GDEP ADVANCE,Inc. Q2 FY2025 earnings call

Core Financial Performance

• 2nd quarter cumulative net sales: 3.995 billion yen, up 111.8% year-over-year, reaching 69.2% of the full-year forecast.
• Operating profit: 498.54 million yen, up 90.6% year-over-year, reaching 73.8% of the full-year forecast.
• Ordinary profit: 497.15 million yen, up 101.8% year-over-year; net profit: 344.13 million yen, up 101.3% year-over-year.
• Gross profit margin: 17.9%, down 4.6 percentage points year-over-year, with the decline already factored into initial forecasts and no expected material impact on full-year results.
• Selling, general and administrative (SG&A) expenses: 216.29 million yen, up 32.4% year-over-year, driven by personnel cost increases (up 18.2% YoY from headcount expansion), Tokyo office depreciation, and higher insurance fees for large projects.

Key Growth Drivers

• The main growth driver is increased demand for GPUs driven by generative AI and large-scale AI development, combined with growing average project size.
• Large-scale projects drove strong revenue growth in the reporting quarter, with quarterly revenue volatility increasing as project size and acceptance timing shift from the historical trend of Q3 being the strongest quarter.

Strategic Priorities

• Aligns with the mid-term management plan to deliver 30% top-line growth this fiscal year, with all incremental profits invested aggressively in talent and infrastructure.
• The company prioritizes the automotive, healthcare, and media & entertainment industries, which have high compatibility with generative AI due to their large existing big data holdings.
• Growth in cloud vendor GPU adoption is expected to create a positive cycle that stimulates demand for the company's core on-premise and private cloud offerings.

By service segment: 1. DX Services: Revenue grew more than 120% year-over-year. 2. Service&Support: Revenue grew more than 20% year-over-year. By customer industry segment: Sales to cloud vendors related to Japan's Ministry of Economy, Trade and Industry (METI) cloud program grew sharply. These sales are systems for domestic cloud service providers investing in GPU server infrastructure using METI program subsidies. No specific absolute revenue values or percentage contribution breakdowns for each segment are provided in the transcript.

• Full-year financial guidance and dividend guidance maintained from the original published plan, with no upward or downward revision.
• The company plans to prepare internal testing and customer trial environments for NVIDIA's latest Blackwell GPUs, which began global shipments at the start of 2025, to generate new sales leads for future fiscal years.
• The company will continue aggressive talent acquisition investment in the second half of the fiscal year, which will increase fixed operating costs.
• Management expects LLM and large-scale text/audio AI applications to drive demand over the next 2-3 years, followed by a gradual shift to robotics, multi-modal AI, and virtual simulation use cases over the next 10 years.

• Yen depreciation and exchange rate volatility increases procurement costs, which the company may have to partially absorb, and can negatively impact customer purchase intent by pushing projects outside customer budget limits.
• Larger average project size leads to longer lead times and lower gross margins, as large projects require higher-level corporate approval that often results in unbundling of hardware and services, reducing the company's higher-margin service contribution.
• Any shipment delays for Blackwell generation GPUs could create uncertainty for future project pipelines, though management currently expects no material near-term impact as long as DGX B200 shipments remain on schedule.

Q: How do exchange rate changes and larger project sizes separately impact gross margin, and did the low-margin METI cloud program drive most of the gross margin decline? / A: Larger projects change customer approval processes, often leading to unbundling of hardware and expertise services, which creates downward margin pressure. Yen weakness directly raises procurement costs that may need internal absorption and can reduce customer purchase appetite. The company does not have specific quantitative data on each factor's relative weight. Any large project exceeding a certain size has an inherent downward trend in gross margin, but expanding top-line revenue still drives sufficient operating profit growth. The METI cloud program projects had low margins, but they were pursued to accumulate internal expertise and build external credibility, which management views as a valuable long-term opportunity.

Q: Has the expected positive spillover from the public cloud program to on-premise and private cloud demand begun generating customer inquiries? / A: The company has already received several inquiries for on-premise and private cloud projects spurred by the growth in public cloud GPU adoption. Inquiries come from a range of sources including private companies, academia, and research institutions, concentrated in generative AI and LLM-related use cases.

Q: What is the impact of reported GB200 GPU shipment delays on the company's pipeline, and is DGX B200 the main Blackwell product the company is importing? / A: The company plans to import DGX B200 units, and has not independently confirmed the reported GB200 shipment issues related to heat dissipation. Management expects no material near-term impact from any GB200 delays, as most customers currently prioritize DGX B200 for their evaluations, with GB200 only used for niche specialized use cases.

Q: Which industries currently have the strongest generative AI demand, and when will manufacturing demand for AI-enabled robotics pick up? / A: The company prioritizes automotive and healthcare for their large big data holdings, with growing generative AI demand also coming from retail for chatbot customer support and media & entertainment. Manufacturing demand exists for material development and early-stage trials of AI robotics simulation (via NVIDIA Isaac Sim and Omniverse), but broader manufacturing demand for these use cases is still 2-3 years out, with the next 2-3 years focused on LLM-driven applications.