Workers at Innodata — a New Jersey data-engineering firm whose 10,000-plus employees label data, safety-test models and evaluate robot performance for the biggest names in AI — spent Monday 26 October losing their jobs mid-shift: laptops logged out one by one, official IDs deactivated, termination letters arriving afterwards on personal email. Nearly 550 India-based employees were affected, with cuts also reported in the Philippines, and the company has not publicly confirmed the total, according to India Today’s reporting.
🔍 THE BOTTOM LINE
The AI industry’s labour story has quietly moved from “AI takes the jobs” to this: the people who train and grade the AI can be exited by lunchtime, by a project accounting decision, with payslips unreachable behind the lockout that delivered the news. Innodata says the cuts were a business-requirements ramp-down, not a performance problem — and the market agreed with the substance, punishing the revenue concentration rather than the workforce.
What happened on Monday
Per India Today’s reconstruction from employee accounts: the morning began normally, with staff logging attendance; around lunchtime, word spread that the company’s Philippines team — raters, quality experts, team leads and operational managers — had been dismissed during a Zoom call; by late afternoon, Indian staff found October 26 marked as their “resignation date” in the HR portal, and device access began disappearing one workstation at a time. A termination email reviewed by India Today cites “the ramp-down of the business assignment for which you were engaged” and states the decision “is not related to your individual performance.”
Employees’ own account of the hours that followed is the part that will outlast the stock chart. A WhatsApp group organised by affected workers grew from roughly 500 members on Monday night to 617 by Tuesday, as people tried to establish who had been cut, and several found they had already lost access to HR and payroll systems before they could download payslips or employment records. As one employee told India Today: “There has been no calls from HR, and we didn’t receive the chance to even download our payslips and other important documents as even the access of the HR payroll app was revoked.” Innodata did not respond to India Today’s request for comment before publication, per that report; the company has not confirmed the reported numbers.
What the market did with the news, and why
The stock action is the second story inside the first. Innodata shares fell 16 percent in premarket trading Tuesday to $49.88 after the layoffs surfaced, per Seeking Alpha’s note via TradingView — and by the close the drop from $59.91 to $44.74 was a 25.3 percent one-day fall on roughly thirteen times normal volume, the stock’s worst session since its February 2024 short-report crash, per Cryptonews’ market analysis, which counted about $500 million of market value erased in the session and left the company worth just over $1.5 billion. Innodata reports Q3 results on 5 November — now the likeliest date for a clearer accounting of the ramp-down.
The fear investors priced was never “550 fewer salaries”; it was concentration. Innodata’s own quarterly report discloses two customers supplying roughly 37 percent and 34 percent of Q2 revenue, in a business where project work is “primarily at-will” and customers can “reduce, delay or cancel projects” at any time. The termination letters cite a ramped-down client assignment, and speculation on X has centred on Meta (a claim reported, not confirmed — Innodata has said nothing, and no client has). Neither the concentration nor the suspected client can be verified from outside; what is verifiable is the at-will structure in the company’s own filing, which means whatever happened on Monday could happen again without notice. Analysis of the company’s results counted twelve consecutive quarters of year-over-year revenue growth to its name; the same at-will clause sits under all of them.
The visible edge of a business the AI story usually skips
Data labelling and AI-adjacent back-office work is the layer of the industry that touched New Zealand media twice this month — an AI firm’s hidden workforce across 31 countries was mapped by a student researcher, and Kenya’s 40,000 freelance AI writers watched the work evaporate ahead of everyone else. Monday was that pattern with share prices attached, and a couple of steps further up the stack: these were mid-seniority raters, quality experts and team leads, not gig-piece workers, and their exit was processed by system automation before human management caught up. The same firms building the evaluation infrastructure that AI safety debates lean on — Innodata itself opened a robot-training motion-capture lab this quarter, including for humanoid work — are also the firms where the labour terms are most exposed to a single client’s procurement decision.
For New Zealand the instructive part is the shape of our exposure: this year’s local AI employment debate has been almost entirely about white-collar adoption — the entry-level ladder AI thins out, the skills job ads now demand. That supply chain is offshore almost by default, which is why a 550-person mid-shift lockout registered on Nasdaq before it registered in the national conversation. If New Zealand wants AI-sector jobs that survive a client’s ramp-down, the honest read of Monday is that it would rather have them in firms with distributed customer bases and employment law nobody can outsource around.
❓ FAQ
How many people were laid off at Innodata? Reports say nearly 550 in India (26 October) with additional cuts in the Philippines. The company has not publicly confirmed a total; India Today reviewed a termination letter and interviewed affected staff.
Why did the stock fall so sharply? Innodata closed 25.3 percent lower on Tuesday (from $59.91 to $44.74), the worst session since February 2024. Investors priced in the risk around customer concentration — two customers supplied about 71 percent of Q2 revenue, per the company’s filings, and the layoffs were tied to a ramped-down client assignment.
Is this an AI-taking-jobs story? Not in the usual sense — the work was AI services (labelling, evaluation, safety testing) and the cuts followed a client ramp-down, per the termination letters. The story is exposure to single-client shifts, not automation replacing the roles.
Has Innodata commented? India Today reports the company had not responded to its request for comment by publication, and no public statement confirming the numbers had been issued at the time of writing. Q3 results are scheduled for 5 November 2026 (UTC), the likeliest first public accounting.
🔍 THE BOTTOM LINE
The AI boom’s visible jobs story is about who gets hired to use the models; this week’s reminder is about who builds them, and how thin a margin of notice they live on. Innodata’s mid-shift lockout processed 550 exits through device access rather than HR conversations — and the market’s 25 percent verdict was about customer concentration, not the people. The structural lesson travels: in AI services, the workers closest to the models are often the furthest from the contract, and that distance is now a business-model feature.
📰 Sources
- India Today (27 + 28 October 2026, two reports)
- Seeking Alpha via TradingView
- Cryptonews.net market analysis
