AI Phase Evolves Towards Agents and Physical AI, IT Services Welcome New Opportunities
AI summary card
AI Phase Evolves Towards Agents and Physical AI, IT Services Welcome New Opportunities
NVIDIA's Computex keynote emphasizes AI agents and physical AI, Nemotron model solves ROI issues, SIer business shifts to agent design and management, Fujitsu has advantages in the physical AI field.
- AI phase evolves from training → inference → agents, enterprises need to design dedicated agents with low cost and high efficiency.
- NVIDIA launches Nemotron 3 Ultra open-source model, improving enterprise AI return on investment.
- SIer business will shift from system development to enterprise AI agent design, management, and security.
- Physical AI (robots and autonomous driving) becomes the next stage, cooperation between Fujitsu and NVIDIA is worth watching.
- Enterprises need to optimize model usage, selecting models based on task difficulty and importance, creating new business for SIers.
Report interpretation
Overview
This research report is based on NVIDIA's Computex keynote speech, summarizing the key impact of AI development on the Japanese IT services industry. The report points out that with the transition to the Vera Rubin era and the acceleration of enterprise AI adoption, the AI phase is shifting from training to inference, and further to AI agents. NVIDIA emphasizes that enterprises need to design low-cost, high-efficiency dedicated agents, and has launched the Nemotron 3 Ultra open-source model to solve AI return on investment (ROI) issues. In addition, NVIDIA is strengthening its layout in physical AI (such as robots and autonomous driving), and business progress related to its partner Fujitsu is worth close attention.
Core views
AI Phase Evolution and IT Services Transformation As AI enters the agent phase, enterprises need to design dedicated agents to reduce costs and improve efficiency. The Nemotron 3 Ultra model launched by NVIDIA, as an efficient and low-cost open-source model, aims to help enterprises solve the problem of high AI Token costs and improve ROI. This means that the business focus of System Integrators (SIers) may shift from traditional system development to enterprise AI agent design, management, and security services. Model Usage Optimization and New SIer Opportunities As enterprises adopt AI agents, how to optimize model usage will become a key issue. Enterprises need to select appropriate models based on task difficulty and importance. This 'model usage optimization' work is expected to become an important business area for SIers in the future. The report mentions that software companies such as ServiceNow, Palantir, SAP, and CrowdStrike are building agent AI based on NVIDIA technology. Physical AI and Fujitsu's Potential NVIDIA views physical AI (robots and autonomous driving) as the next stage after agent AI. In physical AI, robots may use rule-based software to handle specific tasks (such as tax calculations), which is more efficient than repeated AI verification in certain scenarios. The report believes that due to Fujitsu's cooperative relationship with NVIDIA, as well as its accumulation in business software, data, and expertise, its attention may increase as physical AI advances.
Analysis framework
The report mainly sorts out the key information released by NVIDIA in the Computex speech, combines industry development trends, and deduces its impact path on the Japanese IT services industry. First, starting from NVIDIA's product strategy (such as the Nemotron series models), it analyzes how it solves cost and efficiency problems in enterprise AI deployment; second, based on the evolution logic of AI from training to inference to agents, it infers the potential transformation of the SIer business model; finally, combining the forward-looking layout of physical AI, it identifies the benefit logic of Japanese IT service providers (such as Fujitsu) cooperating with NVIDIA.
Methodology notes
Transmission of AI development stages (training → inference → agents) to downstream IT service demand
The report points out that the evolution path of AI technology (from training models to inference applications, and then to autonomous agents) will directly change the demand form of enterprises for IT services, promoting SIers to shift from traditional system development to higher-level AI agent design and operation services.
AI Return on Investment (ROI) is a key consideration for enterprise AI adoption
The report emphasizes that whether enterprises adopt AI depends on whether productivity improvements are sufficient to cover costs such as Tokens. NVIDIA's launch of the low-cost and efficient Nemotron model is precisely to solve this ROI constraint, thereby accelerating enterprise AI implementation.
Substitution relationship between rule-based software and AI verification in physical AI
The report mentions that in specific tasks (such as tax calculations), traditional rule-based software may be more efficient than letting AI repeatedly verify, indicating that physical AI does not completely replace traditional software, but collaborates or complements rule-based software according to task characteristics.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Fujitsu (6702.T)Benefits from NVIDIA's layout in the physical AI field, has business cooperation with it
- Strengths
- Cooperation with NVIDIA; possesses business-specific software, data, and expertise
- NVIDIA (NVDA)As an AI infrastructure and model provider, drives the industry towards agents and physical AI evolution
- Strengths
- Launched low-cost high-efficiency models such as Nemotron 3 Ultra; laid out models such as Cosmos 3 in the physical AI field
Key data
- NVIDIA New ModelNemotron 3 UltraOpen-source, low-cost high-efficiency model, aimed at improving enterprise AI return on investment
- NVIDIA Future Model PlanNemotron 4 and 5Will be released in subsequent stages, the report will continue to track its status as an enterprise AI model provider
- NVIDIA Physical AI Related ModelsCosmos 3Used to strengthen the layout of physical AI (robots, autonomous driving)
- Partner CompanyFujitsu (6702.T)Has business cooperation with NVIDIA, possesses advantages in business software, data, and expertise in the physical AI field
Impact & implications
The report believes that NVIDIA's promotion of AI agents and physical AI development has brought new growth opportunities to the Japanese IT services industry. SIers are expected to benefit from model usage optimization, enterprise AI agent design and management, and physical AI-related system integration. Among them, Fujitsu may receive more attention in the physical AI era due to its cooperation with NVIDIA and its accumulation in industry software and data.
What to watch
- NVIDIA's status as an enterprise AI model provider
- Release status of Nemotron 4 and 5
- Fujitsu's business progress in the physical AI field