The long-held belief that artificial intelligence is stealing jobs has been exposed as a company-by-company divergence across the U.S. Big Tech sector. Companies that moved quickly to ride the AI technology wave have dramatically expanded hiring, while even traditional powerhouses that lagged in AI transformation have been unable to avoid talent outflows.
According to an analysis released on the 27th by Blind, the anonymous workplace community, a survey of roughly 50,000 U.S. members who changed their company settings due to job changes between January of last year and August of this year showed pronounced talent inflows at AI chip and model companies.
Nvidia, AMD, OpenAI, and Anthropic were classified as “inflow-dominant” companies, where new hires outnumbered departing employees. These firms aggressively pursued top talent to maintain AI market leadership, with departure rates classified as “low” while incoming hire rates were classified as “high”—a wide gap.
Intel, by contrast, was singled out as a representative “outflow-dominant” company, with its talent outflow rate classified as “high” during the survey period while its inflow volume was merely “low.” PayPal and Coinbase were also classified as outflow-dominant, with more departures than arrivals, though the absolute scale of outflows was not large.
Amazon, Google, Apple, and Meta exhibited a “revolving door” phenomenon, with both outflows and inflows occurring actively. Employees were effectively circulating among these companies.
The same pattern was observed in the software-as-a-service (SaaS) sector. Adobe, Databricks, and ServiceNow saw more arrivals than departures, while Snowflake, Zscaler, and Asana had more employees leaving than joining.
Blind analysts noted, “Big Tech companies that experienced simultaneous large-scale outflows and inflows during the AI transition were not uniformly reducing headcount but rather restructuring their workforce composition and transforming their organizational DNA.” They added, “When changing jobs now, professionals should not simply look at a company’s name but evaluate whether their role falls within a growth area at that organization.”
This workforce restructuring is attributed to the fact that companies with high-growth business portfolios—spanning AI semiconductors, models, and platforms—and clear strategies have the capacity to reduce legacy headcount while simultaneously increasing hiring of talent with AI development capabilities. Conversely, companies with limited AI transformation results have little choice but to prioritize cost reduction and organizational efficiency. This is why projections suggest that a company’s business direction will matter more than its revenue scale or size in the future hiring market.
New Jobs Created by Data Centers and Power
As the AI industry’s bottleneck has shifted from algorithms to computing power, and now to electricity and land, entirely new job categories are emerging.
James Waddy, who works at a Digital Realty data center in Ashburn, Virginia, walks through a facility spanning roughly 93,000 square meters every day, logging about 30,000 steps. His job involves sniffing for burning smells from equipment, running his hands along pipes to detect cracks, and checking whether hundreds of rooftop condensers have been damaged by birds.
Digital Realty estimates its data centers achieve 99.999% uptime, with a next target of 99.9999%—meaning reducing annual downtime from under six minutes to around 30 seconds.
As power demand surges, a new profession has emerged: the “site selection specialist.” These professionals evaluate remote mountain regions, abandoned mines, and coastal reclaimed land, and negotiate with power companies and regulatory agencies. Google signed a 396-megawatt geothermal power purchase agreement with Fervo Energy on September 1, while Amazon secured 700 megawatts of carbon-free power in Nevada. Meta’s nuclear-related commitments reportedly total up to 6.6 gigawatts.
According to U.S. Census Bureau data, private U.S. companies spent approximately $40 billion per month on data center construction in 2025, a staggering increase from $1.8 billion a decade ago. A single data center building’s power demand reaches 40–60 megavolt-amperes (MVA), and entire campuses often exceed 250 MVA, meaning new grid connections can take years.
According to data from Indeed, more than 1 million people have searched for data center-related jobs this year, with August search volume reaching eight times the level of early 2022. Hourly wages for maintenance and installation workers in data centers were found to be approximately 42% higher than similar roles in other industries. Digital Realty stated that entry-level engineers earn approximately $66,000 annually, with experienced professionals able to exceed $120,000.
AI Safety and Governance Talent Demand Surges
Jobs focused on testing the safety of AI models are also growing rapidly. FAR.AI, a U.S. nonprofit research organization, recently hired a dedicated lead responsible for discovering previously unknown “jailbreak” methods targeting frontier AI models used by hundreds of millions of people. The role involves bypassing multi-layered defense systems—including input filtering, the model’s own refusal mechanisms, reasoning process monitoring, output filtering, and account-level screening—one by one to construct complete attack pathways.
ZipRecruiter, a hiring platform, calculated in early September that the average hourly wage for this role was $67.60, with most positions falling in the $59.60–$78.40 range.
The AI governance field is also expanding. An analysis of AI governance job postings by a U.S. recruiting consultancy found that posting volumes increased in waves, spiking whenever regulatory deadlines approached. The pattern showed low correlation with model release hype; hiring surged when incidents or regulatory issues arose, then budgets contracted again once things quieted down.
SANS’s cybersecurity workforce survey this year found that demand for new-role specialists more than doubled year-over-year, with employers creating positions for AI security engineers and AI governance analysts.
Meanwhile, Microsoft is struggling in the AI assistant market. Its paid Copilot version has reached 30 million enterprise users, but companies are increasingly questioning whether the $21–$38 monthly subscription fee delivers sufficient value. Google’s Gemini and Anthropic’s Claude, both offered free, are expanding market share, while OpenAI’s ChatGPT continues to dominate the free AI assistant market.
The analysis that AI is not simply reducing jobs but reshaping the employment landscape is gaining traction. When electricity, railroads, and the internet emerged, they too created new professions that were difficult to comprehend at the time. According to research published in a quarterly economics journal by David Autor and colleagues, approximately 60% of U.S. jobs in 2018 belonged to job titles that did not exist before 1940. Observers suggest the AI transition is following a similar trajectory.