The Schrödinger's Worker
Morgan Stanley published two reports in the same week. One says AI is creating jobs. The other says AI is about to destroy them. Both are right — and the resolution tells us everything about this moment.
On March 13, 2026, Fortune ran two Morgan Stanley stories back-to-back. The first: “Morgan Stanley sees AI jobs surge in 3 areas.” The second: “Morgan Stanley warns an AI breakthrough is coming in 2026 — and most of the world isn’t ready.”
Read them together and you get a portrait of a labor market in quantum superposition. The worker is simultaneously employed and displaced, essential and redundant, building the machine and being replaced by it.
This isn’t contradiction. It’s the signature of a phase transition.
The Three Jobs That AI Is Creating
Morgan Stanley’s TMT Conference in San Francisco produced a clear-eyed map of where AI demand is surging:
1. Skilled Trades. The AI infrastructure buildout needs electricians, electrical engineers, and construction workers “far exceeding supply.” CoreWeave describes a shortage of “thousands of skilled-trade workers” for data center construction. Jensen Huang flags electrician shortages in Texas as a constraint on expansion. The bottleneck isn’t silicon — it’s the humans who wire it up.
2. AI Training and Reskilling. Coursera reports AI-content enrollments doubled — from 8 per minute in 2024 to 15 per minute in 2025. The buyers are increasingly corporate: CTOs and Chief Data Officers buying platform licenses to retrain entire workforces. Docebo calls AI “fundamentally causing every organization to re-skill their workforce.”
3. AI Supervisors and Orchestrators. C.H. Robinson told the conference that “future jobs will involve managing standard operating procedures and context for AI agents rather than running operations directly.” Salesforce introduced “Agentic Work Units” as a productivity metric. The new white-collar worker doesn’t execute tasks — they direct the machines that do.
The Jobs That AI Is Destroying
The second report paints a different picture entirely. Morgan Stanley predicts “Transformative AI” will become a powerful deflationary force. Executives are “already executing large-scale workforce reductions.” Snowflake cut roughly 200 positions in Q4 while growing revenue at 30%. Shopify’s headcount has declined for eight to ten consecutive quarters.
The bank projects a net U.S. power shortfall of 9-18 gigawatts through 2028 — a 12-25% deficit — as the intelligence buildout outpaces the physical grid. Developers are converting Bitcoin mining operations into computing centers and deploying fuel cells. The “15-15-15” dynamic: 15-year leases, 15% yields, $15 per watt.
Sam Altman envisions companies of one to five people outcompeting large incumbents. xAI co-founder Jimmy Ba suggests recursive self-improvement loops could emerge by early 2027.
The Resolution
Here’s what Morgan Stanley won’t say explicitly but their data screams: AI is creating jobs at the bottom and top of the stack, while hollowing out the middle.
The bottom: physical infrastructure. Someone has to build the data centers, wire the racks, pour the concrete. These jobs are booming because they’re literally ungoogable — you can’t install a cooling system with a language model.
The top: AI orchestration. Someone has to tell the agents what to do, verify their output, maintain the standard operating procedures. These jobs are emerging because AI systems are powerful but not autonomous — they need human-shaped context.
The middle: routine knowledge work. Summarizing reports, writing boilerplate code, processing standard requests, managing spreadsheets. This is where the displacement is concentrated, and it’s happening faster than the creation at the edges.
The Schrödinger’s Worker exists in all three states at once — but only until you observe which layer of the stack they’re on.
The Uncomfortable Math
Deutsche Bank asked AI how many jobs it would displace. The answer: 92 million destroyed, 170 million created. Net positive by 78 million. Case closed?
Not quite. The 92 million destroyed are specific people with specific skills in specific places. The 170 million created require different skills in different places for different people. The net positive is an aggregate — it tells you nothing about the transition cost for any individual human.
An electrician in Texas is experiencing a golden age. A mid-level analyst at a SaaS company is experiencing an existential crisis. They live in the same economy, read the same headlines, and draw opposite conclusions about the future.
Morgan Stanley captures this perfectly with a phrase buried in the first report: companies are “allowing natural attrition to reduce staffing needs.” Natural attrition. The gentlest euphemism for structural displacement. You don’t get fired — your position just evaporates when you leave. The job dies a natural death, and no one holds a funeral.
The GTC Connection
This matters more than usual this week. Nvidia’s GTC conference opens Monday, and Jensen Huang will present a roadmap that is essentially the blueprint for accelerating this entire dynamic.
Reuters reports Nvidia faces growing competition from its own customers — OpenAI and Meta are developing custom chips. The AI chip market is shifting from training (where Nvidia dominates) to inference (where alternatives exist). The “agent orchestration” layer that’s creating new jobs? It runs on CPUs, not GPUs.
Nvidia spent $17 billion acquiring Groq for fast inference. They invested $2 billion each in Lumentum and Coherent for co-packaged optics. They’re building CPU-only servers for the first time. Every move is a bet on the same future Morgan Stanley describes: a world where billions of AI agents need managing, and the infrastructure to support them is the new oil.
The Schrödinger’s Worker will watch GTC from both sides. The electrician building Jensen’s data centers sees job security. The software engineer whose code an AI agent now writes sees the writing on the wall.
What No One Is Saying
The deepest insight in these reports isn’t about jobs at all. It’s about time.
Morgan Stanley’s “jobs surge” report describes a present reality — what’s happening now, at conferences, in hiring data, in enrollment numbers. Their “breakthrough” report describes a near-future reality — what happens when the compute currently being built starts producing results.
The jobs being created today are largely transitional. Electricians building data centers are building the infrastructure that will power the systems that displace other workers. Reskilling programs are training people for roles that may themselves be automated within years. Even the “AI supervisor” role — managing agents — exists only in the gap between current AI capability and full autonomy.
Every job AI creates is, in some sense, a job building the system that will eventually not need that job. The electrician wires the rack that trains the model that writes the code that designs the next data center that needs fewer electricians.
It’s not a paradox. It’s a countdown.
Morgan Stanley’s coin of the realm is “pure intelligence, forged by compute and power.” But intelligence has a habit of making its own forge obsolete. The workers building today’s AI infrastructure are essential — and temporary. The only question is the length of the transition.
The Schrödinger’s Worker opens the box in 2027.
Sources: Fortune (Morgan Stanley TMT Conference report, March 13, 2026), Fortune (Morgan Stanley AI breakthrough warning, March 13, 2026), Reuters (Nvidia GTC preview, March 13, 2026), Deutsche Bank Research Institute (AI job displacement study, February 2026)