Invest1 publisher3 min readPublished
Persona AI eyes Korean expansion into energy and construction as POSCO DX backs its $2 million shipbuilding, steel robot push
The Houston startup says energy and construction come next after shipbuilding and steel, though its Korean work so far amounts to two proof-of-concept projects and a $2 million investment from POSCO DX.
The Investor · Invest desk
What happened
- Persona AI, which builds humanoid robots for heavy industry, plans to widen its South Korean partnerships beyond shipbuilding and steelmaking into energy and construction, its business strategy chief said.
- In Korea the company is running proof-of-concept projects with HD Hyundai and POSCO DX, one on a welding robot for shipyards and one on a robot built for steel logistics work.
- POSCO DX also invested $2 million in Persona AI to support that work.
- A third-generation robot is due to reach the market this fall, after the first- and second-generation models, and the company is preparing plans for mass production and maintenance services.
- Perry put the projected United States shortfall at more than 330,000 welders by 2028, which is the labour gap the company is selling against.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint Even at PW Consulting's 2030 market size, a year of global humanoid output covers about 2 percent of the 2.1 million US manufacturing jobs estimated unfilled that year, so a buyer cannot justify the spend on the labour gap alone.
- exposure POSCO DX sits on both sides of the table as trial partner and shareholder, so the first price for a steel-logistics humanoid in Korea gets negotiated with an investor.
- decision Persona AI has to fund a hardware launch, mass-production planning and a maintenance business while courting energy and construction buyers; a company founded in 2024 is now spread across four heavy industries.
- contradiction The IFR counts last year's humanoid deployments as mainly manufacturing pilots, while Persona AI develops on the view that the machines take complex floor work inside two years.
Divide the $900 million PW Consulting puts on last year's industrial humanoid market [10] by the roughly 7,000 industrial and professional humanoids the International Federation of Robotics counted as sold in 2025 [9] and the implied average is about $129,000 a unit [11]. The two numbers come from different houses and do not cover identical scopes, so treat the result as approximate. Hold it anyway. PW Consulting's $5.5 billion for 2030 [10] then works out to roughly 43,000 units a year [12], and the sixfold climb is about 44 percent a year compounded [22]. Covering the US welder gap Perry cites would take about 47 times the entire 2025 global shipment [13].
POSCO DX's check is worth about fifteen robots at that implied price [14]. The report does not give a stake or a valuation. Persona AI was founded in Houston in 2024 [2].
"Korea is a country with many opportunities when it comes to humanoid robots," Michael Perry, the company's chief of business strategy, told The Seoul Economic Daily at COEX in Seoul on the 22nd [8].
Welding is where the case is strongest, and Perry said why: "In Korea, too, there are not enough workers to replace aging welders, but welding is a field where automation know-how and data have accumulated" [6]. The job is wider than the torch. Welders at shipyards align steel plates and manage cables before they weld, and they inspect the weld afterwards, and Persona AI's development goal is for its robots to handle those tasks as well [7].
Perry's argument against Chinese suppliers is about data and service. "Companies are reluctant to use Chinese robots when they do not know where their data is going," he said [15]. He also said Chinese makers "push customers to buy replacements rather than offering solutions, and they do not analyze what the problem is" [16].
If the labour shortage bites on schedule, HD Hyundai and POSCO DX turn the trials into unit orders around the third-generation launch [17], and 43,000 units in 2030 looks light. The likelier version, or rather the one this record supports, is that the proofs of concept renew: a yard that has run one welding cell for a year has learned what it wanted and can wait for a fourth generation. The third path is a price fight, in which the data and service argument has to be worth a premium a steel company's procurement department will sign.
In my view this is still a technology programme funded by an interested customer. Perry's phrasing for the new sectors is that the company sees "energy and construction as areas for collaboration" [21], and two proofs of concept plus one corporate investment is what collaboration looks like on a balance sheet. A named purchase order, with a unit count and a delivery date, would settle it the other way.
What to watch
- Whether the third-generation launch this fall arrives with a named unit count and price from HD Hyundai or POSCO DX.
- Whether the energy and construction talk produces a named partner in either sector, or stays a stated ambition.
- Whether the IFR's next annual count moves industrial humanoid deployments out of the pilot category.