Industry Wire · Updated weekly

Annotation Radar

What's moving in data annotation and AI training data — acquisitions, funding, and research from across the industry, each with a short read from the H2L Robotics India team.

Last updated 16 Nov 2026 · 19 stories · RSS

Market 📌 Pinned 9 Nov 2026

Sama, the AI data-annotation firm, is winding down and liquidating

H2L takeA letter to creditors blames falling revenue from competition, made worse by class-action litigation. For anyone buying annotation, the lesson is supplier risk: how a vendor treats and pays its annotators is not a side issue — it can end the vendor. Stable, salaried in-house teams are a real advantage here.

TIME Sama Meta OpenAI
Funding 📌 Pinned 19 Aug 2026

Nvidia in talks to back Mercor at a $20bn valuation

H2L takeThe July talks now have a name attached, and it changes the story. If it closes, Mercor passes Scale AI to become the most valuable company in AI data — on a business that barely existed three years ago. Two things sit under the number. Nvidia is reportedly already a customer, paying Mercor to source human-expert data for its Nemotron models, so this is a buyer investing in its own supplier. And the reported $614m first-half revenue leans heavily on a handful of frontier labs — a concentration most annotation businesses would not survive. The money is chasing expert judgement, not labelling volume.

MarketScreener, on reporting by The Information Mercor Nvidia Scale AI General Catalyst
Policy 12 Nov 2026

MeitY examines home-services data collection for physical-AI training

H2L takeIndia’s IT ministry is looking at in-home recording pilots used to train robots. Consent and data handling will decide who is allowed to collect real-world data in India — the same questions any annotation workflow on sensitive images has to answer before the work starts.

Inc42 Pronto Snabbit Human Archive
Market 7 Nov 2026

iMerit’s Radha Basu on how AI is reshaping India’s data-annotation jobs

H2L takeThe direction is clear from India’s largest players: simple labelling is being automated, and the value moves to annotators who understand the domain. That is the bet we make with crop-disease work — people who know what a sick plant looks like, not just how to draw a polygon.

Invezz iMerit
Policy 27 Oct 2026

noyb files GDPR complaints against Scale AI over tracking of contractors

H2L takeFive complaints in four EU countries about tracking software and ID checks on Outlier contributors; Scale calls the claims unfounded, and no regulator has ruled. Either way, how annotators are monitored is becoming a compliance question, not just an HR one.

The Register Scale AI noyb
Product 26 Oct 2026

iMerit launches Scholars, a platform for managed domain-expert work

H2L takeA managed service rather than a marketplace: iMerit runs the projects and the quality checks itself. Expert annotation only works with that kind of oversight — which is why consensus and review sit at the centre of how we work.

SiliconANGLE iMerit EXL
Funding 7 Oct 2026

Mecka raises $60M Series B led by Sequoia for robot training data

H2L takePhysical AI needs its own data: people recording everyday tasks so robots can learn how humans move. Field robots are no different — the camera on a weeding machine is only as good as the labelled images behind it.

Tech Funding News Mecka Sequoia Nvidia
Market 26 Aug 2026

Amazon closes Mechanical Turk on 30 September after 21 years

H2L takeThe platform that defined crowdsourced labelling is gone. The work did not disappear — it moved to managed teams and expert networks that can be held to a quality standard. Micro-task crowds were cheap; trained annotators with consensus and review are what models now need.

TechRepublic Amazon AWS
Market 20 Aug 2026

Micro1 reaches a $500m gross run rate, up from $7m in a year

H2L takeThe number that matters is not the run rate but the margin: Micro1 keeps roughly 60–70% of gross, because most of it is paid straight out to contracted doctors, lawyers and engineers. That is a marketplace, not a studio — it scales by recruiting faster, and it owns no method. It also confirms the market is big enough for several players at once rather than consolidating around one, which is the more useful read for anyone specialising.

TechCrunch Micro1 Mercor Handshake
Policy 18 Aug 2026

Ex-staff describe “hostile” conditions for the labellers behind Mercor

H2L takeReported alongside the funding talks, and worth reading beside them: former employees describe excessive hours and targets across a contracted workforce of around 30,000, and the company says performance-linked bonuses are industry standard. Both things can be true, and that is the point — a marketplace of contractors and a salaried studio are not the same supply chain, and buyers who care where their training data comes from are starting to ask which one they are buying.

Business & Human Rights Resource Centre Mercor
Acquisition 3 Aug 2026

EXL closes the iMerit acquisition

H2L takeThe deal we flagged in June has completed, and founder Radha Basu joins EXL's executive committee rather than leaving after an earnout — which suggests the buyer wants the operating knowledge, not just the delivery capacity. Consolidation in this sector is now a fact rather than a rumour, and the independents left are fewer and more specialised.

EXL newsroom EXL iMerit
Market 30 Jul 2026

Scale AI appoints Francis deSouza as chief executive

H2L takeAn enterprise-software operator from Google Cloud replaces an interim chief, more than a year after founder Alexandr Wang left for Meta. The hire reads as a deliberate turn away from the founder-led, lab-facing era toward selling to ordinary enterprises — which is a different business with different buyers, and it opens room underneath for vendors who still do the specialist work.

Scale AI (company announcement) Scale AI Google Cloud Meta
Funding 9 Jul 2026

Mercor in talks to raise $500M at a $20 billion valuation

H2L takeThe AI-training-data race keeps accelerating — Mercor would double its valuation in under a year, after crossing a reported $2B annualized revenue in June. The money is chasing human expertise in the training loop, not just bulk labeling. Precision, domain-specific annotation is where the margin is moving.

Forbes Mercor
Acquisition 24 Jun 2026

EXL to acquire iMerit for up to $310 million

H2L takeThe biggest annotation-industry consolidation of the year: EXL pays $170M upfront plus up to $140M in earnouts to fold in iMerit's Ango platform and its expert 'Scholars' network for RLHF and model evaluation. It signals that enterprise-AI buyers now want annotation, domain experts and delivery under one roof — good news for specialist studios who can plug into that stack rather than compete with it.

GlobeNewswire (EXL press release) iMerit EXL Read our full take →
Market 17 Jun 2026

Meta redeploys 6,500 engineers to label AI training data

H2L takeEven a hyperscaler with unlimited compute hit a wall on data quality — and had to put expensive engineers on labeling. The lesson for the whole industry: models are now bottlenecked on high-quality human-labeled data, not GPUs. That's a structural tailwind for professional annotation.

TechTimes Meta
Research 1 Jun 2026

AI data-labeling market projected to reach $6.5B by 2031

H2L takeAnalysts peg the annotation market at ~$2.3B in 2026, nearly tripling by 2031. The growth is real, but it rewards vendors who move up the value chain into specialized, high-accuracy work — which is exactly where a focused agricultural-CV studio wins.

HeroHunt (industry overview)
Policy 6 May 2026

Pentagon awards Meta-backed Scale AI a $500 million contract

H2L takeA fivefold jump on last year's $100M defense deal. Government and defense are becoming serious annotation buyers, and both demand exactly what agricultural and industrial CV work demands: auditable quality, traceable provenance, and vetted labelers.

Forbes Scale AI
Funding 26 Feb 2026

Encord raises $60M Series C to build the data layer for physical AI

H2L takeLed by Wellington at a $550M valuation, with revenue from physical-AI customers (robots, drones, autonomous vehicles) up 10x. This is squarely our lane — annotation for machines that act in the real world, not just chatbots. Validates the agricultural-robotics annotation thesis H2L is built on.

SiliconANGLE Encord

No stories match that filter.

Radar links out to original reporting; summaries and analysis are H2L Robotics India's own. Spotted industry news we should cover? Tell us ↗