Analysis · Labor Markets · AI Infrastructure
Jensen's Five-Layer Cake
and the Jobs Nobody
Is Talking About
NVIDIA's CEO framed AI as a massive industrial buildout at Davos 2026. We dug into every job board to see what that actually means for hiring — and the numbers are not what you'd expect.
Data sourced from BLS, Indeed, Veritone, Lightcast · Last verified March 2026
AI Infrastructure Stack — Inspired by Jensen Huang's Davos Thesis
Numbers sourced from BLS, Indeed, Veritone & CHIPS Act reports · March 2026
First principles — what is the AI job market, really?
Pull up any job board and search "AI jobs." You will see prompt engineers, AI product managers, and ML researchers. What you will not see — because it requires aggregating across different job categories — is the much larger set of jobs that AI is creating in the layers beneath the intelligence itself. Every AI model requires power to train. Every inference request requires a chip to run. Every chip requires a data centre to house it. Every data centre requires an engineer to keep it running. The AI job market is not primarily a software market. It is an infrastructure market — and the infrastructure market is enormous, largely overlooked, and hiring aggressively right now.
Jensen's five-layer framework is not just a product roadmap. It is a map of where the jobs actually are.
What Jensen Said at Davos
In January 2026, NVIDIA's founder and CEO Jensen Huang took the stage at the World Economic Forum in Davos and made a deceptively simple argument: AI is not a product. It is not a chatbot. It is not a single model. It is infrastructure — on the scale of the electrical grid or the interstate highway system.
"AI runs on real hardware, real energy and real economics. It takes raw materials and converts them into intelligence at scale. Every company will use it. Every country will build it."
— Jensen Huang, World Economic Forum, Davos 2026The layers, from bottom to top: Energy (the binding constraint — no electricity, no AI), Chips & Computing (the hardware layer where NVIDIA lives), Cloud Infrastructure (data centers, networking, cooling), AI Models (researchers and engineers training the things), and finally Applications (where most people think AI jobs exist).
The Numbers Tell a Different Story
Pull up LinkedIn or TechCrunch and you'd think the AI job market is almost entirely about prompt engineers and AI product managers. That framing misses 92% of the actual hiring happening right now.
Layer 1 · Energy
53%
of all AI-related US openings are in energy & construction trades
Layer 3 · Cloud Infra
140K+
Software/cloud openings per year (BLS, 2023–2033 projections)
Layer 4 · AI Models
35,445
AI-related positions in Q1 2025 alone — 25% YoY growth (Veritone/BLS)
Layer 5 · Applications
~1,800
LangGraph/agent-specific openings on Indeed, March 2026
Layer by Layer: The Full Picture
Layer 1 — Energy: The Foundation Nobody Discusses
Jensen calls energy the "binding constraint" for AI. You can have the best chips in the world and no AI if you can't power them. The surge in data center construction — driven by hyperscalers like Microsoft, Google, Amazon, and Meta racing to build AI infrastructure — has created a skilled trades shortage that is quietly one of the most acute labor market pressures in the country. Construction salaries are growing 8–12% in 2026, outpacing the broader US economy. For young people looking for stable, well-paying work that is explicitly not threatened by AI: the trades are hiring aggressively.
Layer 2 — Chips & Computing: The Bottleneck Layer
This is where AI accelerators, GPUs, and the ASICs that power modern inference are designed and manufactured. There are currently three engineering jobs for every one qualified candidate in the semiconductor sector. The US CHIPS Act has catalyzed domestic fab construction, creating sustained demand for semiconductor process engineers, fab technicians, and VLSI designers — some of the highest-value but lowest-volume positions in the entire stack.
Layer 3 — Cloud Infrastructure: The Quietly Enormous Layer
LinkedIn's 2026 Jobs on the Rise report lists data center technician at #17 fastest growing. But that understates it. When you aggregate SREs, network engineers, data center engineers, and cooling specialists, Layer 3 represents the BLS-projected ~140,000+ software and infrastructure openings per year — significantly more than all of Layer 5 agent-specific roles combined. The cooling engineer role deserves special mention: high-density GPU racks running NVIDIA Blackwell chips generate enormous heat, and every major hyperscaler is now hiring specifically for liquid cooling expertise.
Layer 4 — AI Models: High Value, Severe Shortage
This is the layer everyone is trying to hire for and almost nobody can fill. Q1 2025 saw 35,445 AI-related US positions — a 25.2% jump year-over-year, with AI/ML Engineer the single fastest-growing role at +41.8% YoY (Veritone/BLS). Globally the talent gap is severe: demand consistently outpaces supply across every major market. ML researchers and AI scientists are the scarcest and most compensated. Research scientists at frontier labs — Anthropic, OpenAI, DeepMind — can command total compensation packages exceeding $489K.
Layer 5 — Applications: The Layer About to Explode
With ~1,800 agent-specific openings on Indeed today (LangGraph, LangChain, CrewAI), Layer 5 is still small in absolute terms. But the acceleration is clear: GenAI skill postings grew from 55 in January 2021 to nearly 10,000/month by May 2025 (Lightcast). Forward-Deployed Engineers — the people who take AI from the lab into enterprise production — are the fastest-growing role in this layer. When FDE hiring explodes, it means companies are finally moving from "AI pilot" to "AI in production."
The No-Code Revolution Within Layer 5
A critical development within Layer 5 is the emergence of a distinct sub-layer: no-code and low-code AI roles. Tools like Make, n8n, Zapier AI, Microsoft Copilot Studio, and Salesforce Agentforce are rapidly democratizing AI deployment.
This creates three new job archetypes: the AI Automation Specialist (builds agent workflows without coding), the AI Workflow Designer (maps organizational processes and rebuilds them around agents), and the No-Code AI App Builder (ships production-ready AI apps using visual tools). These roles pay $35–$100/hr and are the most accessible entry point into the AI economy for non-technical professionals.
The Pay Structure: Hourly vs. Salaried
~46% of all tracked AI-related openings have a meaningful hourly pay component, driven by trades work at Layer 1 and the significant contract/freelance market at Layer 5.
| Role | Layer | Pay Type | Avg Rate | Openings |
|---|---|---|---|---|
| Electrician (industrial) | L1 · Energy | Hourly | $28–$58/hr | ~150,000 |
| Data center technician | L3 · Cloud | Hourly | $28–$54/hr | ~20,000 |
| No-code AI app builder | L5 · Apps | Hourly | $35–$100/hr | ~900 |
| Prompt engineer | L5 · Apps | Both | $42–$66/hr | ~3,500 |
| RLHF / fine-tuning eng. | L4 · Models | Both | $58–$173/hr | ~500 |
| ML researcher | L4 · Models | Salaried | ~$109/hr equiv. | ~15,000 |
| AI software engineer | L5 · Apps | Salaried | ~$85/hr equiv. | ~35,000 |
Why Layer 5 Is About to Explode
Every major platform shift in technology history has followed the same pattern: infrastructure builds for 5–7 years, then applications explode for a decade. We are in the final innings of the infrastructure build phase.
1993–1999
Internet infrastructure: Fiber cables, routers, servers. Then Google, Amazon, Facebook exploded 2000–2010.
2004–2008
Mobile infrastructure: Cell towers, chipsets, app stores. Then Uber, Instagram, WhatsApp exploded 2010–2015.
2010–2015
Cloud infrastructure: AWS/Azure/GCP data centers. Then Slack, Zoom, Shopify exploded 2016–2022.
2020–2026 ← we are here
AI infrastructure: GPUs, data centers, foundation models being built now. Application explosion: 2026 onwards.
Jensen's Deeper Point: Purpose vs. Task
Throughout the Davos conversation, Jensen kept returning to a distinction that gets lost in most AI job coverage: the difference between the task of a job and the purpose of a job.
A radiologist's task includes reading scans. Their purpose is diagnosing disease and caring for patients. AI can handle the task of reading scans faster than any human. But the faster it reads, the more patients a radiologist can see. "The number of radiologists has gone up," Jensen said. "The fact that they're able to study scans now infinitely fast allows them to spend more time with patients."
The question Jensen asks is worth internalizing:
What is the purpose of your job?
Not your tasks — your purpose. Wherever the answer is uniquely human — judgment, physical presence, empathy, creativity, accountability — AI makes you more valuable, not less.
The Takeaway
The AI job market in 2026 is not primarily about prompt engineers and AI product managers. It is about electricians in Texas, cooling engineers in Virginia, semiconductor fab technicians in Arizona, and SREs keeping clusters running in Oregon.
Layer 5 — the application layer — represents just 8% of current AI hiring. But the infrastructure beneath it has spent five years being assembled. The models are now good enough. The inference costs are falling. The no-code tooling is maturing.
The application explosion is coming. The only question is how fast — and whether the workforce is ready for it.
Start a Critical Discussion
These questions don't have consensus answers. Share one to LinkedIn or X and see what your network actually thinks.
"Is AI infrastructure more like railroads in 1870 or the internet in 1999 — and does the distinction matter for your career bets?"
"Jensen says Layer 5 (applications) hasn't exploded yet. What's actually blocking it — talent, trust, or tooling?"
"The trades (electricians, cooling engineers) are the #1 AI job category by volume. Is this fact criminally underreported?"
Share this analysis
If this changed how you think about something, share it. The AI workforce conversation needs more data and less hype.
References
All sources accessed March 2026