AI Summit Seoul Opens, Exhibitors Jump from 71 to 118

Looking purely at the numbers, this year's event feels like a different show from last year's. Exhibitors number 118 companies from seven countries, filling 307 booths. Last year's event drew 71 companies and 197 booths, meaning exhibitor count grew 66.2%.

From 71 to 118 companies — what happened in a year

The bigger jump came on the visitor side. Pre-registrations totaled 23,063 — roughly five times last year's 4,452. Organizers expect more than 25,000 total attendees once on-site registrations are counted.

The show floor is divided into four zones: industry-specific AI, enterprise AI solutions, generative AI and content, and hands-on experience areas. New this year is a dedicated humanoid robot demonstration zone.

Exhibitors include Adobe, Oracle, HP, Seagate, and Genspark. On the conference stage, speakers come from Google, Microsoft, and NASA's Jet Propulsion Laboratory (JPL).

The exhibition floor layout follows these same four categories — starting in sections grouped by industry such as finance, manufacturing, and healthcare, moving through enterprise business solutions and generative AI content tools, and ending in the hands-on experience zone.

The star of the show floor has changed — from models to bodies

What stands out in this year's exhibitor lineup is robotics. AgiBot, RealMan Robotics, and Pudu Robotics brought humanoid and service robots. The floor also features hands-on demos, including one that analyzes sports movements to correct posture and another that recognizes taekwondo forms.

"Physical AI" — a term for the shift in which models that once generated text in a screen now gain cameras and limbs to move through the physical world. If the center of gravity at Korean AI events over the past year or two was large language models, this year the first thing visitors encounter walking through COEX's aisles is those models given bodies, standing upright.

The same shift is visible on stage. The conference features 69 sessions across six tracks: AI Mega Trends, AI Transformation, AI and Data, Vertical Industry Use Cases, AI and Robotics, and Agents/Agentic AI. Two of the six tracks are devoted to robots and agents.

This year's theme is "what comes after adoption"

The event's stated theme is "The Age of Innovation: Beyond Adoption, the Rise of the Enterprise Agent." Rather than a forum debating whether to adopt AI, it's designed for companies that have already adopted it and are now asking what comes next.

The speaker lineup reflects this direction, including Steve Chien of NASA JPL, Professor Larry Heck of Georgia Tech, Haipeng Poon of Microsoft Research, Maxime Labonne of Liquid AI, and Maxim Afanasiev of Google Cloud. Topics span autonomous systems for space exploration, conversational agents, medical and scientific AI, and small models.

200 consultations — what this event is really selling

Sandwiched between the conference and the exhibition is "AI Business Connecting," where 80 companies took part in roughly 200 one-on-one consultations.

The side programming leans the same way. On August 20, an AI startup forum and investment pitch session will pair early-stage companies with investors. Practitioner workshops are built around immediately applicable topics like overseas marketing and workflow automation, while a recruitment hall for AI talent and a one-on-one expert consultation lounge sit on the same floor.

This part best captures the nature of the event. Trade shows are usually about browsing new products, but what's really being traded here isn't products — it's meetings. Companies that build AI and companies looking for places to apply it exchange business cards on the same floor.

Jung Heecheol, head of KITA's Overseas Marketing Division, said, "For Korea to become a key player in the global AI supply chain, we need a marketplace where technology companies and industry demand-side companies can actually meet, collaborate, and build business together."

Editor's take

Articles about event scale usually stop at the exhibitor count. But the more telling number this time isn't the 66.2% rise in exhibitors — it's the roughly fivefold jump in pre-registrations. The gap between the two numbers means demand grew faster than supply. Far more people showed up to buy than showed up to sell.

Having attended Korean AI events over the past two years, the tenor of the questions has kept shifting. Through 2024, it was "what is this." Last year, it was "should we adopt this too." That this year's event baked "beyond adoption" directly into its theme signals that organizers judged the question has moved on to "we adopted it — why isn't it working." Indeed, half of the six conference tracks are devoted to industry use cases, data, and agent operations.

The shift in technology's center of gravity shows up just as clearly in the exhibition layout. Language models are now background; robots and agents are front and center. If you compare models to a car engine, the past few years were about competing on engine performance. This year's show floor looks more like a marketplace for where to install that engine and who will service it.

That's also why so many of the 307 booths aren't pitching "our model is better" but rather "here's how it plugs into your process."

For Korean companies, the practical value lies less in the exhibition floor itself and more at the consultation tables beside it. Eighty companies and roughly 200 meetings may not sound like much, but considering that the longest step in AI adoption is often "finding a vendor who understands our problem," it reads differently. Sitting down for 30 minutes with someone who knows your process beats watching ten demos.

Placing the startup pitch session and recruitment hall on the same floor on August 20 reflects the same calculation. Put technology, money, and people on one floor, and a three-day exhibition keeps working long after it closes. That's exactly what this event is betting on.