Galaxy Robot Park attendance: 28,000 paid visitors, 1,000 a day — what the numbers show

Galaxy Corporation’s chief executive Choi Yong-ho said on September 21, 2026 that Galaxy Robot Park in Seoul has drawn about 28,000 paying visitors and up to 1,000 visitors a day. The same round of New York interviews produced a different cumulative figure of 20,000, a daily figure that one outlet limited to weekends, and a claim of robot-business revenue in the “tens of billions of won” — all company-reported, none audited.

Why it matters: the park is the main public evidence that Galaxy’s physical-AI strategy has paying customers, and it now underpins a planned US listing. The figures measure different things, and several cannot be reconciled with each other on the information published so far.

How many people have visited Galaxy Robot Park?

The best-supported current number is about 28,000 paying visitors, a cumulative figure Choi gave to Korean correspondents in New York on September 21, 2026. EDaily reported it on September 22 and Money Today repeated it on September 23, describing the park as having opened “last month” and the 28,000 as paid visitors “so far”. Neither report says whether the count starts at the July preview or at the August 21 full opening.

That baseline matters. Galaxy said on August 21 that more than 20,000 people had visited during the preview. If the 28,000 includes the preview, the park added roughly 8,000 paid visitors in its first month of full operation. If it excludes the preview, the park added 28,000 in about a month. Those are very different businesses, and the published reporting does not let a reader choose between them. Our earlier coverage of the August 21 opening already flagged that the preview figure was self-reported.

Hankyung’s two September 22 articles, from the same trip, put cumulative attendance at “over 20,000” — the preview number, unchanged. The simplest reading is that one outlet used an older company figure; it should not be treated as a second, lower measurement.

What does “1,000 visitors a day” actually mean?

It is a peak-day claim, not an average. NewsPim quoted Choi saying the Korean venue receives 1,000 visitors a day on weekends. Seoul Economic Daily quoted him saying that once tickets went on sale, “visitors come in at 1,000 a day”, without a weekend qualifier, and added that he had originally planned to offer the robot shows free.

The arithmetic shows why the qualifier matters. Between the August 21 opening and September 20 there were ten Saturdays and Sundays. At 1,000 a day, weekends alone would account for about 10,000 visits. That fits comfortably with 28,000 paid visitors since opening; it does not fit with an increase of only about 8,000 over the preview figure. At least one of the three numbers — 28,000, 20,000, or 1,000 a day — is therefore being used on a different basis from the others, or rounded generously.

As of September 28, 2026 the table below is the complete set of attendance and revenue figures we could verify against the original reports.

Figure	Metric type	Source	Date	Caveat
About 8,000	Reservations, initial July booking window	The Korea Herald / The Korea Times (company figures)	Jul 7–8, 2026	Bookings, not attendance
20,000+	Cumulative preview visitors	Opening coverage (company figures)	Aug 21, 2026	Preview period only; paid status not stated
20,000+	“Cumulative visitors”	Hankyung	Sep 22, 2026	Same number as pre-opening; likely not updated
About 28,000	Cumulative paid visitors	EDaily; Money Today (CEO)	Sep 22–23, 2026	Start date of count not stated
About 1,000 a day	Daily visitors	NewsPim (weekends); Seoul Economic Daily (unqualified)	Sep 22, 2026	Peak-day figure, not an average
“수백억 원”	Revenue already earned, physical-AI robot business	NewsPim (CEO)	Sep 22, 2026	Unaudited; period and scope not given
“수천억 원” possible	Future robot-business revenue	Hankyung (CEO)	Sep 22, 2026	Target; physical AI to pass management revenue “within two years”

Can the bookings, visits and paid-visit figures be added together?

No. The July 8,000 were reservations made before anyone walked through the gate; many of those people are inside the 20,000 preview figure. The 20,000 preview visitors may in turn be inside the 28,000 paid visitors. Adding them would count the same person up to three times. The July sell-out coverage measured demand for a limited number of weekend slots, which is a different question from how many people attend once capacity expands.

The comparable series, if Galaxy publishes one, would be monthly paid admissions from a fixed start date, split by ticket type. Nothing in the September reporting provides that.

How much revenue does Galaxy make from robots?

Galaxy has not published a robot-business revenue figure. NewsPim quoted Choi saying the physical-AI robot business is “already generating 수백억 원” in revenue — a Korean expression for an amount in the tens of billions of won, with no single number attached. The article gives no period, no definition of the segment and no audited basis.

Park admissions cannot be the bulk of it. Every ₩10,000 of average spend per visitor adds about ₩280 million across 28,000 paid visitors. Reaching even ₩10 billion from admissions and on-site spending would need an average of more than ₩350,000 per visitor. The larger number, if accurate, must come mostly from elsewhere — robot performances sold to events and brands, the New York activation, fashion work or hardware-related sales — and Galaxy has not broken those out.

For context, Galaxy reported group revenue of about ₩299 billion and operating profit of ₩12.5 billion for 2025, driven largely by G-Dragon. Our business-model case study sets out why group earnings say nothing about the park’s own margin. Hankyung’s interview adds a forward target: Choi said physical-AI revenue would overtake the management business “within two years”, and that “수천억 원” (hundreds of billions of won) is possible. That is a projection, not a result.

What would settle the question?

Three things: paid admissions by month from a stated start date; realised revenue per visitor; and a segment line for robot entertainment in audited accounts. A US listing would eventually force the third. Until then, the defensible statement is narrow: Galaxy reports tens of thousands of paying visitors in its first months and peak days near 1,000, which is evidence of consumer demand. It is not yet evidence of scale or profit. For how this format compares with robots deployed as staff inside an existing resort, see our analysis of the two robot theme-park models.