[nis0s]

If the cost benefit analysis for employers still shows that H1-Bs are cheaper, how will this offset H1-B exploitation? My guess is that this will suppress STEM wages artificially to account for paying a one-time fee for an H1-B, but hiring someone for 1+ years at that suppressed rate will be cheaper. Employers will blame AI for decrease in STEM wages, of course. A complementary solution is to add the $100k fee, and to restrict H1-B per employer per year, or something like that.

[alephnerd]

In 2025, the decision isn't hiring someone on an H1B versus a citizen - the cost is mostly a wash.

The decision is hiring in the US (visa or citizen) versus hiring abroad.

Given that a large number of EMs, PMs, Directors, and even VPs are on some sort of immigration or work visa, this makes it easier to incentivize you as an employer to move some of them back to India or Czechia to open a GCC. This is what has been happening for the past 5 years now.

On top of that, vast swathes of STEM academia are dependent on H1B. You simply aren't going to find enough American citizens with a background in (say) battery chemistry to become a tenure track professor versus from Korea, Japan, or China.

Now you basically created an incentive for large swathes of junior faculty in STEM subfields to return to Asia, leading to a massive reverse brain drain.

[nis0s]

> The decision is hiring in the US (visa or citizen) versus hiring abroad.

True, but there’s a balance that employers have to maintain to get some in-state advantages from local or state governments for job creation.

That said, it makes more sense for America to get trainers or professors for niche subfields than actual workers so you can create homegrown talent, not sure why that isn’t done more.

[alephnerd]

> but there’s a balance that employers have to maintain to get some in-state advantages from local or state governments for job creation

True! The issue is local, state, and federal governments gives limited benefits compared to CEE countries, Israel, India, and others who roll the red carpet with multi-year tax holidays, subsidizes, and targeted hiring pipelines.

> makes more sense for America to get trainers or professors for niche subfields than actual workers so you can create homegrown talent

How? They overwhelmingly came on H1Bs as well, not O-1s.

This is why this is such a stupid approach, and is absolutely showing the hallmarks of a Stephen Miller policy. Interestingly, this seems to have overshadowed the Trump Gold Card and Platinum Card announcements (which part of me thinks was part of the reason this announcement happened).

[porridgeraisin]

> such a stupid approach

What do you think of this alternate one?

Don't make H1-B employer-specific. That way, they automatically have to pay market rates to the guy since otherwise you would sponsor his entry and he'd switch to a market rate employer immediately. This removes the "unfair" aspect of h1bs being cheaper to hire.

[alephnerd]

> Don't make H1-B employer-specific. That way, they automatically have to pay market rates to the guy since otherwise you would sponsor his entry and he'd switch to a market rate employer immediately

Exactly.

That solves the problem of consultancies and firms trying to abuse the H1B program as indentured servitude, and makes it easier for those in that kind of a situation to demand a higher salary.

In product companies, someone on a work visa is paid comparable to an American, and bringing a foreign nation on site is already a bit of a wash savings wise.

That said, with this announcement the ship has sailed, because having to spend $100k per year per H1B filing on top of the salary premium of hiring in the US just made opening a GCC/offshoring even more cost effective. For the top 20% of talent in CEE and India, you're already seeing TC break that $100k mark. The issue is there just aren't enough people in the US in certain subdomains with the right skills.

I blame CS programs over the last 10 years for that by trying to overleverage "Leetcode" and "Fullstack" style courses and increasingly reducing specialized courses.

A CSE major at a decent program in India or the CEE will have studied algorithms, digital signals processing, OS internals, and computer architecture along with the option to take further electives in a specialization of their choice (ML, Security, Systems, HCI, etc). They are much more "well-rounded" for technical roles because they will have dipped their toes in 2-3 technical subfields and did their "data structures" equivalent.

In my subfield (cybersecurity) it's been almost impossible to find the right talent at scale of OS internals, systems programming, CompArch, and CUDA+Infiniband experience in the US for the past 5-7 years. As such, there is a generational skill gap, because there is a gap of people who should be mid-career now but don't exist domestically.

And it's not something a "bootcamp" can solve either. The reality is, if we need to spend 2-3 years retraining people domestically with table stake skills like algos, OS internals, and other courses that are expected in a CS major, we should also dramatically reduce salaries for those employees, becuase I can't justify paying $150k for a bootcamp grad. At $50k-80k the math works out to only hire domestically with that level of skill while also offering training like the ASU BSCS program.

[porridgeraisin]

I'm seeing a mismatch between what was said in news articles and by the govt officials as well as trump in the video, and what is written in the official order[1]

Considering your background, what is your take on this? Many folks I know who are on H1-B don't know what to believe

[1] https://www.whitehouse.gov/presidential-actions/2025/09/rest...

[nis0s]

> A CSE major at a decent program in India or the CEE will have studied algorithms, digital signals processing, OS internals, and computer architecture along with the option to take further electives in a specialization of their choice (ML, Security, Systems, HCI, etc). They are much more "well-rounded" for technical roles because they will have dipped their toes in 2-3 technical subfields and did their "data structures" equivalent.

I am sorry but this seems to lack complete awareness of the standards of many U.S. CS and computer engineering programs.