Hyundai spin-off Dtonic runs data hubs for most of Korea's smart cities. Its next test is Palantir's home turf

Before launching Dtonic, CEO Jeon Yong-joo worked at Toshiba, Hitachi and Hyundai Motor. The business began in 2014 inside the carmaker, spinning off from a Hyundai Motor in-house venture programme under the name Ilmile. Jeon, a division head in Hyundai's venture business team who had been involved since the founding, joined as research director, became chief executive in 2018 and renamed the company Dtonic.

"In a gin and tonic, gin is a relatively inexpensive liquor, but when mixed with tonic water, it becomes a high-value cocktail," Jeon told Korean startup outlet WOWTALE. "Similarly, Dtonic transforms what may seem like low-value data into highly valuable outcomes." The data in question is spatio-temporal: records of where things are and when, generated by vehicles, sensors, cameras, shoppers and city systems.

Twelve years on, the company had 96 employees as of March 2026 and reported 2025 revenue of about 33.9 billion won, roughly $24m. It has built and operated data hubs for metropolitan and provincial governments including Seoul, Busan, Ulsan, Gyeongnam, Chungnam and Chungbuk. "We have secured over an 80% market share in South Korea's municipal Smart City Data Hub projects, validating our technology's high-performance capabilities at a national scale," said Nora Lee, Senior Global Business Manager at Dtonic. The company is now attempting a harder transition, from a Korean public-sector data specialist into what it calls an AI OS: the execution layer beneath models and agents that keeps them running in factories, shops, cities and military settings.

A pandemic contact-tracing system proved the engine under pressure

Dtonic's core engine, Geo-Hiker, was built to collect, store, process and analyse varied forms of spatio-temporal data quickly, cutting both time and cost. Its most public test came in 2020, when the company took part in developing the Covid-19 epidemiological investigation support system for Korea's disease control agency and patented the related technology.

"Dtonic's platform was provided to the disease control agency during Covid-19 and cut epidemiological investigation time from 48 hours to five minutes," Jeon told the Electronic Times. According to WOWTALE, the same work had previously required 1,000 devices, which Dtonic reduced to 20. In practical terms, investigators could reconstruct an infected person's movements within minutes of a confirmed case, a step that had previously consumed two days while the virus kept moving.

"We will be the company that sells jeans in the AI gold rush," Jeon told the Electronic Times, describing a business built to supply the infrastructure others depend on. "Dtonic has recorded an average annual growth rate of 100% since its founding," he added. That record gave the company a national-scale reference most data infrastructure start-ups spend years chasing, and Jeon said he planned to open Geo-Hiker to other industries through an app store model.

Most enterprise AI still stalls between the demonstration and daily operations

"They pick one pain point, execute well, and partner smartly," said Aditya Challapally, lead author of MIT NANDA's 2025 report, describing the companies that succeed with AI. The study found that only about 5% of AI pilot programmes achieve rapid revenue acceleration, with the vast majority stalling and delivering little measurable impact on profit and loss. The research drew on 150 leadership interviews, a survey of 350 employees and an analysis of 300 public AI deployments.

"While many enterprises have invested heavily in AI, they often struggle to move beyond successful pilot projects because data, AI models and infrastructure remain fragmented," Lee said. Dtonic's argument is that a model performing well on a curated dataset breaks when it meets live sensor feeds, legacy databases and mixed hardware owned by separate teams. "This is what we call the 'PoC trap', where promising AI initiatives fail to scale into production," she added.

"Dtonic was built to solve this deployment gap by providing an execution environment that enables AI to operate reliably and scale across complex industrial environments," Lee said. The MIT data offers some support for Dtonic's position as an outside vendor: according to the study's figures, 67% of externally partnered deployments succeed, against 33% of internal builds. For a hospital operator, a city transport department or a retailer, the practical question is whether a vendor can keep an AI system stable across thousands of sites and years of changing data, and that is the claim Dtonic is asking buyers to test.

Two patents sit at the centre of Dtonic's production-grade claim

"To deploy AI in demanding industrial environments, we had to overcome high technical barriers," Lee said. "Specifically, we needed to efficiently orchestrate heterogeneous computing resources, such as GPUs and NPUs, and automate highly complex data and AI workflows." The company's answer rests on two patents.

The first, Ontology-Based Workflow Auto-Generation, addresses how AI understands an organisation's context. Dtonic's flagship platform, D.Hub, integrates large-scale data and uses ontology to structure field expertise and work context so AI can understand and use them. An ontology in this sense is a working model of the entities in an operation, such as a bus, a stop, a traffic signal, a shelf or a shift, and the relationships between them. Generating workflows from that model automates pipeline construction that would otherwise be rebuilt by hand for every deployment.

"These innovations automate complex AI deployment pipelines, enabling AI applications to operate consistently and efficiently across diverse infrastructure environments," Lee said of the second patent, Parallel Query Processing for Heterogeneous Accelerators, and its companion. Enterprises and governments increasingly run workloads across a mix of GPUs and neural processing units, often for reasons of cost, supply or sovereignty, and orchestrating queries across them efficiently is a hard engineering problem.

The company's intellectual property record predates the AI OS positioning. Dtonic won Korea's Excellent Patent Award for five consecutive years from 2019 to 2023, and Jeon received the Silver Tower Order of Industrial Service Merit in 2022.

Korean cities gave Dtonic scale, and a beauty retailer gave it a second business

The public sector remains the anchor. In June 2026, Dtonic was selected as final operator, alongside the cities of Cheonan and Asan and other AI companies, for the Ministry of Land, Infrastructure and Transport's AI Specialised Pilot City project, which carries total investment of 610.9 billion won. The company plans to standardise urban data for AI use, build an urban management system using agentic and physical AI, and run participating companies' AI services on a single platform.

"D.Eview combines electronic shelf labels with big data to provide functions such as stock clearance through real-time price changes and customer pattern analysis to leading retailers including LG Best Shop and Olive Young," Jeon told the Electronic Times. For store staff, that means clearance prices change on the shelf edge automatically, removing hours of manual relabelling. Dtonic's system integrates visitor movement and behaviour data with inputs such as weather and foot traffic, and the company is developing retail AI agents in stages.

"Building on our successful collaboration with CJ Olive Young, Korea's number one health and beauty platform, in the domestic market, we are now actively co-expanding into the North American retail market," Lee said. Dtonic also supplies D.Eview and AI-powered electronic shelf labels to retailers in Japan, the Middle East and Vietnam. "The reliability and data technology we proved in Korea are leading to successive business results and co-expansion in global markets including the US, Vietnam and the Middle East," Jeon said in August.

"Leveraging our proven enterprise-grade AX and DX references with major manufacturers and public sectors, we are continuously expanding the application of our AI OS to national critical infrastructure and high-value enterprise domains, including defence, manufacturing and energy," Lee said. In 2025, Dtonic was designated an AI Factory specialist company under a programme run by Korea's Ministry of Trade, Industry and Energy.

Palantir has shown the size of the prize and the height of the wall

"Demand for AI sovereignty has now been unleashed," Palantir chief executive Alex Karp said as the company reported second-quarter 2026 revenue of $1.935bn, up 93% year on year, with US commercial revenue of $764m, up 149%. Jeon has openly compared Dtonic to Palantir, and the comparison is instructive.

Palantir's growth is concentrated at home. Its international commercial revenue grew 26% to $182m in the same quarter, a far slower pace than its US business. That gap is where regional specialists see an opening: buyers outside the US who want sovereign control of their data, local references and support for domestic hardware. Ontology alone does not set Dtonic apart, since Palantir's Foundry is also built around one. Dtonic's case rests on real-time spatio-temporal processing, heterogeneous hardware orchestration and a cost structure suited to mid-sized cities and enterprises.

"Our core differentiation is in delivering the powerful execution layer that drives foundation models," Lee said, adding that integrating spatio-temporal processing, semantic data integration, workflows, agents and infrastructure in a single AI OS gives the platform its stability in production. "Whether it is smart cities, retail, manufacturing, defence, energy or mobility, enterprises can rapidly deploy, adapt and scale AI services across any physical domain using our single, unified platform," she said.

"We restructured Sales. We aligned cash outflows with cash inflows," C3.ai said in its latest results, after founder Thomas Siebel returned as chief executive. The enterprise AI software company reported fiscal 2026 revenue of $250.3m, down from $389.1m a year earlier, and a net loss of $470.4m. The lesson for platform companies is that a compelling architecture does not guarantee revenue; sales execution, services discipline and customer concentration decide outcomes.

Dtonic also faces pressure from below and beside. Cloud data platforms such as Microsoft Fabric, Databricks and Snowflake are moving up the stack into agents and orchestration, while Korea's large IT services groups compete hard for public digital transformation contracts. In smart cities, global engineering and networking groups have long sold integrated urban platforms. Dtonic's niche is narrower and deeper, which is both its defence and its ceiling.

The accounts show a company still converting projects into platform economics

Dtonic's 2024 results marked a turning point. Revenue rose about 81.9% to roughly 31.5 billion won, gross margin improved from 9.3% to 22.7%, and the company swung from an operating loss of about 6.1 billion won to an operating profit in the 300 million won range.

Growth then slowed sharply. Revenue for 2025 was 33.9 billion won, up 8%, with an operating loss of 2.43 billion won. A gross margin in the low twenties reflects a revenue mix that includes hardware and integration work, and it sits far below the software economics investors associate with platform businesses; Palantir's adjusted operating margin was 62% in its latest quarter.

"We operate a flexible hybrid business model that combines platform licensing, tailored integration projects and software as a service," Lee said, adding that the approach lets clients adopt AI in a way customised to their own infrastructure and business needs. The model helps Dtonic win complex public contracts, though the challenge is to keep integration work from defining the company.

Investors have nonetheless backed the platform story. Dtonic closed a $10m pre-Series A round from six venture firms in 2023, after revenue grew from $1.72m in 2020 to $10.9m in 2022. Hyosung Ventures, the corporate venture arm of Hyosung, invested in 2025. Later that year, the company raised about 6 billion won in pre-IPO funding from Korea Development Bank, IBK Capital and Aon Investment at a post-money valuation just under 200 billion won, roughly double its valuation from less than a year earlier. It filed for preliminary review for a KOSDAQ listing with NH Investment & Securities as lead manager, and Jeon was the largest shareholder at the end of 2024. The route is well trodden: technology special listings on KOSDAQ passed 300 companies this month and account for 62% of the market's capitalisation, with sectors diversifying from biotech into AI, semiconductors, defence and robotics.

Presight and Kuala Lumpur will test whether an urban data playbook travels

"We will link complex city and national data to better decision-making and execution," Jeon said in September, as Dtonic formed a strategic partnership with Presight, a subsidiary of Abu Dhabi AI group G42. The companies plan to apply Dtonic technology to Presight's existing projects and use the UAE as a base for expansion into Europe, the Middle East, North Africa, Central Asia and Africa.

"South Korea has a deep technology ecosystem developing advanced capabilities across AI, data and intelligent infrastructure," said Presight chief executive Thomas Pramotedham. The deal followed Dtonic's selection as one of 12 companies, from 376 applicants across 62 countries, for Presight's global AI accelerator cohort.

"Many Southeast Asian countries are accelerating their adoption of smart city technologies," Jeon said as Dtonic launched an Edge AI and data hub-based smart parking pilot in Kuala Lumpur with Korea's transport ministry and Kuala Lumpur City Hall. The project targets a daily frustration for drivers circling congested blocks. The company also won the AI & Data Award at the 2025 World Smart City Expo.

State backing remains both the company's strength and its exposure. Much of Dtonic's overseas momentum runs through Korean government programmes, and public buyers move on budget cycles. Palantir itself warns investors that large government customers are subject to uncertainties over budgets, spending levels and priorities that make the timing and size of contracts hard to predict. For Dtonic, the next two years will show whether the execution layer that runs Korean cities can win commercial customers abroad on its own terms, at margins that justify the valuation its investors have already paid.