Of course the right answer here was always to open more slots and train more doctors.
To me the clear answer here is to tie $100k-exemption to hospitals funding non-medicare GME slots, and generally handing the AMA the lobbying optimization problem of reversing any barriers to training they've put in. Reasonable to expect the current administration can't competently pilot that and want them to just drop the approach entirely, to be fair.
The Trump Administration is attempting to put a $100k fee on future H1-B applications, including those that are exempt from the lottery and cap, unless of course they choose to waive it for you. I say attempting because Trump’s legal ability to do this appears dubious.
This post mostly covers the question of whether this is a reasonable policy poorly implemented, or a terrible policy poorly implemented.
Details offered have been contradictory, with many very confused about what is happening, including those inside the administration. The wording of the Executive Order alarmingly suggested that any H1-B visa holders outside the country even one day later would be subject to the fee, causing mad scrambles until this was ‘clarified.’ There was chaos.
Those so-called ‘clarifications’ citing ‘facts’ gaslit the public, changing key aspects of the policy, and I still do not remain confident on what the Trump administration will actually do here, exactly.
How Does Any Of This Work?
Trump announced he is going to charge a $100,000 fee for H-1B visas, echoing his suspension of such visas back in 2020 (although a court overruled him on this, and is likely to overrule the new fee as well). It remains remarkable that some, such as those like Jason and the All-In Podcast, believed or claimed to believe that Trump would be a changed man on this.
But what did this announcement actually mean? Annual or one-time? Existing ones or only new ones? Lottery only or also others? On each re-entry to the country? What the hell is actually happening?
I put it that way because, in addition to the constant threat that Trump will alter the deal and the suggestion that you pray he does not alter it any further, there was a period where no one could even be confident what the announced policy was, and the written text of the executive order did not line up with what administration officials were saying. It is entirely plausible that they themselves did not know which policy they were implementing, or different people in the administration had different ideas about what it was.
It would indeed be wise to not trust the exact text of the Executive Order, on its own, to be reflective of future policy in such a spot. Nor would it be wise to trust other statements that contradict the order. It is chaos, and causes expensive chaos.
RTFEO (Read the Executive Order)
The exact text is full of paranoid zero-sum logic. It treats it as a problem that we have 2.5 million foreign STEM workers in America, rather than the obvious blessing that it is. It actively complains that the unemployment rate in ‘computer occupations’ has risen from 1.98% to (gasp!) 3.02%. Even if we think that has zero to do with AI, or the business cycle or interest rates, isn’t that pretty low? Instead, ‘a company hired H1-B workers and also laid off other workers’ is considered evidence of widespread abuse of the system, when there is no reason to presume these actions are related, or that these firms reduced the number of American jobs.
As I read the document, in 1a it says entry for such workers is restricted ‘except for those aliens whose petitions are accompanied or supplemented by a payment of $100k,’ which certainly sounds like it applies to existing H1-Bs. Then in section 3 it clarifies that this restriction applies after the date of this proclamation – so actual zero days of warning – and declines another clear opportunity to clarify that this does not apply to existing H1-B visas.
I don’t know why anyone reading this document would believe that this would not apply to existing H1-B visa holders attempting re entry. GPT-5 Pro agrees with this and my other interpretations (that I made on my own first).
Distinctly, 1b then says they won’t consider applications that don’t come with a payment, also note that the payment looks to be due on application not on acceptance. Imagine paying $100k and then getting your application refused.
There is then the ability in 1c for the Secretary of Homeland Security to waive the fee, setting up an obvious opportunity for playing favorites and for outright corruption, the government picking winners and losers. Why pay $100k to the government when you can instead pay $50k to someone else? What happens when this decision is then used as widespread leverage?
People Notice They Are Confused
Here’s some people being understandably very confused.
People React Sensibly, By Which I Mean They Panic
Here’s some people acting highly sensibly in response to what was happening, even if they had closely read the Executive Order:
Any remotely competent drafter of the Executive Order, or anyone with access to a good LLM to ask obvious questions about likely response, would have known the chaos and harm this would cause. One can only assume they knew and did it anyway.
Clarification And Gaslighting Are Offered
The White House later attempted to clear this up as a one-time fee on new visas only.
It is not only incompetence, it is very clearly gaslighting everyone involved, saying that the clarifications were unnecessary when not only are they needed they contract the language in the original document.
That post has the following highly accurate community note:
None of these three seem, to me, to have been described in the Executive Order. Nor would I trust Leavitt’s statements here to hold true, where they contradict the order.
There’s also the issue that this is unlikely to stand up in court when challenged.
Meanwhile, here’s the biggest voice of support, which the next day on the 21st believed that this was a yearly tax. Even the cofounder of Netflix is deeply confused.
The issued FAQ came out on the 21st, two days after some questions had indeed been highly frequently asked. Is this FAQ binding? Should we believe it? Unclear.
Here’s what it says:
If we are going to ‘prioritize high-skilled, high-paid aliens’ in the lottery, that’s great, lose the lottery entirely and pick by salary even, or replace the cap with a minimum salary and see what happens. But then why do we need the fee?
If they had put out this FAQ as the first thing, and it matched the language in the Executive Order, a lot of tsouris could have been avoided, and we could have a reasonable debate on whether the new policy makes sense, although it still is missing key clarifications, especially whether it applies to shifts out of J-1 or L visas.
Instead, the Executive Order was worded to cause panic, in ways that were either highly incompetent, malicious and intentional, or both. Implementation is already, at best, a giant unforced clusterfuck. If this was not going to apply to previously submitted petitions, there was absolutely zero reason to have this take effect with zero notice.
There is no reason to issue an order saying one (quite terrible) set of things only to ‘clarify’ into a different set of things a day later, in a way that still leaves everyone paranoid at best. There is now once again a permanent increase in uncertainty and resulting costs and frictions, especially since we have no reason to presume they won’t simply start enforcing illegal actions.
If this drives down the number of H1-B visas a lot, that would be quite bad, including to the deficit because the average H1-B visa holder contributes ~$40k more in taxes than they use in benefits, and creates a lot of economic value beyond that.
How To Clear The Market
Could this still be a good idea, or close enough to become a good idea, if done well?
Maybe. If it was actually done well and everyone had certainty. At least, it could be better than the previous situation. The Reed Hastings theory isn’t crazy, and our tax revenue has to come from somewhere.
I would be willing to bite the bullet and say that every non-fraudulent H1-B visa issued under the old policy was a good H1-B visa worth issuing, and we should have raised or eliminated the cap, but one can have a more measured view and the system wasn’t working as designed or intended.
Use for middling talent and wage arbitrage (at least below some reasonable minimum that still counts as middling here) is not something I would have a problem with if we could issue unlimited visas, but I understand that others do have a problem with it, and also that there is no willingness to issue unlimited visas.
The bigger problem was the crowding out via the lottery. Given the cap in slots, every mediocre talent was taking up a slot that could have gone to better talent.
This meant that the best talent would often get turned down. It also meant that no one could rely or plan upon an H-1B. If I need to fill a position, why would I choose someone with a greater than 70% chance of being turned down?
Whereas if you impose a price per visa, you can clear the market, and ensure we use it for the highest value cases.
Talking Price
If the price and implementation were chosen wisely, this would be good on the margin. Not a first best solution, because there should not be a limit on the number of H-1B visas, but a second or third best solution. This would have three big advantages:
The central argument against the H1-B is that you’re taking a job away from an American to save money. A willingness to pay $100k is strong evidence that there really was a problem finding a qualified domestic applicant for the job.
You still have to choose the right price. When I asked GPT-5 Pro, it estimated that the market clearing rate was only about $30k one time fee per visa, and thus a $100k fee even one time would collapse demand to below supply.
I press X to doubt that. Half of that is covered purely by savings on search costs, and there is a lot of talk that annual salaries for such workers are often $30k or more lower than American workers already.
I think we could, if implemented well, charge $100k one time, raise $2 billion or so per year, and still clear the market. Don’t underestimate search and uncertainty costs. There is a market at Manifold, as well as another on whether the $100k will actually get charged, and this market on net fees collected (although it includes a bunch of Can’t Happen options, so ignore the headline number).
I say second or third best because another solution is to award H1-Bs by salary.
Awarding by salary solves a lot of problems. It mostly allocates to the highest value positions. It invalidates the ‘put in tons of applications’ strategy big companies use. It invalidates the ‘bring them in to undercut American workers’ argument. It takes away the uncertainty in the lottery. IFP estimates this would raise the value of the H1-B program by 48% versus baseline.
Indeed, the Trump Administration appears to be intent on implementing this change, and prioritizing high skilled and highly paid applications in the lottery. Which is great, but again it makes the fee unnecessary.
Is The Value Crazy High?
There is a study that claims that winning the H1-B lottery is amazingly great for startups, so amazingly great it seems impossible.
This is a double digit percentage increase in firm value, worth vastly more than $100k for almost any startup that is hiring. It should be easy to reclaim the $100k via fundraising more money on better terms, including in advance since VCs can be forward looking, even for relatively low-value startups, and recruitment will be vastly easier without a lottery in the way. For any companies in the AI space, valuations even at seed are now often eight or nine figures, so it is disingenuous to say this is not affordable.
If we assume the lottery is fully random, suppose the effect size were somehow real. How would we explain it?
My guess is that this would be because a startup, having to move quickly, would mostly only try to use the H1-B lottery when given access to exceptional talent, or for a role they flat out cannot otherwise fill. Roles at startups are not fungible and things are moving quickly, so the marginal difference is very high.
This would in turn suggest that the change is, if priced and implemented correctly, very good policy for the most important startups, and yes they would find a way to pay the $100k. However, like many others I doubt the study’s findings because the effect size seems too big.
The Fee Might Be Too Damn High
A $100k annual fee, if that somehow happened and survived contact with the courts? Yeah, that would probably have done it, and would clearly have been intentional.
Even then, I would have taken the over on the number of visas that get issued, versus others expectations or those of GPT-5. Yes, $100k per year is quite a lot, but getting the right employee can be extremely valuable.
In many cases you really can’t find an American to do the job at a price you can socially pay, and there are a lot of high value positions where given the current lottery you wouldn’t bother trying for an H1-B at all.
Consider radiologists, where open positions have often moved into the high six figures, and there are many qualified applicants overseas that can’t otherwise get paid anything like that.
Consider AI as well. A traditional ‘scrappy’ startup can’t pay $100k per year, but when seed rounds are going for tens or hundreds of millions, and good engineers are getting paid hundreds of thousands on the regular, then suddenly yes, yes you can pay.
I think the argument ‘all startups are scrappy and can’t afford such fees’ simply flies in the face of current valuations in the AI industry, where even seed stage raises often have valuations in the tens to hundreds of millions.
The ‘socially pay’ thing matters a lot. You can easily get into a pickle, where the ‘standard’ pay for something is let’s say $100k, but price to get a new hire is $200k.
If you paid the new hire $200k and anyone finds out then your existing $100k employees will go apocalyptic unless you bump them up to $200k. Relative pay has to largely match social status within the firm.
Whereas in this case, you’d be able to (for example) pay $100k in salary to an immigrant happy to take it, and a $100k visa annual fee, without destroying the social order. It also gives you leverage over the employee.
The One Time Fee Is Still Rather High
If you change to a one-time fee, the employer doesn’t get to amortize it that much, but it is a lot less onerous. Is the fee too high even at a one time payment?
One objection here and elsewhere is ‘this is illegal and won’t survive in court’ but for now let’s do the thought experiment where it survives, perhaps by act of Congress.
Another issue is that they are applying the fee to cap-exempt organizations, which seems obviously foolish.
Presumably they won’t allow the L visa loophole or other forms of ‘already in the country,’ and would refuse to issue related visas without fees. Padme asks, they’re not so literally foolish as to issue such visas but stop the workers at the border anyway, and certainly not letting them take up lottery slots, are they? Are they?
On compensation, yes presumably they will offer somewhat lower compensation than they would have otherwise, but also they can offer a much higher chance of a visa. It’s not obvious where this turns into a net win, and note that it costs a lot more than $100k to pay an employee $100k in salary, and the social dynamics discussed above. I’m not convinced the hit here will be all that big.
I certainly would bet against hyperbolic claims like this one from David Bier at Cato, who predicts that this will ‘effectively end’ the H-1B visa category.
Research universities will recruit a lot fewer foreign-trained scientists if and only if both of the following are true:
That does seem likely to happen often. It also seems like a likely point of administration leverage, as they are constantly looking for leverage over universities.
Need a Doctor, No Cap
American intentionally caps the number of doctors we train. The last thing we want to do is make our intentionally created doctor shortage even worse. A lot of people warned that this will go very badly, and it looks like Trump is likely to waive the fee for doctors.
Note that according to GPT-5-Pro, residencies mostly don’t currently use H-1B, rather they mostly use J-1. The real change would be if they cut off shifting to H1-B, which would prevent us from retaining those residents once they finish. Which would still be a very bad outcome, if the rules were sustained for that long. That would in the long term be far worse than having our medical schools expand. This is one of the places where yes, Americans very much want these jobs and could become qualified for them.
Of course the right answer here was always to open more slots and train more doctors.
Our Price Cheap
Our loss may partly be the UK’s gain.