Axing jobs to fund GPUs: Meta job cuts could see surge in low paid ghost workers : will you soon be one of them?
As Meta prepares to axe 8,000 staff to fund a staggering $135 billion AI bet, a new hierarchy is emerging. If the AI revolution continues its trajectory, it isn’t just automating tasks and burning through billions of dollars each month; it could be restructuring the global workforce into a narrow elite and a vast, low-paid “ghost” class
Making humans redundant to fund silicon
In May 2026, thousands of high-skilled tech workers will face a cold reality. They are being traded for Graphics Processing Units (GPUs), those chips that have become the black gold of the 2026 economy. Originally designed to make video games look pretty, they are now the engines that drive every AI interaction you have.
Meta’s latest internal memo has confirmed a massive wave of 8,000 job cuts effective 20 May 2026. That’s roughly 10% of its global workforce alongside the cancellation of 6,000 open roles.
Let it be said, as it has before, this isn’t a sign of financial distress; Meta remains highly profitable. Instead, it is a calculated trade-off to fund the most expensive corporate expansion in history.
In 2026, the cost of GPU chips has reached astronomical levels. A single high-end AI chip (like the NVIDIA H100 or the newer Vera Rubin platform launched in early 2026) can cost upwards of $40,000.
To stay competitive, Meta needs hundreds of thousands of these chips. By cutting 8,000 expensive human salaries, Meta frees up the billions of pounds needed to buy the GPUs and build the massive, liquid-cooled data centres required to house them.
These GPUs are so energy-intensive that AI data centres are expected to consume nearly 4% of all global electricity by the end of this year.
Meta is not on its own in the AI race. Microsoft is offering about 7% of its US workforce the option to retire early, CNN reported. This is the latest attempt by a major tech firm to downsize while ramping up artificial intelligence investments.
The one-time retirement program, a company first, will be available to employees whose age and years of service total 70 or higher, Microsoft said.
CNBC and Bloomberg first reported the voluntary buyouts, which the company confirmed to CNN.
Who are the AI winners and losers?
Meta has raised its 2026 capital expenditure guidance to between $115 billion and $135 billion, a nearly 70% increase from 2025. This money is being poured into data centres (like the new $10 billion Hyperion facility in Louisiana) and AI research.
While tech jobs are being scrapped, shorter-term construction jobs are being created. More than 3,700 construction workers have been working on the original 2,250 acres, which was initially set to cost $10 billion. Last October, Meta formed a $27 billion joint venture with Blue Owl Capital to fund the project, suggesting that its ambitions for the site have grown, Data Center Dynamics reported.
Meanwhile, a new Model Capability Initiative is taking place. Meta is installing software on US employees’ laptops to capture keystrokes and screenshots. The goal? To train AI agents to replicate the work of the very people they are laying off.
According to a Platformer report, when one Meta employee asked if there was any way to opt out of the initiative, Meta CTO Andrew Bosworth said “No, there is no opt out on your work-provided laptop.”
EU-based Meta employees are not part of this initiative thanks to GDPR and data protection laws.
Casey Newton, the founder of Platformer, wrote, “Meta contractors have long laboured under much worse conditions. In 2019, I began writing about the lives of Facebook content moderators, whose work was closely monitored by automated systems and could be fired for making just a few errors in a week. Data labellers and model raters for Meta and other companies operate under similar levels of surveillance and job precarity.”
While 8,000 workers are being let go, Meta is reportedly offering packages worth up to $1.5 billion for single elite AI engineers and has invested $14.3 billion into Scale AI to manage its sprawling network of low-paid data “ghost” workers.
Who are the ghost workers and why should you care?
They are the people performing the hidden digital tasks that make modern life convenient. They carry out microwork or crowdwork, such as data labelling for AI or content moderation. In the United States, they are called data workers.
In Nairobi, activists such as Mophat Okinyi have exposed the gruelling reality of data labelling. For years, thousands of Kenyans have spent eight hours a day staring at grainy footage to train models for companies like OpenAI and Scale AI.
This labour is essential for safety, yet it remains obscured. According to a TIME Magazine’s investigation, these workers were often paid less than $2.00 per hour to perform the Human Intelligence Tasks that make algorithms appear brilliant. As the researchers behind the Ghost Work Project suggest, the computer is just a student, and the teachers, located primarily in the Global South, remain invisible to the end-user. Here you can get a glimpse of where ghost workers are most active in Europe.
According to news site Pin, the data labelling market hit $4.89B in 2025 (Grand View Research) and is growing 28.4% annually. Scale AI and Surge AI dominate at the top, while specialists like Sama and iMerit win on quality. RLHF demand is shifting hiring from gig workers to domain experts earning $30-$100+/hr.
While data for these sub-sectors is often fragmented, earnings are largely dictated by national minimum wage laws and the lower-paid standards of the gig economy. While their labour is essential to make AI safe, the corporations that outsource this work arguably rarely value worker contributions since they often outsource the work to markets with low wages. That said, Europeans are among the ghost data workforce, and so are Americans.
The 2026 Ghost Economy: A fragmented pay scale
The “ghost economy”, consisting of data labelling, Reinforcement Learning from Human Feedback (RLHF), and content moderation, operates on a disjointed sliding scale.
The 2025 report by the Communications Workers of America (CWA) and TechEquity, Ghost Workers in the AI Machine, challenges the assumption that exploitative data work is exclusively offshored to the Global South by revealing that US-based workers face strikingly similar conditions.
According to the report, the AI industry is driving a “race to the bottom” where US data workers earn a median hourly wage of just $15, which, combined with a median workweek of 29 paid hours, results in annual earnings of $22,620.
Financial precarity is widespread, with 86% of surveyed US workers worrying about meeting their financial responsibilities and 25% relying on public assistance, including Medicaid or food stamps. The report highlights that US workers suffer from the same “unpaid overhead” seen globally: 66% spend at least three hours a week waiting for tasks to become available, yet only 30% are compensated for that time, effectively pushing their take-home pay well below standard professional levels.
While the UK and Western Europe offer higher rates than the global average, the work is plagued by “unpaid overhead”.
Stated rates rarely reflect take-home pay. Research into 2026 trends shows that up to 30% of a worker’s day is spent on unpaid qualification tests.
Plus, when it comes to LLM contamination, where workers use AI to complete tasks meant for humans, platforms implement stricter, unpaid verification checks, further eroding hourly earnings.
UK vs. Europe: Estimated hourly rate comparison (2026)
Wages are typically split between managed services (contract-based) and independent crowdwork (task-based).
| Region | Managed AI Contractor (Hourly) | Independent Microworker (Hourly) | Key Platforms |
| United Kingdom | £12.00 – £18.00 | £5.00 – £10.00 | Appen, Telus, Amazon MTurk |
| Western Europe | €20.00 – €45.00 | €8.00 – €15.00 | Clickworker, Lionbridge |
| Eastern Europe | €10.00 – €25.00 | €3.00 – €7.00 | Toloka, Sama |
Sources: AI Multiple, GOV.UK – The National Minimum Wage in 2026, Oxford Internet Institute – Fairwork Cloudwork Ratings, The 180i – High Paying Remote AI Jobs in Europe (2026), Eurostat – Minimum Wage Statistics (January 2026), Uvik Software – Offshore Development Rates Guide 2026, Alcor – Software Development & Outsourcing in Eastern Europe 2026
Reclaiming the human narrative
To dismantle exploitation within the ghost economy, all workers must demand visibility. Experts argue that companies should be forced to disclose their human-in-the-loop percentages. That’s the frequency at which human intervention is required to validate, correct, or refine AI output. However, transparency must extend to data worker wages, payment frequency and actual working conditions.
After reading this report, perhaps the next time you interact with an LLM or a chatbot, you’ll appreciate the global network of hidden workers behind those responses. They are often exploited to satisfy a thirst for modern convenience. Conveniences we never truly requested, but which could soon cost knowledge-sector workers their livelihoods.
Get the latest news impacting independent workers each week by signing up for The Freelance Informer newsletter:
