting  
that funding to paid influencers to promote 
effectively the same narrative as Mr. Wonderful,  
just with a little bit less transparency. A AI is 
something that needs to be kept in America. And B,  
we can't afford to let China win. The end goal 
is pretty openly to pave the way for a government  
bailout should open AI need it hypothetically. 
Broadly speaking, they understand the AI is not  
politically popular. Government bailouts are 
not particularly popular either, but they are  
hoping that Chinese AI is even less popular. 
Now, the channel Wall Street Millennial did a  
great video on this earlier this week, literally 
as I was putting this video together, that went  
into detail about how they are channeling 
this money and crafting this message. So, I  
will leave a link to his video in the description 
alongside Taylor Loren, the journalist influencer  
who originally broke the story when she herself 
was offered money to repeat their talking points,  
but decided to expose them instead. Either way, 
it's pretty clear that this is happening, and this  
is the message they are running with. So, while we 
might not have a 9-figure lobbying budget to push  
back against this, we do have basic logic to point 
out the flaws in their argument. The first irony  
is the most basic one. The same companies warning 
about Chinese dominance are also the ones lobbying  
hardest to sell their cuttingedge technology 
directly into China. For the better part of  
2 years, the headline messaging out of Nvidia, 
OpenAI, and most of the major AI labs has been  
that letting China get ahead is a civilizational 
risk that the US simply cannot allow. And then  
late last year, when it started to look like the 
AI sector might be headed for a market correction,  
uh that messaging got a little bit more flexible. 
In December, the government announced that Nvidia  
would once again be allowed to sell its H200 
chips into China with the regulation formally  
taking effect around the Department of Commerce in 
January. The H200 is, by Nvidia's own description,  
one of the most advanced production AI chips on 
the planet. China bound shipments are technically  
capped at 50% of US sales and routed through 
US territory for inspection. But in practice,  
the cap on the H200 alone could allow up to 1 
million units to flow into China with potentially  
another million H100s on top. According to the 
Council on Foreign Relations, that is enough  
silicon to build the largest AI data center on 
Earth and to add roughly 2 and 1/2 times Chinese  
AI compute compared to a scenario where they only 
had domestic chips to work with. So the same year  
we were told that China overtaking the US and 
AI was a generational threat. We also decided  
that selling them the actual chips required to do 
exactly that was perfectly acceptable as long as  
the right American company was getting paid. Now 
that Nvidia is allowed to sell into China again,  
they are not exactly thanking everyone and getting 
on with it. They are now lobbying to sell even  
more advanced chips on the basis that the H200 
isn't competitive enough on its own to actually  
win them the Chinese market, which is impressive, 
confusing, both when you remember that this is the  
same Nvidia that has spent 2 years explaining 
to investors that CUDA makes their hardware  
essentially irreplaceable across the entire AI 
stack. One of those two stories has to be wrong  
because they directly contradict one another. 
Either Nvidia chips are uniquely irreplaceable, in  
which case Chinese labs will be paying through the 
nose to use them no matter how stripped down they  
are, or they aren't. In which case, the entire 
valuation of every AI company downstream of Nvidia  
is built on a much shakier foundation than the 
market has been pricing in. The contradiction does  
not stop with chips either. The same companies 
that will on a Tuesday tell a Senate committee  
that their products are essentially harmless and 
don't need any kind of liability framework will on  
a Wednesday tell a podcast host that these are the 
most powerful creations in human history with a  
real chance of wiping out the species. Do you have 
a bunker? I have like underground concrete heavy  
reinforced basement, but I don't have anything I 
would call Hold on. Hold on. Hold on. Dude, look,  
I'll let you I'll let you keep me on the ropes in 
a lot of this conversation, but I am going to call  
that out of the dang bunker, dude. The harmless 
version comes out in front of regulators, where  
the goal is to keep liability frameworks off the 
table. The species ending version gets saved for  
investor decks and government grant committees, 
where bigger stakes mean bigger checks. Both  
can't be true, but both are extremely useful 
depending on which room the lobbyist happens  
to be standing in. And that's just irony number 
one. The next two are even harder to explain away  
and they reveal a lot more about what this 
lobbying campaign is actually buying. So,  
it's time to learn how many works to find out why 
the loudest voices warning about losing this arms  
race might be the exact same voices guaranteeing 
that we do and what their version of winning  
actually looks like. The biggest companies in the 
world are spending hundreds of millions of dollars  
on lobbyists to make sure the AI race plays out 
in their favor. And we are going to get into all  
of that. But while we are doing that, a lot of 
regular Americans are quietly going in the other  
direction, building something for themselves 
that does not depend on what way a senator  
votes. New business applications have stayed 
near record highs in the United States for 4  
years straight. And a lot of those start the same 
way with somebody who has a design or an audience  
trying to put it on a hoodie or a mug or a tote 
bag without becoming a logistics company to do it.  
This video is sponsored by Printify, a print 
on demand platform that handles every part  
of merch. Nobody actually wants to deal with 
the inventory, the packaging, the shipping,  
and the returns. You design what you want in 
their editor, mock it up on the actual product,  
and that becomes your product. Nothing 
exists physically until a customer buys it,  
so there is no inventory order, no warehouse, 
and no boxes piling up in your living room. When  
the order comes in, Printify routes it to one of 
their print partners near the customer, prints it,  
and ships it directly. It connects directly to 
Shopify, Etsy, Tik Tok Shop, eBay, Wix, and a  
handful of other platforms. So, your products show 
up wherever your audience already shops, and you  
keep the profit between Printify's cost and what 
your customer pays. You can even determine what  
you could make using their profit calculator. So, 
if you have a design idea, a podcast, a community,  
or a brand, this is about as low friction as it 
gets to start a real business and an income on  
the side. Let's start today building a business 
with Printify. Use my link in the description  
to sign up for free and use code how many works 
for 15% off your first order. Only the first 100  
people get this offer. Okay, so the people giving 
the loudest warnings about losing the AI race to  
China are also the people lobbying hardest to 
sell China the chips and tools that would in  
turn let them actually win it. Not a great start, 
but the second irony is bigger because it really  
isn't about who is selling what to who anymore. 
It's about what kind of regulation is actually  
getting pushed through under the banner of beating 
them. Even if you accept that an AI arms race  
is a real and imminent threat, almost none of 
the deregulation that has been pushed through  
under that banner is actually aimed at winning 
it. Most of the recent deregulatory effort has  
gone into two areas. Environmental protections 
and energy prioritization, both designed to get  
private sector data centers online faster. 
The vast majority of the compute that those  
data centers will produce is not going to advance 
defense applications, secure cyber infrastructure,  
or any of the surveillance and war fighting 
systems that actually would matter in a real  
strategic competition with another superpower. 
It is going to add targeting, AI customer service  
replacements, image generators, the 10th 
iteration of a chat product, and the fifth  
season of Fruit Love Island. If this really were 
a national security priority, the simple version  
of policy is straightforward. narrow the scope of 
what counts as a strategic AI project, fasttrack  
those, and treat the rest of the industry like the 
consumer product business it actually is. Instead,  
the policy has been to treat the entire industry 
as if it were the Manhattan project. It might not  
be the best analogy, but something like steel 
clearly has civilian and military applications.  
So, maintaining a ready supply of the stuff is 
a matter of national security. But we aren't  
going to let a car manufacturer ignore safety 
standards just because their products are made  
out of the same stuff as our artillery shells. 
Again, I know it sounds dumb, but the parts of  
regulation that have been most fiercely fought 
against are the parts that have nothing to do  
with national security at all. Liability for 
inappropriate imagery generated by these models,  
mental health protection for kids being exposed 
to AI companions, and actual responsibility  
for crimes these tools have been helped to 
facilitate, if not directly encourage. None  
of those rules would slow down America's ability 
to build a defensegrade AI system. If anything,  
forcing companies to actually clean up the worst 
behavior of their models would probably make them  
better at the things they are pitching to the 
Pentagon in the first place. The fact that those  
are the rules being lobbied against the hardest 
tells you what this campaign is actually about,  
and it has very little to do with China. Now, if 
you genuinely thought you were in an existential  
civilization ending arms race, the obvious 
template is right there in the history books.  
We have done arms races before and we have rules 
for them. Industries with the kind of national  
security exposure these companies claim to have 
are usually regulated like weapons manufacturers  
or nuclear energy operators, meaning strict 
licensing on supply chains, mandatory tracking  
of where the technology ends up, background 
checks on consumers, and end use restrictions  
with criminal penalties for violations. It might 
shock you to learn that none of the people warning  
that AI could end humanity have actually asked to 
be regulated like people who already built things  
that could end humanity. Of course, they don't 
actually want that because that level of oversight  
would massively limit their ability to scale, take 
on capital from foreign sovereign wealth funds,  
and throw a lot of impressive but ultimately 
unfounded numbers around in the press to drive  
a hype. The arms race framing, in other words, 
is really just a convenient line, so long as it  
begins and ends with the government clearing the 
runway for them and never actually putting them  
under the kind of scrutiny it would imply. Which 
is why, going back to where this video started,  
Kevin Oolirri may be cooking something here. 
His actual plan, as far as anyone can tell,  
isn't really to build the world's largest AI data 
center in Utah, or before that in Alberta. it is  
to get on television, sell a story about competing 
with China to as many state and local governments  
as he can find, and then use that uh personality 
to stack up regulatory allowances, tax exemptions,  
and lock in cheap natural gas contracts. And 
then if a real consortium with actual capital  
wants to come along and use those allowances, he 
flips the package to them, potentially with some  
kind of perpetual royalty. In the very likely 
event that nobody actually shows up, he has  
been very careful to keep his side of all of this 
remarkably non-committal. Heads, he wins. Tails,  
the taxpayers of two countries lose. It's the Mr. 
Wonderful metaphor for the whole industry. Get the  
governments to bend over backwards to clear 
the path under the banner of beating China,  
pocket the value of those allowances, and worry 
about whether the underlying project ever ships  
later. It sounds bad, but it gets dumber. The 
China narrative has been effective at getting  
governments on board to cut red tape, but so far 
this is ignoring the biggest irony of all. China  
is regulating its AI sector like crazy. China's 
amended cyber security law, which took effect on  
January 1st of this year, included a dedicated 
AI compliance provision that puts firms on  
the hook for ethics, risk monitoring, and safety 
assessments. The US has no equivalent federal law  
and a real chunk of the current lobbying effort 
is dedicated specifically to making sure no state  
level law fills that gap either. China requires 
mandatory algorithm registration for any service  
offering recommendations or generative content. 
Companies have to file the name of the service,  
the application domain, the algorithm type, and a 
self- assessment report. The US has no equivalent.  
China made AI generated content labeling mandatory 
across text, images, audio, and video with rules  
that came into force on September 1st last year. 
The US has only proposed it. China has published  
a set of national standards on data security, 
pre-training data security, and basic security  
requirements for generative AI services, all 
of which became enforceable last November.  
Last month, a Chinese court in Hao ruled that 
companies cannot fire workers solely to replace  
them with AI and that any restructuring on 
those grounds requires retraining offers,  
fair reassignment, and proper severance. The US 
has had no comparable ruling in part because there  
is essentially no statutory basis from which one 
could even be made. It is also worth pointing out  
that overall investment from China into AI is 
considerably lower than it is here in America.  
So, if this is really a race, spending alone isn't 
going to be the deciding factor. Now, to be clear,  
China is not exactly a model citizen of any of 
this. And there is plenty to criticize about how  
these laws actually get applied. Enforcement is 
real, but selective and politically motivated,  
with rules getting deployed against companies 
that fall out of favor and conveniently ignored  
for companies that haven't. Many provisions are 
written vaguely enough that they can really only  
be enforced when the party feels like enforcing 
them. And most of the financial penalties by  
tech company standards are small enough to be 
priced in as a cost of doing business rather  
than treated as a serious deterrent. So no, China 
is very obviously not running a civics class on AI  
ethics here. A lot of these rules are functionally 
about state control of information rather than  
user safety in any meaningful sense. But that is 
sort of the point. China is strongarmming its tech  
sector into developing AI in a way that for better 
or worse forwards the interests of the state.  
America is told that the way to beat this is to 
let its tech sector do whatever they want in the  
vague hope that private interest and geopolitical 
interest align. From the market's point of view,  
this is already working exactly as intended. 
And if you want further proof of that,  
go watch this video next to find out how OpenAI's 
problems are already becoming our problems.  
And don't forget to like and subscribe 
to keep on learning how money works.

May 13, 2026

Watch Read Summary

Liver Lobules (Portal Triad)

Dr Matt & Dr Mike

MIT OpenCourseWare

May 18, 2026

Watch Read Summary

PDF

The Three Big Ironies

The Reality of Data Center Projects

Comparative Regulation: US vs. China

Takeaways

Frequently Asked Questions

Why do AI companies claim the "China threat" while selling advanced chips to China?

How does China’s AI regulatory approach differ from the United States?

Who is How Money Works on YouTube?

Does this page include the full transcript of the video?

Helpful resources related to this video

Share This Summary

Embed This Summary