License-based revenue expanding, led by Data for AI; build-out in developed markets complete, shifting to harvest phase. Progress also seen in validating demand for AI-native data

Summary

1Q FY3/27 revenue came in at JPY 1,147M (-21.3% YoY), with adjusted EBITDA of -JPY 667M. License-based revenue reached JPY 645M (+10%), surpassing a year-ago quarter that included a large deal, driven mainly by additional corporate license orders for AI applications (Data for AI) from a major overseas semiconductor manufacturer. This underscores continued progress in shifting the revenue mix toward licensing. The revenue decline reflects a payback effect in project-based work following completion of new build-out in North America. Management left full-year guidance unchanged (revenue of JPY 7B, adjusted EBITDA of +JPY 50M). Foundations for the Data for AI business also advanced, including the release of a dataset on Hugging Face and partnerships with NVIDIA and MathWorks.

Key Points (Results Takeaways and Growth Initiatives)

• Management Strategy and Market View
  • The physical AI market is projected to grow at roughly 30% CAGR; the company aims to transition beyond an autonomous-driving-related business into a data infrastructure company.
  • Management frames AI development around three elements — compute resources, model development, and data — and positions NVIDIA as complementary rather than competitive.
  • Even for AI applications, real-world data reflecting region-specific road environments and traffic rules is required; global coverage of 1.8M km is the source of the company's one-stop capability.
  • The WP.29 international standard agreed in June 2026 (expected to take effect around January 2027) is viewed as a medium-to-long-term tailwind that should lift demand for ODD management and simulation-based validation.
• Current Business Progress and Drivers
  • License-based revenue rose +10% YoY on expansion in corporate license deals centered on Data for AI, advancing the shift in revenue structure. The overall decline reflects payback in automotive project-based work following completion of new build-out in North America.
  • Cost of goods sold rose roughly JPY 400M (higher depreciation on North American data-related assets plus FX effects), and SG&A increased about JPY 70M. The new build-out phase is largely complete, and the company now moves into a harvest phase driven by license revenue growth.
  • Roughly 70% of the 249 employees are engineers, with a Japan-US development structure maintaining technological advantage.
  • Vehicle models equipped with the company's data increased by one to a total of 39 models across 6 OEMs; disclosure permission for model names is expected after the Obon holiday.
• Key Strategic Initiatives and Inflection Points
  • Data for AI supply to a major overseas semiconductor manufacturer has expanded from North America into Europe, South Korea, and Japan, and is being leveraged as a reference case for new customer acquisition.
  • An AI-native dataset was released on Hugging Face, generating roughly 4,000 downloads with sales discussions and data evaluation processes underway.
  • The Tokyo Ryutsu Center premises were converted into high-precision 3D data and provided to companies participating in the Heiwajima Autonomous Driving Council — a showcase for area-specific deployment.
  • Rikanosu was acquired in April 2026 as the second deal in the surveying company roll-up; downstream M&A is also under consideration.

Outlook and Strategy

• Full-year guidance unchanged (revenue of JPY 7,000M, license-based revenue of JPY 3,000M, adjusted EBITDA of +JPY 50M). 1Q progress stands at 16.4% (our estimate), but as in the prior year the revenue structure is weighted to 2H.
• The new build-out phase in developed markets is largely complete, shifting to a data supply and update phase. Management expects gross margin and EBITDA margin improvement alongside revenue growth.
• Over the medium to long term, the company targets 10-30% revenue growth centered on license-based business with gross margins above 80%.
• AI-native data is expected to progress from R&D and ADAS development use cases into evaluation applications, and subsequently into mass-production and in-vehicle applications — positioned as a medium-to-long-term growth driver.
• M&A will continue with successive acquisitions of profitable surveying companies, targeting both upstream (survey analysis, data collection, 3D modeling) and downstream (GIS, simulation, AR/VR, 3DCG).
• The Middle East is viewed as a promising market on the back of government-led smart city investment, with a new build-out project for a 27th country currently underway. Emerging markets will be selected based on market potential and investment returns.

Positive Factors

• License-based revenue of JPY 645M (+10%) exceeded a year-ago quarter that included a large deal, confirming the ongoing shift in revenue structure.
• Additional orders for European, South Korean, and Japanese data followed positive evaluation of North American data usage, advancing monetization of global data assets.
• Roughly 4,000 downloads and increased inquiries following the Hugging Face release, with sales discussions underway with automakers, Tier 1 suppliers, autonomous driving developers, and robotics companies.
• 3DGS generation using both point cloud data and HD maps delivers real-world-scale geometric accuracy while removing moving objects, differentiating from image-only generation methods.
• Launch of bridge and tunnel management solutions for North American transportation authorities, with participation in the Esri partner network expanding DOT-facing channels.
• 3Dmapspocket® has been adopted by Tokyo Tatemono and other major real estate players, and was released free of charge for areas affected by the Kumamoto earthquake, broadening social-use applications.

Concerns and Risks

• Adjusted EBITDA of -JPY 667M represents a JPY 692M deterioration YoY. Achieving the full-year target of +JPY 50M requires substantial earnings improvement over the remaining nine months, with the focus on license revenue expansion and the timing and scale of pipeline deals.
• Depreciation of JPY 293M (+JPY 80M) pushed up costs, resulting in a gross loss of -JPY 297M (vs. +JPY 436M a year ago).
• Risk that order and acceptance timing fluctuates with customers' development plans and investment decisions.
• While adoption of 3D data licenses in infrastructure management and real estate development is expanding, revenue contribution remains at an early stage (company self-assessment: △).
• FX is running above the JPY 145/USD assumption, but the profit impact is limited given offsetting overseas costs.
• Heightened geopolitical risk and rising crude oil prices are recognized as business environment risks affecting the automotive industry.

Performance Highlights

1Q FY3/27 revenue was JPY 1,147M (-JPY 311M), with an operating loss of -JPY 999M and adjusted EBITDA of -JPY 667M. Revenue declined on payback in automotive project-based work following completion of new build-out in North America, while license-based revenue increased, led by corporate licenses. The adjusted EBITDA loss landed within budget, and full-year guidance was left unchanged.

Segment Results

• Domestic Revenue: JPY 142M (-62.2% YoY)
• Overseas Revenue: JPY 1,005M (-7.2% YoY)
• Gross Profit: -JPY 297M (vs. JPY 436M a year ago)
• SG&A: JPY 702M (approx. +JPY 70M YoY)
• Depreciation: JPY 293M (+37.6% YoY)
• Net Loss Attributable to Owners of Parent Company: -JPY 980M (vs. -JPY 285M a year ago)
• Cash and Deposits: JPY 4,253M (+JPY 595M vs. end-March 2026)
• Vehicle Models Equipped with Company Data: 39 models across 6 OEMs
• High-Precision 3D Data Coverage: 1.8M km
  ※Some YoY figures are our estimates

Q&A List

• Q: We understand international standards for Level 3 and Level 4 autonomous driving were agreed at WP.29 in June 2026. Could you explain the expected impact of this agreement on your business?
  A: We view this WP.29 agreement as a critically important milestone toward the social implementation of autonomous driving. This regulatory framework should elevate the importance of ODD management — defining under what conditions and where a vehicle may operate and when autonomous mode may be engaged — as well as safety assessment and post-deployment monitoring. In particular, the use of simulation for proving autonomous driving safety is set to expand, which should drive greater demand for evaluation and validation using the high-precision 3D data we presented today. Internally, we expect use cases for ODD management and simulation-based evaluation to expand alongside the regulatory build-out. Our high-precision 3D data can be applied not only to in-vehicle licensing but also to safety assessment and ODD management — what we call Data for AI — which could expand our business opportunities over the medium to long term. That said, the regulatory framework itself will not immediately have a material impact on revenue or profit; rather, we see it as supporting the future expansion of the autonomous driving market and, in turn, our business growth.
• Q: Could you introduce, on your website or via video, the vehicle models equipped with your high-precision 3D data and models planned for adoption over the next one to two years?
  A: From our regular dialogue, we recognize that where our high-precision 3D data is actually used in vehicles and services is a topic of high interest to investors. We intend to communicate as much as possible through briefings, press releases, and social media. However, please understand that contractual constraints with automakers and their suppliers may prevent disclosure of specific models or future adoption plans. Future adoption plans in particular are tied to our customers' product plans and announcement timing, so the information we can disclose is limited. For example, while we have said that adoption in a 39th model has been decided, we cannot yet name the specific model as disclosure permission has not been granted. That said, we are progressing through the process and expect to obtain permission around after the Obon holiday, so we ask for your patience. Within the scope of what we can disclose, we will continue to communicate adoption cases, use cases, and installation track records as clearly as possible via our website and briefing materials.
• Q: You announced participation in an overseas partner network. Is your North American subsidiary actively pursuing sales to target municipalities and road administrators? We assume the 3Dmapspocket data supply to Tokyo Tatemono was won through the Japan headquarters' sales efforts — are similar efforts underway at the North American subsidiary? We also hope to see Japanese initiatives such as snow removal support and airport-related work replicated in North America.
  A: We understand this question concerns our participation in the partner network of Esri, the world's largest GIS software company, headquartered in California. We are actively pursuing sales to municipalities and US state departments of transportation (DOTs) in North America through our subsidiary DMP North America. We have already conducted proposal activities with multiple DOTs and, as updated in today's pipeline disclosure, are currently receiving inquiries for proof-of-concept projects and progressing several sales discussions. We are also beginning to see order wins materialize, such as the recently announced bridge and tunnel management solution for North American transportation authorities. Regarding Esri, we joined the partner network to expand sales channels for infrastructure management in the public sector and for infrastructure operators. We position this as a key initiative in deploying high-precision 3D data originally built for autonomous driving into fields such as infrastructure management and urban planning. DMP North America conducts sales activities for road infrastructure management and public-sector applications in addition to autonomous driving data supply, and the Esri partnership is part of that effort. On snow removal support and airport-related initiatives, our policy is to build a track record in Japan and then expand overseas. We believe market demand in the US is strong for these as well, and we intend to pursue overseas replication aggressively.
• Q: Please introduce any specific business plans outside of autonomous driving — real estate, construction, drones, gaming applications, and so on.
  A: At present, automotive-related business accounts for the majority of revenue. However, by leveraging the data assets built for autonomous driving, we aim to expand our business domains over the medium to long term into real estate, construction and infrastructure, logistics, AI, and entertainment, as well as automation at airports, ports, factories, and logistics facilities.
• Q: Is 1Q adjusted EBITDA tracking in line with plan? And based on 1Q results, is there any need to revise full-year guidance?
  A: 1Q adjusted EBITDA declined YoY.
  However, versus the internal plan, the adjusted EBITDA loss was within expectations — in fact, better than assumed. License-based revenue exceeded the year-ago level on expansion in corporate license deals centered on AI applications, and the shift toward a license-centric model continues to progress.
  Our corporate license and project deals tend to be weighted toward revenue recognition in the second half of the fiscal year.
  At this point, there is no downside in the deal progress underpinning our full-year guidance, and we have left guidance unchanged. We will continue to carefully assess deal progress and disclose appropriately.
• Q: What drove the YoY increases in cost of goods sold and SG&A?
  A: Cost of goods sold rose JPY 400M and SG&A rose JPY 70M, both YoY.
  The increase in cost of goods sold was driven mainly by higher depreciation on high-precision 3D data-related assets in the North American business, plus FX effects.
  We have been building out high-precision 3D data globally, including in North America, and depreciation has risen alongside the growth in data assets. The bulk of the YoY gross profit decline reflects depreciation on past investments.
  That said, the new build-out phase, centered on developed markets, is largely complete, and we are now transitioning to a supply and update phase. Over the medium to long term, we are shifting toward a business structure in which the completion of depreciation translates into improved profitability.
  On SG&A, the JPY 70M increase was driven mainly by M&A-related expenses, new business development expenses, and FX effects. In particular, M&A costs related to building out the surveying company network and expenses related to new business development including Data for AI are not simply cost increases — they also strengthen our future data build-out and update capabilities, the utilization of global data assets, and the business foundation for license revenue expansion.
  For the full year, SG&A is not expected to increase versus the prior year, and we will continue to focus on cost control.
• Q: What do you see as the current downside risks to achieving full-year guidance? What points warrant particular attention, such as deal slippage or the ramp-up of Data for AI demand?
  A: The largest downside risk to full-year guidance would be a change in the external environment — for example, a slower-than-expected ramp in the autonomous driving market, or a recession dampening customers' appetite to invest. In addition, since autonomous driving and Data for AI projects tend to be large in scale, customer-side development plans and investment decisions are also a factor. Taken together, the biggest risk is variability in order timing, acceptance timing, revenue recognition timing, and deal size across our projects, centered on Data for AI. While we recognize these risks, based on the current deal status and progress we do not see a need to revise full-year guidance at this point.
• Q: Full-year guidance assumes JPY 145/USD, but the actual rate is running above that. Since you are maintaining guidance despite this, are there factors you are viewing cautiously beyond the FX tailwind? Please share your FX sensitivity and the potential for upside.
  A: Our guidance for this fiscal year was formulated at the start of the year assuming JPY 145/USD. The current FX level is running above that initial assumption, which is a positive for overseas revenue. In terms of the profit impact, however, costs corresponding to overseas revenue are also incurred overseas, so revenue and costs offset each other and the FX impact is limited. We do not disclose FX sensitivity, but given that a high proportion of our revenue is overseas while much of our overseas business cost base is also denominated in local currency, we believe FX risk on a profit basis is hedged to some degree relative to the revenue impact. As for why we are maintaining guidance: alongside the FX tailwind, we are factoring in business environment uncertainty, including the timing of project-based order wins and the ramp-up timing of certain license deals. For large overseas deals in particular, revenue recognition timing can shift with customer decision-making, so we are assessing the situation cautiously at this point. In addition, FX markets can move sharply over short periods, and we recognize that volatility has been elevated recently. We have no intention of revising guidance solely on the basis of temporary FX movements. We will consider deal progress and FX effects comprehensively and disclose appropriately if we determine a guidance revision is warranted.
• Q: We see AI-native data as a growth area. When do you expect it to begin contributing meaningfully to earnings? Please clarify whether this is a near-term revenue contributor or a medium-to-long-term growth driver.
  A: We expect AI-native data to be used first for R&D purposes and ADAS development, then adopted for evaluation purposes, and subsequently — once vehicles based on it come to market — to expand into mass-production and in-vehicle applications. This fiscal year, we are concentrating resources on generating AI-native data and pitching it to customers. Releasing the data on Hugging Face has allowed audiences worldwide to see it, and customer proposals and sales discussions are progressing steadily. We have confirmed that concrete demand exists. In addition, as explained today, we have begun partnering with major players in this space such as NVIDIA. We intend to grow this business in line with the market ramp and expect it to become an important growth driver over the medium to long term.
• Q: You are partnering with NVIDIA and MathWorks to supply AI-native data, but if NVIDIA enters this space in earnest, would DMP become unnecessary? Where exactly does DMP add value?
  A: We see NVIDIA as complementary rather than competitive.
  In physical AI development, including autonomous driving, three elements are critical: compute resources, model development, and data.
  NVIDIA's business is GPUs — the compute resource for AI — recently branded as "AI factories." Compute providers build open ecosystems with the aim of increasing GPU consumption and AI adoption.
  DMP, by contrast, holds the world's largest volume of high-precision 3D data reproducing real-world roads, including HD maps, point clouds, imagery, and semantic information.
  Because physical AI presumes that AI operates in the real world, data accurately reflecting real-world road environments, traffic rules, signage, lane markings, and road structures is essential.
  Our ability to partner with NVIDIA and MathWorks is one example demonstrating that our data assets and technical know-how are valuable within the AI development ecosystem.
  The remaining element is model development. Our intent is a division of roles in which we provide the real-world data foundation within NVIDIA's compute platform and ecosystem, while companies worldwide advance model development. Of course, training a model is not sufficient on its own — validation, evaluation, and certification are required, and each of these requires data.
  This is not limited to NVIDIA; other companies are advancing physical AI development, including autonomous driving, along the same lines. We believe that the more AI advances, the more important high-quality real-world data becomes, and we aim to deliver value as the real-world data foundation for the AI era.
• Q: Partnerships with NVIDIA and others are progressing in AI-native data generation. Who do you see as competitors in this area? And how do you view the risk that major tech companies build similar datasets in-house?
  A: First, we do not view NVIDIA as a competitor at all — the relationship is complementary. In physical AI development, including autonomous driving, we see three critical elements: compute resources, model development, and data. NVIDIA's business is GPUs, the compute resource for AI, which they now call AI factories. Compute providers form open ecosystems designed for ease of use, with the aim of increasing GPU consumption and AI adoption. We, on the other hand, hold the world's largest volume of 3D data that reproduces real-world roads with high precision — the high-precision 3D data generated from it, raw point cloud data, imagery, and semantic information layered on top. Because physical AI presumes AI operates in the real world, it must accurately reflect and understand real-world road environments, traffic rules, signage, lane markings, and road structures. We present our current partnerships with NVIDIA and MathWorks today as one example demonstrating that our data assets and technical know-how are valuable within the AI development ecosystem. The only remaining element is model development, and we are proceeding with the intent that companies worldwide advance model development under a division of roles in which we supply real-world data within the AI factory, compute platform, and ecosystem NVIDIA provides. Of course, training a model is not enough — validation, evaluation, and certification are required, and each requires data. This is not limited to NVIDIA; with various companies providing compute resources and pursuing model development, we believe our position — holding high-precision digitized real-world data — allows us to supply data broadly without dependence on any single platform. Our data is therefore valuable for AI training and evaluation. Building on that, over the long term we aim to become a data infrastructure company underpinning society and industry, as a corporate group supporting the so-called cyber-physical loop: collecting data from the real world, performing analysis, prediction, and control, and feeding results back into the real world.
• Q: Please explain the earnings impact of the Kumamoto earthquake.
  A: Following the Kumamoto earthquake of July 2026, we have confirmed that road geometry and other features have changed in certain areas, creating discrepancies with our high-precision 3D map data.
  We have been assessing conditions in the affected areas and considering our response since immediately after the event. At this point, while data update costs may be incurred, the scope and costs cannot be reasonably estimated, so the earnings impact is undetermined.
  Should we determine that there is a material impact on earnings, we will disclose it appropriately.
• Q: What kind of company does DMP aim to become in the physical AI era?
  A: We aim to establish a position as the platformer of real-world data underpinning physical AI. In the physical AI era, beyond the AI model itself, the source of competitive advantage will be what real-world data is used for training, validation, and operation. As we demonstrated today, our data is valuable for AI training, evaluation, validation, and simulation. Building on that, over the long term we aim to become a data infrastructure company indispensable to society and industry — a corporate group supporting the cyber-physical loop of collecting data from the real world, performing analysis, prediction, and control, and feeding results back into the real world.