Amazon’s latest PR trick goes something like this: “AI made me do it.” First, the company blames artificial intelligence for a wave of layoffs. Then, according to a group of US senators, it quietly backfills roles with lower-paid H-1B workers. If that sounds like corporate aikido—use the hype cycle’s energy to justify headcount cuts while still hiring—well, that’s because it is.
The allegation, laid out in letters and prodding from Capitol Hill, is simple: Amazon publicly tied job cuts to automation and “efficiency” gains, then sought to import talent on H-1B visas at wages below what comparable US workers earn. If true, that’s a neat piece of narrative arbitrage. It also puts a match to two of the most combustible topics in tech right now: AI’s impact on labor and the long-running abuse of the H-1B system as a cost-control valve.
Let’s separate the theater from the mechanics. Companies aren’t required to stop hiring just because they lay people off. Teams expand and contract. Projects die; others spin up. But senators say the timing and the roles don’t pass the smell test. On one hand, Amazon telegraphed that AI would eliminate or “redefine” jobs. On the other, it filed Labor Condition Applications—the paperwork prerequisite for H-1B hiring—for roles that look awfully similar to ones recently cut. The implication: AI wasn’t the whole story. Cost was.
“Cheap” in H-1B land doesn’t always mean poverty-level wages. It often means cheaper than the local median for that job. The Department of Labor’s prevailing wage system sets four levels tied to local markets. Level 1 is roughly the 17th percentile, Level 2 around the 34th, Level 3 at the median, Level 4 near the 67th. Plenty of top-tier tech firms legally anchor offers at Level 1 or 2 when they can. The Economic Policy Institute has long flagged that a majority of certified H-1B positions industrywide cluster at those lower levels. That’s all within the rules; the question is whether the rules still reflect reality in a market where “entry level” software engineers are shipping production AI models.
Also worth remembering: The H-1B cap remains 85,000 new visas a year (65,000 general, 20,000 for US grad degrees), a number that hasn’t budged in forever despite tech’s footprint exploding. Scarcity keeps wages high for domestically hired talent. That creates an obvious incentive to use every legal lever—global distributed teams, contractors, or visas—to lower average labor costs across the org chart. You don’t need a PhD in labor economics to see how this story writes itself.
The senators’ demand for receipts focuses on two things. First, wage parity. Are H-1B software engineers, data scientists, and cloud roles paid at or above the prevailing wage? Are those wage levels calibrated to real market rates, or anchored to outdated surveys that undercount AI-era skills? Second, displacement. Did Amazon lay off US workers and then put H-1B workers into “substantially equivalent” roles within the non-displacement window? The law is stricter for “H-1B dependent” employers, but that threshold is comically high for mega-corps with hundreds of thousands of employees. Translation: It’s easy to be compliant on paper and still play the edges.
Amazon, for its part, will say it follows the law, pays competitively, and hires H-1B workers to fill hard-to-staff specialties. All of that can be true. It can also be true that messaging “AI eliminated this job” while hiring new talent on the cheap is a contradiction, not a strategy. Most Big Tech cuts since 2022 have been about margins, not algorithms. Wall Street demanded discipline after the pandemic sugar high. CFOs obliged. AI became the perfect narrative cover: futuristic, inevitable, and conveniently abstract enough to avoid specifics.
That being said, there’s a bigger, messier truth here: AI isn’t replacing most of these jobs outright—yet. It’s changing the mix. Companies are swapping generalists for staff who can ship AI-assisted features, tune inference costs, wrangle vector databases, and build retrieval-augmented apps that don’t hallucinate themselves into a lawsuit. That talent is scarce. Scarce talent is expensive. If you can legally buy some of it at Level 2 instead of paying Bay Area cash comp at the 75th percentile, you do it. This is not a morality play. It’s a spreadsheet.
But it is also a policy failure. The H-1B program was designed to fill genuine skill gaps, not to arbitrage local wage medians. If senators want to make noise that isn’t just election-season theater, there are obvious fixes:
Update prevailing wage data and push more roles to Level 3 as the default. If the market has moved, the baseline should too.
Tighten non-displacement rules for very large employers, not just firms that meet the technical “H-1B dependent” threshold.
Increase transparency. Publish LCAs with richer job detail and salary bands that allow outside auditors—and, yes, journalists—to spot patterns in near real time.
Raise the cap alongside tougher enforcement. If the US actually wants to win the AI talent war, gating entry at 85,000 while playing whack-a-mole with wage gaming is self-defeating.
None of this should become an excuse to demonize immigrant workers. They’re not the villains of this story; they’re pawns in a system designed by Congress, optimized by corporate legal teams, and enforced by perpetually under-resourced agencies. The solution is better rules and better enforcement, not nativist cosplay.
For Amazon, the reputational risk is bigger than a stern letter. The company’s AI pitch rests on trust: trust that it can deploy automation without wrecking livelihoods, trust that its models won’t melt down in production, trust that it’s a responsible steward of an insanely powerful technology. “We cut you because of AI, then we hired cheaper people” undermines that story. It also gives regulators a clean narrative thread that connects labor, antitrust, and AI governance into one tidy oversight package.
And for workers? Expect more of the same turbulence. AI will keep reshaping role definitions and team structures. Companies will keep using every lawful means to lower blended costs. The best defense remains the boring one: build skills that sit closer to revenue, infrastructure, and model performance. If you’re proving or protecting ROI, you’re harder to replace—by a bot or by a cheaper human.
What to watch next: whether the Department of Labor opens audits off the back of these letters, whether USCIS pushes forward with H-1B “modernization” that tightens wage rules, and whether Amazon’s next quarterly filing shows net headcount growth in technical roles even as it retires others. Also keep an eye on LCA filings; they’re public, and they tell you a lot when you aggregate them across job families and locations.
AI didn’t force anyone’s hand here. It just made a convenient scapegoat while the usual cost-cutting machinery kept humming. Call it automation theater. The plot is familiar, the special effects are new, and the ending depends on whether policymakers decide to actually rewrite the script. (via Ars Technica)
No Comments