TL;DR — Big Tech is spending $725 billion on AI infrastructure in 2026 while cutting roughly 170,000 jobs. The headlines target executives. Almost nobody is writing for engineers 1–3 years into their career — the demographic feeling the squeeze hardest. Here is the honest read: which roles are actually shrinking, which are expanding, what your skill stack should look like for the rest of 2026, and what to do if you are currently looking.
The Numbers
- ›Meta: 8,000 cut. Reallocating budget into the Llama 5 frontier model program.
- ›Amazon: 30,000 cut across corporate and AWS. Largest single round in the company's history.
- ›Microsoft: 125,000 voluntary buyouts offered. Of those, roughly 18,000 accepted as of May 2026.
- ›Google: Multiple smaller rounds totaling ~12,000. Recruiting, sales, and middle management hit hardest.
- ›Capex on AI infrastructure across the four hyperscalers + Oracle + Meta: ~$725B for 2026, up from $475B in 2025.
The pattern is unambiguous: capital is moving out of headcount and into GPUs, datacenters, and frontier model training. The dollars are not disappearing — they are migrating.
What Actually Got Cut (And What Did Not)
The narrative "AI is replacing engineers" is mostly wrong. The narrative is "Big Tech over-hired during 2020–2022, then used AI productivity as cover for the correction." Here is the breakdown I am seeing across friends and former coworkers in the cuts:
| Role Category | Trend in 2026 |
|---|---|
| Junior/mid software engineers (1–3 YOE) | Headcount flat to slightly down. Hiring bar much higher. |
| Senior+ engineers (5+ YOE) on AI infra, ML systems, or security | **Aggressively** hiring. Comp up 15–30% YoY. |
| Recruiters, program managers, middle managers | Cut hardest. Most exposed to layoffs. |
| Customer support, content moderation, basic SRE | Heavy AI automation. Headcount declining. |
| Data engineers, platform engineers | Stable to growing. AI workloads need pipelines. |
| Frontend / app engineers | Stable. Less AI-disrupted than expected. |
If you are 1–3 years in and you are scared, here is the realistic picture: the floor is not falling out, but the floor is also harder to get onto if you are looking for your first role. The hiring bar moved up. The bar to keep your seat did not.
What This Means If You Are 1–3 Years Into Your Career
I shipped my first full-stack production system as a junior engineer 24 months ago. I have watched the bar move in real time. Here is what I would do if I were where you are.
1. Ship Production AI, Not Tutorials
Every junior portfolio looks the same: a CRUD app, a Next.js blog, maybe a fine-tuned chatbot. None of that signals 2026-grade.
What does signal: one production system you built that uses an AI model in a non-trivial way, with eval, observability, and at least one real user. That is the bar that lets you bypass the resume screen.
If you do not have access to "real users", build the eval harness as if you did. A working LLM-as-judge pipeline, a Langfuse dashboard, three failure-mode case studies — that is more interview-defensible than a year of tutorial-following.
2. Pick a Vertical That AI Cannot Easily Eat
The roles being automated are the ones with narrow, deterministic, low-variance tasks. The roles that are not: anything involving judgment under ambiguity, cross-system reasoning, or production accountability.
Concretely, the safest places to be in 2026 are:
- ›Infrastructure / platform engineering. Distributed systems, cost engineering, observability.
- ›Security and compliance engineering. The Mythos / Glasswing era (backstory here) pushed AI-aware security up the priority list at every serious org.
- ›Production ML systems. Not training models — *operating* them. RAG infra, eval pipelines, agent orchestration, model routing.
- ›Developer tooling. The teams that build the AI coding tools that everyone else uses.
3. Stop Optimizing for Big Tech
Big Tech is not the only path, and in 2026 it is not the best one. Mid-size AI-native companies (Anthropic, Modelia, Cursor, Vercel, Linear, Cohere, Mistral) are hiring at the 1–3 YOE level with cleaner mandates, better learning velocity, and equivalent or better comp once equity is honest.
The pattern: when Big Tech contracts, the talent flowing out of it lands at the AI-native mid-stage companies. Those companies are growing into the gap.
4. Get Excellent At One AI-Augmented Workflow
You should be 3–10x more productive than a 2022-trained engineer doing the same task. Not because AI does the work — because AI does the typing, the docs lookup, the boilerplate. You still do the design.
Pick one stack and go deep. Mine:
- ›Editor: Claude Code + Cursor. Daily driver.
- ›Agent runtime: Anthropic Agent SDK + custom MCP servers (see my MCP production guide).
- ›Eval: Langfuse + LLM-as-judge.
- ›Observability: Sentry + Helicone for LLM calls.
You do not need this exact stack. You need a stack you can defend in an interview and have shipped with.
5. The Honest Conversation About Comp
Comp expectations need to recalibrate. The 2022 "junior engineer at FAANG making $250K" reality is fading. What is replacing it: smaller orgs, equity-heavy comp, more responsibility earlier, faster shipping.
Total comp at AI-native mid-stage startups for 1–3 YOE engineers (US): $140K–$220K base, $30K–$100K equity. India / remote: ₹25–60 LPA at strong AI-native companies, with the upper bound stretching for engineers who ship in public.
The deal has changed. The deal is still good.
If You Are Currently Looking
A grounded checklist:
- ›Update your resume to lead with one shipped AI feature. Not "used GPT in a side project" — *shipped*, with metrics.
- ›Get your GitHub profile working as a portfolio. Pinned repos, clean READMEs, one project with the eval harness visible.
- ›Write one technical blog post that proves you understand a production AI system end-to-end. This is more interview-impactful than a leetcode grind.
- ›Apply to AI-native mid-stage companies first, Big Tech second. Better hit rate, better learning, faster ship cycles.
- ›Engage publicly. Linkedin, X, Hashnode. The Big Tech recruiter pipeline is dead for juniors in 2026; the warm-intro pipeline is alive and well.
Bottom Line
Big Tech spending $725B on AI while laying off 170,000 people is not a contradiction — it is exactly what happens when capital migrates from headcount to compute. The bar to enter the industry is higher. The path through it is more interesting than it has been in years, if you build for the world that actually exists in 2026, not the one that existed in 2022.
If you are 1–3 years in: keep shipping, pick a vertical AI cannot eat, get excellent at AI-augmented workflows, and stop optimizing for the Big Tech logo. The compounding works in your favor for the next decade — but only if you start now.
Frequently Asked Questions
Are software engineers being laid off because of AI?
Mostly no. The 2026 layoffs are primarily a correction of 2020–2022 over-hiring, with AI productivity used as cover. Roles being cut hardest are recruiters, program managers, middle management, content moderation, and basic SRE. Most engineering roles are flat or growing, but the hiring bar is higher.
Which engineering roles are safest in 2026?
Infrastructure and platform engineering, security and compliance engineering, production ML systems (eval, RAG, agent orchestration), and developer tooling. The common thread: judgment under ambiguity, cross-system reasoning, production accountability — things AI cannot easily eat.
What should a junior engineer ship to stand out in 2026?
One production system that uses an AI model non-trivially, with eval, observability, and at least one real user. If you don't have real users, build the eval harness as if you did — LLM-as-judge pipeline, Langfuse dashboard, three failure-mode case studies. That bypasses the resume screen.
Is Big Tech still the best career path in 2026?
Not for most engineers 1–3 years in. AI-native mid-stage companies (Anthropic, Modelia, Cursor, Vercel, Linear, Cohere, Mistral) offer cleaner mandates, better learning velocity, and equivalent total comp once equity is honest. Big Tech hiring is slower and the bar is higher.
What is realistic comp for a 1–3 YOE engineer in 2026?
US AI-native mid-stage: $140K–$220K base, $30K–$100K equity. India / remote at strong AI-native companies: ₹25–60 LPA, with the upper bound stretching for engineers who ship in public. Big Tech junior pay has softened from 2022 highs.
Should I learn AI/ML to stay employable?
Yes, but as a user, not a researcher. Most engineers don't need to train models — they need to operate them. Get excellent at one AI-augmented workflow (editor, agent runtime, eval, observability), then ship something using it.
Written by Harsh Rastogi — Full Stack Engineer building production Generative AI systems at Modelia. Connect with me on LinkedIn for more on Shopify, Generative AI, agentic systems, and production engineering.