As the annual NVIDIA GTC Developers Conference is set to kick off next week, this event, hailed as the "annual barometer of the AI field," has reached new heights in terms of热度 and significance this year.
According to Zhitong Finance, as the annual $NVIDIA (NVDA.US)$ GTC Developer Conference is set to kick off next week, this event, hailed as the 'Annual Bellwether of the AI Field,' has reached new heights in terms of热度 and importance this year. When CEO Jensen Huang walks into a packed ice hockey arena on Monday (local time, March 16), all eyes from global investors will be on what strategies he unveils to address intensifying market competition and solidify NVIDIA's position as the leader in artificial intelligence (AI) chips.
This four-day GTC conference is not only a stage for NVIDIA to showcase its latest advancements in chips, data centers, the CUDA software platform, AI agents, and physical AI like robotics, but also a critical test of the company’s strategic direction. After delivering better-than-expected earnings that failed to drive a significant rise in stock price, investors are eager for reassurance that NVIDIA’s strategy of reinvesting profits into the AI ecosystem is paying off.
Jacob Bourne, an analyst at market research firm eMarketer, stated: "I expect NVIDIA to present an updated full-stack roadmap from Rubin to Feynman, with a particular emphasis on inference, agent AI, networking technologies, and AI factory infrastructure."
The Focal Point of Competition in the 'Post-Training Era': Inference Chips
As the AI industry accelerates its transition from the 'training' phase of large models to the 'inference' phase where AI agents execute tasks across applications, the competitive landscape is undergoing profound changes. Although NVIDIA currently holds over 90% of the market share in both training and inference, analysts widely believe that the erosion of its market share is inevitable, particularly within the inference domain.
Sid Sheth, founder and CEO of d-Matrix, a startup specializing in inference chips, stated that while NVIDIA will maintain its dominance in the training space, "inference is an entirely different story." He added that CUDA, NVIDIA’s core software that underpins most AI training and locks developers into its ecosystem, plays a weaker 'moat' role in the inference realm. Developers can turn to competitors outside NVIDIA because running completed AI models does not require the complex programming needed during training.
To counter this trend, NVIDIA is expected to unveil new products optimized specifically for inference workloads at the conference. Reports suggest that an inference chip incorporating technology from Groq, the AI startup acquired by NVIDIA in December for $1.7 billion, may debut. This chip aims to deliver rapid and cost-effective inference computing capabilities. Groq’s ultra-fast AI technology will be integrated into NVIDIA’s vast CUDA ecosystem to further reinforce its software moat.
Potential Threats and NVIDIA’s 'Defensive Fortifications'
However, challenges remain formidable. On one hand, key NVIDIA clients such as OpenAI and $Meta Platforms (META.US)$ have initiated the development of their own chips. Meta Platforms, in particular, has explicitly stated it will release a new AI chip every six months. The rise of Application-Specific Integrated Circuits (ASICs) poses a long-term threat to NVIDIA’s general-purpose Graphics Processing Units (GPUs). These chips, tailored for specific functions, demonstrate significant efficiency advantages in inference scenarios.
KinNgai Chan, Managing Director at Summit Insights Group, stated that compared to a year ago, NVIDIA will undoubtedly face more intense market competition. It is projected that by 2027, as companies achieve scaled deployment of their self-developed ASIC chips, NVIDIA’s market share will decline, particularly in the inference chip market.
To counter these challenges, NVIDIA is taking a multi-pronged approach to strengthen its defenses. In addition to acquiring Groq, the company recently invested $2 billion each in optical communications firms $Lumentum (LITE.US)$ and $Coherent (COHR.US)$ , aiming to advance the application of Co-Packaged Optics (CPO) technology. This technology uses light instead of electrical signals to transfer data between chips, which is expected to significantly improve connectivity efficiency in hyperscale data centers while reducing power consumption. William Blair research analyst Sebastien Naji predicts that CPO will be a core breakthrough area for the next-generation Feynman chip architecture.
eMarketer’s Bourne added that NVIDIA is likely to position CPO technology as a key solution for efficiently connecting large-scale AI clusters at GTC. However, the current production scale of this technology still cannot match the shipment volume of NVIDIA's chips, and the cost and feasibility of its scaled deployment will also be a focus for investors.
On the other hand, central processing units (CPUs), which have long been dominated by Intel (INTC.US) and AMD (AMD.US), are regaining prominence in AI tasks. Third Bridge analyst William McGonigle noted 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 to showcase server products that rely solely on its CPUs to respond to this emerging trend.
AI Agents and Robotics: Driving the Next Wave of Growth
Beyond competition at the hardware level, the market is also closely watching whether the prospects of AI applications can sustain ongoing demand for computing power. Jensen Huang previously emphasized that agent-based AI will become the next significant driver of inference demand. Sheth from d-Matrix stated that as the potential of voice, video, and multimodal AI agents gradually unfolds, this field is expected to usher in a new wave of inference computing.
Robotics is seen as another layer of growth opportunity. Daniel Newman, CEO of The Futurum Group, pointed out that NVIDIA reported approximately $6 billion in robotics-related revenue last quarter and predicted an extremely “aggressive” timeline for the development of humanoid robots. This suggests that physical AI may become a reality sooner than anticipated.
Geopolitics: The Sword of Damocles Hanging Over Chip Giants
Beyond technological competition, geopolitical factors are increasingly becoming a critical variable affecting NVIDIA’s future. As the U.S. considers further expanding export restrictions on AI chips and access to key markets such as China becomes limited, NVIDIA’s global sales landscape is being reshaped. Reports indicate that after facing a complete downturn in the Chinese market, NVIDIA has ceased production of the H200 chip and shifted its capacity to the next-generation Rubin platform.
In this context, massive investments in AI infrastructure in countries like Saudi Arabia and the UAE in the Middle East hold significant importance for NVIDIA. However, factors such as regional conflicts, energy costs, and the pace of data center construction add uncertainty to the demand in these emerging markets.
Editor/Stephen
