Exclusive: Ex-Meta AI Director discloses $10M financing for Noosphere Labs

RUNTIMEWIRE INVESTIGATION — Exclusive

Original reporting by RuntimeWire.

Why it matters

Noosphere Labs has secured eight-figure backing before revealing a product, funding Carlberg's attempt to move physical AI research from Meta and Sandia into a startup.

RuntimeWire is first to report that Noosphere Labs has raised approximately $10.25 million, according to an SEC filing submitted Sunday.

Kevin Carlberg, a former Meta AI research director who spent years developing faster physics simulations, has raised about $10.25 million for Noosphere Labs, a new physical AI startup that has yet to publicly define its product.

A Form D filed with the Securities and Exchange Commission on August 10th shows that Noosphere Labs sold $10,249,995 of a planned $10,499,991 equity offering. The Kirkland, Washington-based corporation reported 10 investors, all accredited, and said the total includes the conversion of a simple agreement for future equity, or SAFE.

The filing names Carlberg as chief executive, executive officer and director. It also lists Tyler Simpson and Trilogy Equity Partners Managing Director Amy McCullough as directors, giving Noosphere Labs a board that combines physical AI research, startup operating experience and early-stage investing.

An eight-figure round with a SAFE conversion

Noosphere Labs recorded the offering's first sale on July 24th and filed the notice on August 10th. The filing leaves $249,996 available in the offering and reports no sales commissions or finder's fees.

The SAFE conversion matters when reading the headline number. Form D reports the value of securities sold, rather than a clean measure of new cash deposited in connection with the current closing. Because Noosphere Labs included converted SAFE securities in both its offering size and amount sold, some portion of the $10.25 million represents earlier capital that has now converted into equity.

The filing does not break out that portion. It also does not name the 10 investors or provide a valuation. Form D notices generally offer only a narrow view of privately placed securities, leaving ownership, preferred-stock terms and board rights outside the document.

Noosphere Labs conducted the sale under Rule 506(b), an exemption commonly used for private offerings to accredited investors. It declined to state its revenue range and classified itself under the SEC's broad "Other Technology" category. The filing says Noosphere Labs was incorporated in Delaware in 2026, making the financing an unusually large initial disclosure for a corporation formed this year.

Carlberg is building around physical intelligence

Carlberg's own account supplies the clearest description of the business. His personal site identifies him as founder and CEO of a stealth AI startup focused on "physical AI and real-world intelligence." The Noosphere Labs filing uses the same executive title and identifies him as its sole executive officer.

Carlberg previously spent more than five years at Meta, where he says he became director of AI research science and led a team spanning Reality Labs Research and Fundamental AI Research. That group worked on physical AI and simulation technology for wearable computers and virtual- and mixed-reality devices.

Before Meta, Carlberg spent more than eight years at Sandia National Laboratories. His work there centered on model reduction, a collection of techniques used to make computationally expensive physics models run faster while preserving the behavior that matters to a particular analysis. Carlberg says his team applied those methods to near-real-time simulations for national-security uses.

That background points to a technically demanding company thesis: building AI systems that reason about physical environments, rather than limiting the product to language or other digital inputs. Carlberg's public description does not narrow Noosphere Labs to robotics, wearables, simulation software or another specific application. His research history spans all three of the core ingredients such a product could require: machine learning, computational physics and systems designed to operate under real-world constraints.

Carlberg earned a doctorate and master's degree in aeronautics and astronautics from Stanford University after studying mechanical engineering at Washington University in St. Louis. He also holds an affiliate associate professorship in applied mathematics and mechanical engineering at the University of Washington.

A board built for research and commercialization

Tyler Simpson, listed as a Noosphere Labs director, brings experience converting enterprise software into an exit. Simpson was a co-founder and chief technology officer of Luum, a Seattle developer of commute-management software. HealthEquity acquired Luum in 2021, adding the product and its roughly 50 employees to HealthEquity's commuter-benefits operation.

Simpson had previously worked at Microsoft, according to GeekWire's coverage of the acquisition. His presence gives Noosphere Labs an operator who has already carried a Seattle software business from formation through a strategic sale, a different skill set from Carlberg's research-heavy career.

The third director, Amy McCullough, is a managing director at Bellevue, Washington-based Trilogy Equity Partners. McCullough has been part of Trilogy's investing team since 2007 and previously worked in JPMorgan Chase's investment bank and Microsoft's treasury department.

The filing identifies McCullough personally as a director and does not assign the board seat to Trilogy or identify the firm among the offering's investors. Her appointment nonetheless puts a longtime Seattle seed investor inside the boardroom as Carlberg and Simpson move from research and company formation toward a commercial product.

Noosphere Labs is still presenting itself as a stealth operation, but the financing gives Carlberg substantial room to recruit specialized researchers and engineers before a public launch. Physical AI companies typically face costs that pure software startups can defer, including simulation infrastructure, proprietary datasets, sensors, hardware integration and testing in environments where model errors carry consequences outside a browser.

The Form D establishes the scale and structure of the financing: $10.25 million sold, 10 accredited investors, an equity offering and a SAFE conversion. Carlberg's record supplies the direction. He is using an eight-figure pool of private capital to turn more than a decade of work in computational physics and physical AI into a standalone company.