Almost certainly these companies are using similar strategies as the Chinese models which they fear will be made illegal.

It's also the case that it makes it hard to attract customers if your openweight model is banned. A major reason the likes of Qwen, Kimi, GLM, and Deepseek are popular (well, at least highly talked about) is because of the open weight models they gave away.

To some degree even unintentionally this is all inevitable. I know this isn’t what you’re taking about, but it’s a similar line of thought. These models are consuming public free information, and they’re all also producing free public information, so they’re all pissing and drinking into same pool.

If we go down the line of dead internet theory which I’m becoming more convinced of these days, the volume of information that’s not necessarily original or extracted from reality and just interpolated and extrapolated from existing information in different ways by LLMs will greatly outnumber human information coming up.

In which case these models should.. start to converge on the same data I imagine, with slightly different behaviors within that. One big generative orgy feedback loop.

Many of the Chinese optimizations are public because they are published by the Chinese labs themselves. Hard to say what the labs are doing of course.

I wonder if it has anything to do with Apple’s legal case against OAI.

If it doesn't involve a walled garden subject to Apple's domination and control, it's not an Apple product.

DeepSeek has published some really good papers. Lately they're pushing really hard for dramatically cheaper serving costs.

If there are effectively only two AI companies allowed to train or run model inference then demand for hosting, GPUs and data centers goes down, so they'd be able to squeeze their suppliers that much more too.

They've seen the government officials parroting OpenAI and Anthropic lobbies talking points, and they have smart enough people to see the signs of upcoming regulatory capture.

If I were cynical, I'd say it isn't helped by the current US administration.

Let me put it this way: if they're not ripping off all that they can, their investors need to shitcan the leadership.

Apple is allergic to industry-spanning consortiums.

This would support the hypothesis that they are incentivized as compute providers and not simply because they're losing the frontier llm race

Google and Amazon are Anthropic investors right?

> I have no idea why HN still treats them like their the ethical good guy.

I don't know that HN usually does. But they did buy a decent amount of publicity by (initially) telling the US government no in using the models for war.

Nvidia is subject to export restrictions on GPUs to China. So Chinese labs have responded by allowing American hosts to run their open-weight models in the US, where there are no restrictions. We see this in the EU too where Scaleway, and now Hetzner are getting into the inference host space serving Chinese open-weight models.

So Nvidia benefits from more inference providers running many Chinese models that they otherwise would not have been able to service.

Yeah, this letter is just proof of (the by now long standing fact) of corporate power/corporate capture of American governance. Tons of popular protests against big AI? Makes no difference. Joint corporate letter about big AI? Possibly the government will do something.

The future will probably have most of all companies running local models, simply because the alternative would be essentially every company that uses LLMs ending up becoming completely dependent upon Anthropic et al. And that dependence would be milked to the point of absurdity once solidly established.

And as a more general point - more major competitors in a domain is very good for everybody except those competitors themselves, who would rather there be as few as possible.

Different companies have different motivations. I'm sure some are genuine, but a lot of them simply lost the race (to OpenAI/Anthropic/Google) and know they can't compete anymore, so are shifting strategies. How many times has Microsoft done exactly this in the past?

Probably a real effort to push for some kind of sanctions or commerce block on Chinese models. Anthropic just doubled their political spending to $40 mil for the midterms to push for "AI Safety."

Maybe less shocking, but still, so is google. I know they are one of the labs, but that being said they rely far less on gemini for revenue as opposed to the big 2.

Makes sense Microsoft is placing it self as a model host and os providing open weight models for much cheaper on their infra already.

No wonder openai and anthropic are terrified

I may be remembering wrong, but reasoning was first demonstrated by DeepSeek. Edit: I am indeed remembering wrong, seems o1 was first.

They are worried about OpenAI and Anthropic making their own processors and controlling the whole stack is my guess.

NVIDIA sells the tools. The more the better it is for them.

Microsoft and Meta are also-rans at this point. Their best hope of catching up is probably leveraging their infra and open models.

> Equally telling is the list of companies that did not sign it -- including Google, Amazon.

Could you elaborate? I don't understand your implication.

Amazon in particular hosts open weight models via Amazon Bedrock (e.g. Ministral 14B 3.0, DeepSeek V3.2, ...)[1].

In case it changes the context of this comment, I am a Meta employee, all opinions are my own.

[1] https://docs.aws.amazon.com/bedrock/latest/userguide/model-c...

> I'm not sure how else to say it other than that it's honestly just embarrassing to watch China and others run circles around us in the USA.

The gap is shrinking, but considering their models still lag our frontier, I wouldn't say they "run circles" around us.

…and with the size of each class determined by common target memory sizes i.e. 32GB or 64GB or 128 GB minus cache and a little overhead

> (I did apply for OpenAi's Cyber Pilot but got no response)

Same experience with Anthropic's. I applied for my employer, and.. 0 response.

Commoditization of the model layer is to Azure's benefit. It would help to avoid the model layer becoming a platform that commoditizes cloud into a pure bare metal business. It also redistributes 'consumer' surplus back to the end user via a lower effective price, which would conceivably increase demand for IaaS and/or PaaS which are complements.

My hunch is they are all seeing the problems at OpenAI and some other collateral damage on the horizon. They don’t want the US government intervening to try to save OpenAI. A little controlled burn is probably good for the industry long term.

Signatories appear to group into

(a) sellers of hardware (Nvidia, IBM) and those who rent hardware out to others (Amazon, Microsoft)

They want open models because they don't have to license it to run it on their hardware, so they can offer lower cost and inspect what they are offering to clients

(b) open model creators (Arcee, Perplexity, Mistral, IBM, Microsoft)

...who would be unable to work if open-weight models are banned

(c) Software-heavy companies ( Mozilla, The Linux Foundation , ServiceNow, Microsoft, IBM, heck all of them)

...who want models that are cheap to run because that reduces the cost of running an LLM (whether you are trying to replace a software developer or just assist them, the value of a reduced cost LLM is directionally the same)

> Microsoft is somewhat unexpected to me

MS is playing literally all the sides in this whole AI thing. They are a massive investor in oAI, and have lots of rights to use their models for a long time. They also have the cloud where they offer inference w/ all the benefits of already being there, data retention, etc. They also have a research wing that can build models (rumours are they're aiming for frontier-ish). I don't think there's currently an angle in this AI thing that MS hasn't aimed for.

> The call to action is all about keeping the American open weights ecosystem competitive. I didn't even realize that was under threat, so I feel like I'm missing something.

OpenAI and Anthropic can bring American open-weight model providers like Meta to court but not the Chinese. Chinese companies breaking American IP law (whether the law is correct or not) are basically immune. This has been going on forever, I guess enough is at stake finally for the federal government to care.

Google and Amazon both have incredibly large compute deals with frontier labs and GovCloud. They may not want to sour on that cash.

It helps to understand there are multiple factions within the white house and trump admin. One faction is very hawkish towards allies and enemies, especially China. Another is the one sending ICE to invade American cities, the Christian nationalism and such. Another is very focused on wealth expansion for friends and family. There is an amount of Venn to these groups.

The important thing is that Sam Altman and Dario Amodei become trillionaires. Everything must be decided in service of that one primary goal.

Apple would be rather unaffected since they already signed a deal with Google/Gemini.

Because their long-term goal is to make AI a walled garden subscription.

What is a "Chinese model"?

If I change one weight in Kimi, is it still a Chinese model?

If I fine-tune Kimi on pro-America nationalistic freedom loving anti-Chinese propaganda, is it still a Chinese model?

If I distill Kimi from one server to the next without ever directly transferring any weights, is it still a Chinese model?

If Claude accidentally trains on some of Kimi's output, is Claude then considered to be a Chinese model?

If China's next open-weight model is released secretly through a European company, is it still a Chinese model?

i'm glad someone else sees this too. There's nothing open source about an open-weight model. I feel like they're just trying to associate themselves with open source and the positive impacts it has had on software technolgoy.

I’ll never forgive Microsoft for their hostile actions against open source and open standards in the 90s. I’ll never believe any endorsement from them of open-anything.

Yeah, the endorsement logos at the bottom are majority bad actors with regards to openness in code. Many are bad actors with regards to openness in _society_.

Yeah, never ask a fox to count your chickens.

Ok, but please don't fulminate or post unsubstantive comments here.

https://news.ycombinator.com/newsguidelines.html

It's not worthless, you don't have to agree with P on everything, or make it personal - it works as strictly business - only where and when appropriate.

The product strategy of 'consumer-grade' AI making deliberately insecure software, and then selling you limited access to the model that can fix it (if they think you deserve to pay them) is just diabolical.

(Grade 3 AI which can hack both previous tiers is exclusively sold to the highest bidder.)

> K3 is the only frontier model I can have a serious conversation with about my product's security.

This is wrong IMO. You should have a serious conversation about your products security with someone who is actually trained on that subject. LLMs are useless if you don't already know more about the thing than the LLM, or if you don't care too much about the outcome (internal tools etc.)

Both the CEO of Palantir and the CEO of Microsoft have recently criticized very strongly the business model of OpenAI and Anthropic.

Therefore it is no surprise that both companies are among the signatories.

The motivation of Palantir is very transparent, they have launched a product created in cooperation with NVIDIA, which is a turnkey system (Sovereign AI) that includes all hardware (i.e. a rack with servers full of NVIDIA GPUs) and all software needed by a company to run on its premises LLM inference and also training/fine tuning.

Thus the Palantir product competes with OpenAI and Anthropic and it benefits from the existence of open weights LLMs.

The motivation of Microsoft is more obscure for now, but because its criticism was very similar to that of Palantir, about how dangerous it is for corporations to delegate the processing of their AI needs to external entities like OpenAI and Anthropic, I assume that Microsoft is also going to propose an alternative to using remotely the OpenAI and Anthropic APIs.

Unlike Palantir, Microsoft is not likely to suggest self-hosting on premises, but I suppose that they might promote self-hosting on some kind of instances rented from Azure.

> But the letter is really about _American_ open weights models.

AI models don’t really have nationality. They don’t have race/ethnicity fields on their model cards. It’s impossible to determine the country of origin of a model by looking at their weights. There is no DNA test for AIs. They exist in a mathematical space where human-made borders make no sense.

This means IF there is a regulation for open weight models, it has to apply to ALL open weight models. There is no any other option, since you can always train/fine-tune a model to change it’s weights and rebrand it as a new model.

And the verdict is still out on if it's going to be cheaper than employees in the long run. Maybe for companies in HCoL areas (SV/NYC), but any smaller mid-market org in a LCoL area, frontier model token spend may not end up all that cheaper when your employee salaries are $55k-$75k/year so you're mostly evaluating it as an additional force multiplier expense rather than a headcount replacer.

What does copyright have to do with anything? You also cannot copyright an encryption algorithm, but the US Government banned essentially mathematics for YEARS[0].

[0] https://en.wikipedia.org/wiki/Export_of_cryptography_from_th...

You can randomly shift every pixel in a movie by +/-1, doesn't make the original not copyrightable.

The conversation going on in the industry is a bit broader than that. OpenAI's head of strategic futures just said last week that open models are inherently decelerationist, ungovernable, and will slow development on the frontier. You'll notice OpenAI does not appear as a signatory here despite their relationship with Microsoft.

> ...the issue is Chinese models, whether open or not, and that's what's looking to be banned...

Is the US planning to invade China and take away their models? Because otherwise China is still going to have the models even if the US bans them. I don't think they're going to be going with the US position on this one.

Why anyone would care what Hegseth thinks about open weight models? What's wrong with Chinese models? The censorship? I think it's good to have diversity of opinion. Get US flavored output on sensitive subjects from US models and ZH on ZH models. That's a useful tool.

Ideology? I think this administration is the least captured by ideology in the last few decades. It's mostly brazen self-interest, for both entertainment & profit.

> Start with the outcome you believe will be the most in line with the spirit and traditions of the open source community

there's nothing open source about an open weight model. It's more like the binary blobs Linux fought so hard against. If you can't compile, or in this case train, from source then you're missing the "source" part of open source.

They did. Check out the Phi models (e.g., https://huggingface.co/microsoft/phi-4).

Nvidia benefits no matter what because their hardware is being used. Microsoft and Meta benefit by making sure that the gap between "Anthropic/OpenAI" and "everyone else" doesn't widen, or they'll be hopelessly dependent on those frontier labs.

I think if anyone at Apple signs a plea that has the word 'open' in it they spontaneously combust into flames

People unfortunately keep voting for “corporate power” politicians, so corporations continue to be in charge.

If you support the ideas described in Nvidia's letter, you're going against popular opinion. Most Americans do not want to "make advanced AI more accessible, adaptable, and widely available". Indeed, many areas of the country including the entire state of New York have enacted datacenter moratoriums, responding to the widespread public concern that we can't have too much AI or it'll drink all our water.

Deepseek’s R1 paper was the first paper to describe how to do it. O1 was released before R1 came out.

A month before the R1 paper came out, they released the Deepseek math paper which described their method for MoE load balancing.

> Could you elaborate? I don't understand your implication. Amazon in particular hosts open weight models via Amazon Bedrock (e.g. Ministral 14B 3.0, DeepSeek V3.2, ...)

It means Google and Amazon do not care and would not voice their opposition against the government action.

Agreed for NVIDA's strategy. Saw from a previous thread but Joel's commoditize your complement essay makes a lot of sense for NVIDIA. https://www.joelonsoftware.com/2002/06/12/strategy-letter-v/

I think Google can play both sides here: they may end up a frontier lab if OpenAI/Anthropic dissolve, or a regular LLM provider, in which case they could be serving open models.

looking at both openai and anthropic; they are winning something, but a sober look at their finances will make you silently mouth "but what did they win".

Is that different from how companies are reliant on cloud computing?

Oh, that’s interesting. I think the general expectation was that blocking exports to China would result in companies there building their own cards competing with NVIDIA. If it has incentivized them to release open weight models instead, that seems like an unexpected win.

They never did that. They said all they required was a human in the loop for kill decisions. No autonomous kill drones.

Yeah I fully expect companies with lots of GPUs but not a good model like Microsoft and Amazon to just take these open weight models and make money, the GPU expense is the only moat at this point.

It’s classic commoditize your complement, nobody can replicate the cloud providers, everyone can replicate the models with open weights.

Makes sense for Microsoft. Copilot is already capable of being model agnostic (both GH Copilot, obviously, and M365 as well). They're an enterprise software & services company, they want to sell integrated AI tooling in their stack, powered by whatever models their customers want rather than any specific model.

Models are quickly becoming a dumb pipe, like an ISP. Just a commodity. The labs should rightfully be terrified, eventually the value isn't going to come from the model itself but the tools & integrations built on top. Non-tech businesses and non-tech employees don't want to buy API access to a model, they want to buy off the shelf software with its capabilities packaged up into a pretty, easy to use GUI.

I suspect that whatever Musk has agreed to spend on the midterms in exchange for those sweetheart deals the SpaceX IPO got will make that look absolutely comical, but we’ll see.

"leveraging their infra" is a pretty solid play, considering the scope of investments these two companies have made thus far.

They really don't need to be very worried about that. Google, Microsoft, Amazon, et al have decades of experience in shipping custom hardware at scale, and it is not a simple field to pivot into

> AI making deliberately insecure software, and then selling you limited access to the model that can fix it (if they think you deserve to pay them) is just diabolical.

It's also not a very good marketing strategy, secure software and quality also goes in pair and it just makes me doubt about the output of Fable/Sol

I don't think it's nefarious, but the end result leads to a pretty frustrating experience by anybody needing actual security work. (and without the organizational deep pockets to obtain SOC 2 attestation)

I think security is an integral part of any production software, and if you get value from LLM's in software development, it seems likely they can be useful in security too.

My point and frustration is that gatekeeping in the name of Security makes the Chinese models actually better at security than USA models.

Surprised to see I partially agree. They are indeed ungovernable and that is the best part.

Banned in the US, obviously, not worldwide. The problem is if the US wanted to, they actually could enforce such a ban nearly worldwide, by putting Chinese AI companies on a sanctions list and disallowing any US company (or anyone that does business with a US company, which is much of the world) from touching them.

I mean I agree with you, as what I said above is not my position, that's the US government's position.

I think I like this perspective because it points to where regulation might be more usefully applied than “the oracle must be prevented from answering certain question” and more like, “if you handle PII or provide services as a defense contractor, you may not take security advice from an oracle”

Yeah copyright is kinda just like "if we say it's so then it's so"

> but I suppose that they might promote self-hosting on some kind of instances rented from Azure.

Yeah, Microsoft wants to sell Azure compute. They also want to sell their upcoming surface ultra hardware with the Nvidia chip in it. They preached hard on "unmetered intelligence" at this year's BUILD conference, went hard on local AI, and Azure Foundry, Windows Foundry Local, etc.

Copilot, both GH and M365, are also made to be model agnostic. Microsoft sells enterprise services and software, they'd prefer (I'm assuming) to not be locked in and dependent on any 1 or 2 model providers and would prefer plenty of options and competition in that space because they can just easily offer a model agnostic harness, integrated with the rest of their stack.

People certainly don't act that way. A lot of the talk is about how great free Chinese models are. Like, they're grouped by the fact that they're Chinese.

How much of that is organic versus something a agitprop agent is commenting in places like HN would be interesting to know, but people very much talk about that from a nationality perspective.

To be clear: This isn't a technical discussion, it's a political one.

While your point is valid on its own merits, it isn't relevant here.

For hosted models, there's no way to know what the real costs are. But we know what it costs to run the open models, and moore's law tells us that it's going to get cheaper, at least for the same capabilities they have today.

I'm not certain about this statement because Huawei created their own chip.

China is doing open models better. I think there's little going in Hagueseth's head other than the usual reactionary nationalism reflex.

yes, this seems to me to be the reason why these three companies want open weight. everyone wins (but especially big tech) in the open weight world except anthropic and openai. i suspect even google could live with the open weight scenario; they are in a sense the only tech company that is perfectly hedged in this respect.

i suppose this is part of why openai and anthropic have not been shy about advertising the scarier strains of their models, these past few months; they basically have to force the white house's hand or risk being commoditized among the various other options in a microsoft/openrouter dropdown.

I mean movies are up on YouTube slightly pitch shifted lol.

It's a mix of grifts, gaffes, and Project 2025.

This has been the prevailing advice all along, and yet we have security vulnerabilities everywhere that LLMs are good at spotting and exploiting. I think we need more options on the menu.

>open models are inherently decelerationist, ungovernable, and will slow development on the frontier

Ungovernable makes intuitive sense, but I don’t understand how open weight models could be decelerationist or slow the development of the frontier. Was there an argument for those positions?

55k??

The fully loaded cost of employees at that salary level is likely over $100K. You may still be right on the relative expenses, but you can't just look at the top-level salary line.

It's the same patronizing nonsense as when all the US webcos pulled out of the Chinese market in some misguided supposed protest believing the Chinese so incapable of rolling their own Google, Amazon or WhatsApp they will come begging the americans to go back. Now the western tech universe goes to China for ideas.

Well of course OpenAI and Anthropic say those sorts of things, it's what they've always said, particularly Anthropic, but that doesn't necessarily mean that's the position of the US government. In other words the government seems to be listening to some but not all of the lobbying done by the big two.

Human in the loop is infeasible in comms denied areas which is what we have in Ukraine/Russia right now.

It's why Ukraine and Russia are now using fully autonomous drones. That's the only real solution to jamming.

Those salaries seem low for mid sized cities even no?

Great, how about some frontier models like China seems to be doing on a weekly basis at this point. Give us a CoPilot variant that's open source & open weights.

Yes. No law prevents you from provisioning and using your own hardware.

They did something Sam wouldn't do at the very least.

It's all things at once. Perhaps it's incentivized the open weight strategy. It's also prompted Huawei to ramp up production of AI chips to try and create a local competitor to Nvidia. It's also also created a grey market for Nvidia chips within China.

They want to be the successor to Enron.

I don't see why convincing people to overpay for something based on hype and artificial scarcity required any "sweetheart deals" via congress.

True, but Google has their own models and did not sign this either.

The other thing that comes to mind is they want to avoid the government from banning open models as has been suggested by some of the closed models providers.

Isn’t Microsoft perpetually short on capacity?

https://windowsforum.com/threads/azure-capacity-crunch-exten...

I don't know about that, a lot of those bloated enterprise solutions are becoming obsolete fast, and integrations becoming a moot point as these models advance.

It does however totally abandon the question of making a profitable mass-market AI product upon which the scale of those investments was based.

You hire someone that knows security and that guy uses an LLM. Ignorant business guy or the research engineer genius guy won’t be able to do much with just an LLM.

You can easily see this if you are using LLMs in a field you are an expert in

Here's his original post https://x.com/i/status/2078133895766114412

Open weight models challenge the "winner takes all" hypothesis, which motivates the astronomical valuations and acts as a justification for trains of cash dumped into AI development.

In other words, if AI models become a commodity, the "AI leaders" will not be able to burn the cash like they do today, thus resulting in a "slower" (one could say instead "more sustainable") development of AI models.

That is a completely normal programmer salary in many countries, Silicon valley wages are just nuts. For example in Finland 55k before taxes is pretty good for ordinary non-wizard guy who can code, if you are a junior 40k is more realistic (and finding job in the first place can be a huge chaĺlenge..).

There are a lot of unknowns. It will probably be harder to get an acceptable quality outcome in most fields than in the text-based, limited vocabulary, comparatively highly structured, granularly documented, relatively unambiguous world of code. For the ones where it’s even possible with our current approach, the economics might not even be feasible to provide the service. Based on how much money these companies need to start making to keep the infrastructure running, let alone expanding, we just don’t know what the landscape will even look like in a couple years.

It's a copy/paste of the arguments against open source software from decades ago. The argument goes: innovation requires profit motive, and profit motive requires IP rights. Of course, this was soundly disproven in software and I don't see any reason it should be different here.

If by "pulled out of the Chinese market" you mean "were hacked by the Chinese government in a way that compromised their intellectual property and the safety of dissidents", then yes [0]

[0] https://en.wikipedia.org/wiki/Operation_Aurora

What does that have to do with LLMs hosted by Anthropic. They didn't say the Dept of War couldn't make kill bots. They said no using Claude to make kill decisions. A quadcopter drone running on a Raspberry Pi isn't running an Anthropic model locally.

anthropic's models would also be useless in a jammed environment though

They could start by releasing the MAI-1 models they announced recently. https://microsoft.ai/models/

The guy who said those things just joined OpenAI in the last month. He was previously a senior policy advisor, specifically on AI, to the current administration in the White House.

I don't know that the US government has a unified position on open source AI. But it's not like this isn't a discussion happening among policymakers.

I think they're generalizing beyond software engineers with what they're saying. G&A, Marketing, etc roles pay significantly less than software engineers and that $55-75k figure is realistic in lower cost of living areas. Even in Colorado, which the Denver Metro Area is HCOL, there are companies (smaller, regional, not national/global) still offering that level of salary.

Nope, not low at all. In the Midwest that’s exceedingly normal, even for non-entry level. $70k a year is living good in the Midwest, especially if dual income household ($140k a year is plenty when a decent house goes for $300k).

There absolutely are laws restricting what hardware you are allowed to use. e.g. https://www.fcc.gov/faqs-recent-updates-fcc-covered-list-reg...

Maybe for startups or simple orgs, but in any regulated industry/high compliance requirements they are still useful or outright required.

Claude Team w/ Cowork still only lets you set R/W permissions globally for each MCP connector for the whole team, they still don't offer config on a per-user basis. The enterprise controls are sorely lacking first party. Maybe fine for most startups or flat orgs, but any established enterprise with strong identity & access governance is going to have an issue with that. You can make it work, but you need to make your own agent/harness and give it an identity in Entra or whatever else you use, so for a non-tech company without a dev team, right back to square one of preferring to just buy a COTS solution vs trying to hand out direct model access with minimal governance.

(M365) Copilot sucks but Microsoft is putting a ton of work into agent governance & identity that sorely lacking elsewhere.

Not entirely, if models become commodity, then Microsoft's play as an infrastructure & enterprise apps company is the better bet. They don't need to make profitable mass-market AI models, they just need the best tooling on top of any model their customers pick, remaining fully model agnostic.

"AI Product" doesn't necessarily mean "We sell API access to a our models." Ton of companies out there itching to buy a commercial off the shelf product to deliver whatever AI capabilities they need packaged in a nice GUI, with enterprise governance controls, without needing a dev team/team of engineers to integrate it or develop harnesses, etc.

Buy it->have IT click a few buttons in an admin console->Deploy and have it be immediately useful is the play.

Well maybe they should find better training methods or architectures

Tough shit welcome to capitalism

It's also a completely normal salary for non-dev, non-IT roles in most of the US that aren't major HCoL cities, which are the roles I was referring too mostly where the cost tradeoff of AI vs. Hiring humans isn't as clear cut as the labs marketing departments want everyone to believe.

Even in the US, only about 23% of workers earn $100k/year or more.

According to the US BLS, the median income for "knowledge workers" is ~$65k-$72k/year.

People in SF, NYC, etc. live completely different lives from the rest of the country.

Chinese providers are actively debunking that argument by innovating on and releasing open weight models.

??? These are not unknowns. It is pretty well documented how much more an employee costs over their outright comp

We are talking about the U.S. here, and no, 40k or 55k are not good salaries at all. It may be OK in Europe with the lower cost of living, etc., but the price of a new car is 50k in 2026, and a basic fast food meal pretty much anywhere is well north of $10.

The point is did the US reaction work or was it completely counter productive?

Then will attempting to cutoff China again today work or do we expect that also will prove to achieve the exact opposite of what its proponents are claiming?

Yeah, correct, I wasn't thinking of devs specifically, more other office/administrative roles threatened by automation. Even where I live in a relatively HCOL area, any non-dev office role you can still expect to see a$65-$80k salary range, and $80k is pushing toward the top end for non tech roles.

This argument would make more sense if the cloud providers used these devices but you were not allowed to.

Yeah that would make more sense. Though it seems to me AI is currently better at reducing dev headcount than most other roles.

Those are routers - not general purpose computers.

And yet ~64+% of the US makes less than $74k/year. The median, once you filter out part-time and seasonal workers is ~$65k/year.

You're right, they are not good salaries but they are the reality of wages in the US outside of the SF/NYC bubble.

This is extremely regional dependent. Please don’t think 70K is living large across the majority of the Midwest, it is not.

AI services becoming a commodity by definition means that the price will hover around cost to operate. That's just what commoditization means in tech. Free, as in Google creating a free browser for no profit in order to increase the size of their ad market.

The entire justification for the AI buildout, and the valuations of all the major AI labs, is that the service will be priced at some significant fraction of the knowledge work it's replacing. If you think it will become commoditized in the future, you are saying these companies are significantly overvalued at present.

Fair enough, 55k would’ve seemed pretty low to me in the NE for a fresh dev. Not 75k though.

We didn't cut off anything. American companies would have been more than happy to continue offering products and services in the Chinese market.

The Chinese government didn't like that idea. They attacked American companies, stole the IP, and created their own analogs of American corporate offerings. If you are a company and are targeted by a totalitarian government, it's not unreasonable to pull out of that totalitarian government's jurisdiction.

Honestly, the American response to that was not nearly stout enough. Chinese goods and services had been sold in the US without issue for decades at that point, often to the detriment of both domestic and other international market players. When China was expected to play by those rules and allow American companies to do business in a mostly-unfettered way, they attacked the companies and drove them out of their market.

Depending on industry and public versus private here, that $80k+ barrier for non-tech is mostly management positions.

It might not be living large, but it's a very comfortable income for a single person in pretty much the whole region.

55k is low for a dev, but not for a ton of other office/administrative roles that are threatened by AI automation.

> you are saying these companies are significantly overvalued at present.

I do lean toward them being significantly overvalued at present. OpenAI & Anthropic's valuations assume high margin product pricing on raw intelligence itself. Even if assume the labs will start charging value-based pricing, enterprise buyers will pay that pricing to whoever saves them engineering hours to make any cheap model work inside their compliance boundary, no guarantee that's going to be OpenAI or Anthropic.

Microsoft wins either way, proprietary, expensive models or cheap commoditized inference, because they monetize the workflow on top, not the raw intelligence. I also happen to think it's everything "on top" where a lot of value lives. Plenty of non-tech enterprises and companies out there itching for a ClickOps style AI product, especially if they don't have an engineering team, that they can buy off the shelf, meets all the compliance checkboxes, and does what they need without having to build the agents and harnesses themselves with the APIs.