NVIDIA Corporation (NVDA.US) GTC conference outlook: Can the AI leader maintain its dominance, with the market closely watching the new strategy of the "post-training era"?

With the annual NVIDIA Corporation (NVDA.US) GTC developer conference set to kick off next week, this event, known as the "AI field annual barometer", is hotter and more important than ever this year. When CEO Jensen Huang steps into the packed hockey arena next Monday (March 16 local time), the global investors' eyes will be focused on the cards he plays to deal with the increasingly fierce market competition and consolidate NVIDIA's position as a leader in AI chips. The four-day GTC conference is not only a stage for NVIDIA Corporation to showcase its latest developments in chip, data center, software platform CUDA, AI agents, and physical AI fields such as Siasun Robot & Automation, but also a crucial test of the company's strategic direction. After delivering better-than-expected financial reports that failed to drive a significant increase in stock prices, investors are eager to be assured that NVIDIA Corporation's strategy of profiting back to the AI ecosystem is taking effect. Market research firm eMarketer analyst Jacob Bourne said, "I expect NVIDIA Corporation to showcase updates from Rubin to Feynman in their full-stack roadmap, while emphasizing inference, intelligent agents, Network-1 Technologies, Inc., and AI factory infrastructure." The focus of the competition in the "post-training era": Inference chips As the AI industry transitions from the "training" phase of large models to the "inference" phase where AI agents shuttle through tasks in applications, the competitive landscape in the market is undergoing profound changes. Although NVIDIA Corporation currently holds over 90% of the market share in the training and inference markets, analysts generally believe that its market share loss is inevitable, especially in the inference field. Sid Sheth, founder and CEO of inference chip startup d-Matrix, said that while NVIDIA Corporation will still maintain its dominance in the training field, "inference is a completely different thing." He added that CUDA, as the core software that supports most AI training and locks developers into its ecosystem, has a weaker "moat" in the inference field. Developers can turn to competitors outside of NVIDIA Corporation because running completed AI models does not require complex programming like training them. To address this trend, NVIDIA Corporation is expected to launch new products optimized for inference workloads at the conference. There are reports that an inference chip integrating the technology of AI startup Groq, acquired for $1.7 billion last December, is expected to be unveiled, aiming to provide fast and cost-effective inference computing capabilities. Groq's ultra-fast AI technology will be integrated into NVIDIA Corporation's extensive CUDA ecosystem to solidify its software moat. Potential threats and NVIDIA Corporation's "defense works" However, challenges remain severe. On one hand, core customers of NVIDIA Corporation, including OpenAI and Meta (META.US), have started to develop their own chips, with Meta even explicitly stating that they will release a new AI chip every six months. The rise of Application-Specific Integrated Circuits (ASICs) is seen as a long-term threat to NVIDIA Corporation's General-Purpose Graphics Processing Units (GPUs), as these chips customized for specific functions demonstrate higher efficiency advantages in inference scenarios. Kin Ngai Chan, Managing Director of Summit Insights Group, stated that compared to a year ago, NVIDIA Corporation will undoubtedly face more intense market competition, and it is expected that by 2027, as enterprise-developed ASIC chips are scaled up, NVIDIA Corporation's market share will decline, especially in the inference chip market. To address the challenge, NVIDIA Corporation is taking multiple measures to strengthen its defense. In addition to acquiring Groq, the company recently invested $2 billion each in optical communication companies Lumentum (LITE.US) and Coherent (COHR.US) to promote the application of Co-Packaged Optics (CPO) technology. This technology uses light instead of electrical signals to transmit data between chips, potentially significantly improving the connection efficiency and reducing power consumption of large-scale data centers. Sebastien Naji, research analyst at William Blair, expects CPO to be the core breakthrough direction of the next generation Feynman chip architecture. Bourne from eMarketer added that NVIDIA Corporation is likely to position CPO technology as a key for efficiently connecting large-scale AI clusters at GTC, but currently, the scalability of this technology cannot match the shipping volume of NVIDIA Corporation chips. The cost and feasibility of its large-scale deployment will also be the focus of investors. On the other hand, the Central Processing Units (CPUs) that have long been dominated by Intel Corporation (INTC.US) and AMD (AMD.US) are regaining their position in AI tasks. William McGonigle, analyst at Third Bridge, pointed out that with the rise of agent-based AI, the "agent orchestration layer" managed by CPUs is becoming a new performance bottleneck. Therefore, the analyst expects NVIDIA Corporation to showcase server products that use only its CPU to respond to this new trend. AI agents and Siasun Robot & Automation: Driving the next wave of growth In addition to hardware competition, the market is also concerned about whether the prospects of AI applications can support sustained demand for computing power. Jensen Huang has emphasized that intelligent agents will be the next important DRIVE for inference demand. Sheth of d-Matrix said that as the potential of speech, video, and multimodal AI agents is gradually unleashed, this field is expected to bring about a new wave of inference computing. The technology from Siasun Robot & Automation is seen as another layer of growth space. Daniel Newman, CEO of The Futurum Group, pointed out that NVIDIA Corporation reported about $6 billion in revenue related to Siasun Robot & Automation last quarter and predicted that the development schedule for humanoid Siasun Robot & Automation will be very "aggressive". This suggests that physical AI may become a reality faster than expected. GEO Group Inc politics: Damocles' sword hanging over the chip giant's head In addition to technological competition, political factors from GEO Group Inc are increasingly becoming a key variable affecting the future of NVIDIA Corporation. With the United States considering further expanding restrictions on the export of AI chips, and access to key markets such as China restricted, NVIDIA Corporation's global sales map is being reshaped. According to reports, after completely cooling down in the Chinese market, NVIDIA Corporation has stopped the production of H200 chips and shifted capacity to the next-generation Rubin platform. In this context, large-scale AI infrastructure investments in the Middle East countries such as Saudi Arabia and the United Arab Emirates hold significant meaning for NVIDIA Corporation. However, factors such as regional conflicts, energy costs, and the speed of data center construction add uncertainty to the demand in these emerging markets.