46 investors put $65.4M into Persona AI inside a 17-day selling window

A Form D reports securities already sold. The $65.36 million is what had actually been bought when the filing went in on September 14th, and the first sale is dated August 28th ^[1][2]. So 46 buyers closed inside a 17-day window ^[3][2]. In my view a close that fast means the allocations were agreed before the selling window opened.

Against the target, $65.36 million is 46 percent ^[1]. The unsold $76.6 million is the other 54 percent ^[2][4]. Split across 46 investors, the average position is about $1.42 million ^[3]. The filing does not name the investors, identify a lead or disclose the valuation or price per security ^[5]. It is filed under Rule 506(b) and covers equity, options or warrants, and the securities issued when those rights are exercised ^[4], so one line carries more than one instrument class.

None of that measures whether the welding works. Persona AI's own site describes its technologies as conceptual designs and prototypes under development, not yet commercially available ^[8], and the company expects initial pre-production deliveries around the end of 2026 ^[9]. The shipyard schedule comes from HD Hyundai, which said Persona AI would work with HD Korea Shipbuilding and Offshore Engineering, HD Hyundai Robotics and Vazil Company on humanoid welding systems, with prototypes by the end of 2026, then field testing, then commercial deployment starting in 2027 ^[10].

The welding-first choice is defensible on engineering grounds. Persona AI's argument is that an expensive bipedal machine is easier to justify when it carries specialized tools through a facility already built around human movement, and that shipyards are full of ladders, hatches, catwalks and work areas that change ^[14]. The target task list is welding, grinding, inspection, fabrication, equipment maintenance and material handling ^[13].

ABS is involved on safety and classification requirements, and Persona AI has described projects with POSCO and SSE Steel Fabrication in Louisiana ^[11]. A separate early-stage collaboration with Under Armour is testing textiles and external materials under heat, friction, abrasion and repetitive motion ^[12]. Those are the loads a welding cell puts on a housing, and they are cheaper to characterise in a materials lab than on a hull.

The founding group's record is in getting machines out of the lab. Radford spent roughly 15 years at NASA's Johnson Space Center on Valkyrie, Robonaut and other robotics programs, then founded Houston Mechatronics, which became Nauticus Robotics ^[15]. Pratt led walking-robot research at the Florida Institute for Human and Machine Cognition for about two decades, placed second in the 2015 DARPA Robotics Challenge and was CTO of Figure AI ^[16]. Akinyode worked on robotics at NASA before joining Nauticus, where he helped move the Aquanaut underwater robot from concept toward production ^[17]. Radford and Pratt discussed starting a humanoid business as early as 2008 and held off, according to a 2024 IEEE Spectrum interview, because the machines were hard to operate autonomously and hard to commercialize; faster perception systems and stronger industrial demand changed that ^[18].

The filing sizes the balance sheet behind the next 15 months of hardware, and only the part already sold. The May 2025 pre-seed was $27 million, co-led by Unity Growth and Tides Ventures ^[6][7]. The amount sold this time is 2.4 times it ^[6].