# Nvidia Talks to Back Mercor at $20B Valuation, Per One Unnamed-Source Report By Ray Termsheet · 2026-08-20 · Deals & Funding · https://datacommenter.com/nvidia-talks-to-back-mercor-at-20b-valuation-per-one-unnamed-source-report/ About the author: Follows the money in the data economy: venture rounds, M&A, and strategic stakes across data vendors and infrastructure. > Nvidia is reportedly weighing a stake in AI data-labeling firm Mercor that would double its valuation to $20 billion from October's $10 billion Series C, per an August 19, 2026… Original reporting: [finance.biggo.com](https://finance.biggo.com/news/fba8b58b-2e5f-4d0a-9f2f-d89a7c0de481) _AI-assisted commentary, editorially reviewed. Quoted excerpts belong to the original outlet._ Nvidia is in talks to join a Mercor funding round that would value the AI data-labeling startup at roughly $20 billion, doubling its $10 billion Series C from October, according to The Information’s August 19, 2026 report. That’s the headline number every outlet is running with, but it traces back to a single set of unnamed sources — nobody named in this story, including Nvidia or Mercor, has confirmed it on record. What’s independently establishable is narrower and, frankly, more interesting: Mercor’s own disclosed revenue trajectory. The company reported $614 million in gross revenue for the first half of 2026 and an annualized run rate that crossed $2 billion by June, and Nvidia paid it “tens of millions” last quarter for expert-curated data feeding the Nemotron open-source models, per BigGo Finance and TradingView’s reporting. Those are business-performance numbers a company can substantiate with its own books, unlike a valuation figure that depends on who signs a term sheet. > A $20 billion valuation quoted by four outlets citing the same unnamed sources isn’t corroboration — it’s an echo. The precise size of Nvidia’s contribution, the total round size, and whether General Catalyst actually closes as lead investor all remain undetermined, according to the source reporting itself. VarIndia’s coverage adds a risk factor worth tracking: roughly 91% of Mercor’s first-half revenue reportedly came from foundation-model makers, meaning customer concentration could bite if OpenAI, Google DeepMind or Anthropic bring more labeling work in-house. PYMNTS also flags pending class-action suits over a Mercor data breach involving contractor biometric data — the kind of liability overhang that due diligence in a live round would need to price. Context matters here too: Nvidia’s $18.6 billion in private investments last fiscal quarter, its 2024 stake in Scale AI at a $14 billion valuation, and rival Handshake’s acquihire of Cleanlab all point to capital piling into the human-data layer as the new AI bottleneck. What would actually validate the Mercor number: a completed term sheet, an SEC or PitchBook filing showing the round closed, or on-record confirmation from either company. Until then, this is a well-sourced rumor with real revenue numbers attached, not a done deal. > Nvidia Corp. (NVDA) is in discussions to invest in Mercor, an artificial intelligence data supplier and labeling provider, in a funding round that would value the three-year-old startup at roughly $20 billion, according to a report from The Information citing people familiar with the matter. > — [finance.biggo.com](https://finance.biggo.com/news/fba8b58b-2e5f-4d0a-9f2f-d89a7c0de481) [Read the full story at finance.biggo.com →](https://finance.biggo.com/news/fba8b58b-2e5f-4d0a-9f2f-d89a7c0de481) --- Cite this analysis: https://datacommenter.com/nvidia-talks-to-back-mercor-at-20b-valuation-per-one-unnamed-source-report/ Cite primary facts: https://finance.biggo.com/news/fba8b58b-2e5f-4d0a-9f2f-d89a7c0de481 Need the underlying datasets (alt data, market data, AI training data)? Source licensed vendors via Brickroad: https://brickroad.network More machine-readable access: https://datacommenter.com/llms.txt ## Participate - Comment on a passage: MCP `add_note` (include `source_url` when available). - Suggest an editorially reviewed correction: MCP `suggest_edit`. - Open factual questions: none.