Grepped, a new data broker, claims a $4M monthly run rate selling company data to AI labs

What was sold

Grepped is a data-brokerage startup. According to a Dealroom report from October 15, 2026, sellers list proprietary internal records such as chat logs, codebases, documents and accounting exports from tools like Salesforce and QuickBooks. Grepped de-identifies and packages the data for AI labs to buy, and says it pays sellers within seven days of a sale.¹

This entry is different from the other cases on this tracker. Grepped is an intermediary, not a company selling its own archive. The figures in the Dealroom report are about the broker's business and not about a single deal.

How the deal came together

The report is based on job listings that Grepped founder Erik Newsham posted on X on October 13, 2026. Newsham said he entered the business two weeks earlier and had reached a $4 million monthly run rate, roughly $48 million annualized.¹ The listings include a VP of Semiconductors at $4 million to $9 million a year plus equity, and neural data and chip design data leads at $1.5 million a year each.¹

Newsham named accounting, legal, chemistry and semiconductors as hot areas and called the space "the most lucrative space I think has ever existed."¹ He has also published a five-step playbook for cloning the brokerage model.¹

For context on how other buyers and brokers describe this market, micro1 announced on October 9, 2026 a plan to spend $1 billion over 12 months buying and licensing operational data from companies, financed by Citi and Hercules Capital. Its program page lists payment tiers of more than $100,000 for a qualified partnership, more than $500,000 for large datasets or ongoing work across teams, and more than $1 million for highly unique proprietary data.² Those are published program tiers and not completed deals. SimpleClosure, a wind-down firm, reports typical payouts of $10,000 to $100,000 per shut-down company.³

What went wrong or drew pushback

The main pushback is about the numbers. Dealroom reports that some commentators on X questioned the $4 million monthly run rate because the business is only weeks old.¹ The report does not describe any funding round or investors, and it does not name any sellers or buyers.¹ The run rate is a founder's own statement and cannot be verified from public sources.

There are also open questions that apply to any broker model. The report says Grepped de-identifies the data but does not describe how, who audits it, or what contract terms sellers and buyers sign.¹ The wider debate about selling workplace data, including employee consent and who receives the proceeds, is described in MIT Technology Review's reporting on shut-down startups.³

The job listings themselves drew attention for their hours, listed as 9am to 9pm, seven days a week, and for an application question: "what does humanity know that it doesn't know it knows?"¹

What it means for companies with data

Grepped's claim, if accurate, would mean demand for company data is strong enough to support a fast-growing broker within weeks. The evidence for that is thin. A single report, built on the founder's own posts, is not enough to treat the $48 million annualized figure as a market indicator.

What the report does support is the shape of the market. Brokers are positioning themselves as the point where messy internal records become packaged, de-identified data for labs. For a company with data, the practical questions are who does the de-identification, whether the company keeps ownership of the underlying records, how much of the sale price the broker keeps, and whether the buyer may resell. The Dealroom report does not answer those questions for Grepped.

Compare with micro1's stated terms, under which companies keep ownership of their underlying data, micro1 removes personal information "when relevant" and companies can review samples before use.² Those terms are published; Grepped's are not in the sources reviewed.

See /how-it-works/ and /resources/where-to-sell/ for how to evaluate intermediaries.