The Wage Gradient
How the AI boom is strip-mining industrial competence
Welcome back to The Ledger - a weekly briefing of what’s happening inside complex systems around industrial decarbonisation. Welcome to Issue #14.
THE OPENING ENTRY
A controls engineer in the North East gets a LinkedIn message at 2am. A big tech company’s London office. They need someone who understands industrial cooling. £180,000 plus stock. Triple his current salary.
He’s the third controls engineer from his plant to leave this year. All to data centres.
His manager stares at the resignation letter: “We can’t match that. Nobody can.”
The plant makes the precursor for 40% of Britain’s insulin supply. Next month, they won’t have anyone who really understands the cooling loops. They’ll run them manually. Energy consumption will rise 30%.
This is happening everywhere. The AI boom isn’t just consuming electricity - it’s consuming electricians. The data-centre build-out isn’t just using resources - it’s stealing resourcefulness.
We’re watching the greatest transfer of industrial competence in history.
From making things to cooling chips.
THE TALENT MAP
Here’s who’s being hunted:
HVAC Engineers
Industrial: £45–60k → Data centres: £120–150k
Controls Specialists
Manufacturing: £50–70k → Tech facilities: £140–180k
Grid Engineers
Utilities: £55–75k → Hyperscalers: £160–200k
Commissioning Experts
Plants: £60–80k → AI infrastructure: £150–220k
Facilities Managers
Factories: £40–55k → Tech campuses: £130–170k
The velocity: 18 months ago, none of these roles existed at these prices. 18 months from now, the gradient will be twice as steep.
This isn’t a pay rise inside one market.
It’s a different economic physics.
FIELD REPORTS
1. The Cooling Crisis
Chemical plant, Scotland. They lose their chief refrigeration engineer to another Big Tech firm. Six months later, an ammonia leak they would have prevented. £2.4m clean-up. Forty-eight hours of lost production.
“He would have heard it coming,” the ops manager says. “The new guy’s qualified on paper. But twenty years of listening to compressors - that walks out the door.”
His new employer needs him for a new 100MW facility. One building.
The chemical plant cools seventeen processes that make precursors for pharmaceuticals.
“They’re cooling ChatGPT. We’re cooling cancer drugs. Guess who pays more.”
2. The Grid Brain Drain
National Grid has lost a couple hundred engineers to data-centre developers in 18 months. The expertise that once managed 30GW of national infrastructure now optimises connections for individual 500MW sites.
“We’re solving connection puzzles for single buildings that are more complex than entire cities,” says a former Grid engineer now at Amazon. “And the pay reflects that complexity.”
Back at Grid, interconnection applications take three times longer. The people who understand how power systems behave are designing private ones.
Renewables developers complain they “can’t get to yes.” The quiet reason: the engineers who used to approve grid connections now work for the developers applying for them.
The public grid is being run with whoever’s left. The wage gradient has pulled the steepest competence into private networks.
3. The Maintenance Hollow
Steel plant, Wales. Their predictive maintenance team - four specialists who kept 40-year-old kit alive with vibration analysis, thermal imaging, oil sampling.
All four now work for a global tech company’s facility in Denmark. Remote roles. Four times the salary.
Maintenance at the plant is now reactive. Wait until it breaks. Downtime up 300%. Energy use up 15%.
“We had guys who could tell you which bearing would fail next month by touching a gearbox,” the maintenance manager says. “This other company is using them to predict hard-drive failures.”
The knowledge transfer: zero. You can’t export intuition via handover notes. When they left, decades of tacit knowledge left with them.
The knowledge that walked out can’t walk back in.
4. The Construction Cannibalism
Every major contractor has moved their A-teams to data-centre projects.
An industrial project manager shows me his WhatsApp:
“Sorry mate, pulled to Virginia. [tech firm] job. You understand.”
His biomass boiler install - critical for a decarbonisation target - is now running with the C-team. Six months late. Forty per cent over budget.
“Data centres pay penalties for delays that are bigger than our entire project budget,” he explains. “When one of the big data centre firms says jump, every contractor asks how high.”
The wage gradient doesn’t just move individuals. It drags entire supply chains.
5. The Training Tragedy
Technical college, Midlands. Their industrial automation course: 8 students. Their new “Data Centre Operations” course: 400 applicants.
“Why would kids learn to fix factories when factories pay half what data centres pay?” the principal asks.
The automation instructor is leaving too. To teach at a hyperscaler’s training centre. They’re building their own pipeline and hoovering up the existing one.
“We used to train for industry,” he says. “Now industry doesn’t exist at the salary level our graduates expect.”
WHAT THE LEDGER REVEALS
This isn’t just labour-market competition. It’s industrial capability collapse.
Big Tech has quietly discovered something obvious in hindsight: the skills needed to run complex industrial systems are exactly the skills needed to cool and power AI at scale.
Fluid dynamics
Thermodynamics
Control systems
Power management
Predictive maintenance
Same expertise. Just applied to silicon instead of steel.
But here’s the asymmetry no one is pricing:
Every data-centre engineer is a plant that runs 20–30% less efficiently.
Every hyperscale cooling expert is a factory that won’t decarbonise.
Every poached grid engineer is a renewable project that won’t connect.
On paper, productivity rises. In reality, the hands that keep physical systems alive now keep servers comfortable.
THE CASCADE EFFECTS
The wage gradient doesn’t stop at poaching. It cascades:
Level 1: Direct Poaching
Best operators leave for tech → efficiency drops immediately.Level 2: Experience Vacuum
No seniors to train juniors → knowledge transfer breaks.Level 3: Contractor Capture
Service firms prioritise hyperscalers → industrial projects stall.Level 4: Education Exodus
Training pivots to tech → talent pipeline dries up.Level 5: Innovation Freeze
No competence to implement new tech → upgrades stay on slides.Level 6: Systemic Fragility
Ageing systems run by inexperienced teams → accidents and failures rise.
We’re not just losing people. We’re losing the ability to trust that large, dangerous, essential systems are in steady hands.
THE SILICON SUBSIDY
Here’s the hidden economics:
Every data centre is subsidised by the industrial decline it causes.
When a chemicals plant can’t optimise processes, energy consumption rises 20%. That waste becomes the wage premium.
When factories can’t maintain equipment properly, it fails early. Those replacement costs underwrite tech salaries.
When industrial projects slip because contractors migrated to AI campuses, those overruns are part of the subsidy too.
Industrial energy efficiency has fallen 8% since 2022 - the same period data-centre capacity expanded nearly 800%. Nobody will say these are connected. They don’t have to be. The talent numbers already tell the story.
A hyperscaler executive told me: “We’re not competing with industry for talent. Industry doesn’t exist at our pay scale.”
He’s right.
And that’s the problem. The wage gradient doesn’t look like industrial policy. But it functions like one.
THE COMPETENCE INEQUALITY
We’re creating two economies.
The Silicon Economy
Practically unlimited capital
Winner-takes-all dynamics
Salaries unmoored from local productivity
Focused on bits
The Physical Economy
Constrained capital
Thin, competitive margins
Salaries tied to cost models and regulators
Focused on atoms
Between them: a gradient so steep that talent can only flow one way.
A senior controls engineer summarises it cleanly:
“I spent twenty years learning to optimise chemical processes. Now I optimise TikTok’s server cooling. The processes I left behind are falling apart. But my mortgage doesn’t care about industrial poetry.”
The competence that took decades to build is being repriced in months.
THE SOCIETAL SIGNALS
The competence drain surfaces uncomfortable truths:
We value computing over manufacturing. The market is clear: cooling chips is “worth” more than making chemicals, steel, food or medicines.
Experience is being privatised. Publicly funded education and training now compound to private hyperscaler moats.
Industrial knowledge is dying unrecorded. When these engineers retire from tech, their industrial intuition retires with them. There is no shared memory.
The transition needs hands we don’t have. Every heat pump, retrofit, grid upgrade needs exactly the people currently building data centres.
We’re abstracting ourselves to death. More compute, less competence at the physical layer that keeps societies running.
We’re building an extraordinary brain on top of a body we’re quietly neglecting.
THE OPERATOR’S PLAYBOOK
For industrial companies:
You will not win the salary war. Stop pretending you can.
Compete on ownership, meaning, and mastery, not just money.
Lock in key contractors with multi-year agreements before they’re fully captured by hyperscalers.
Assume every expert can leave tomorrow. Document ruthlessly. Shadow them. Turn intuition into at least partial process.
For policymakers:
Treat this as critical infrastructure risk, not “the market at work.”
Consider talent-retention credits or salary support for critical industrial roles.
Tie data-centre approvals to local industrial-impact assessments and skills compacts.
Fund “industrial fellowships” that let senior engineers split time between public infrastructure and private projects.
For individuals:
If you’re industrial, your leverage has never been higher. Price it accordingly.
If you move to tech, negotiate not just salary but what problems you get to work on. Don’t waste deep competence on trivial optimisation.
If you’re senior, treat documentation as legacy work. The notes you leave may be the only thing standing between a safe system and a fragile one.
THE LEDGER LINE
The wage gradient isn’t just moving people. It’s moving possibility.
Every engineer optimising AI cooling is a factory that won’t decarbonise.
Every technician maintaining servers is a grid that won’t stabilise.
Every contractor building data centres is an industrial project that won’t complete.
We’re building the infrastructure for artificial intelligence by quietly harvesting the competence that keeps the physical world running.
The gradient is steepening. The flow is accelerating. The industrial base isn’t simply declining.
It’s being strip-mined.
The AIs we’re building will soon design perfect factories. They’ll run simulations of workers who no longer exist, optimising processes no one remembers, making things we’ve forgotten how to make.
The future will know exactly what to do.
Just not how.
Thanks for reading, and please share The Ledger if you have found it useful or insightful in any way.
Here’s to what’s possible
Dom