From training to inference, NVIDIA's barriers are further strengthened! Analysts: It's hard to imagine how competitors can compete

According to Zhitong Finance APP, at NVIDIA's (NVDA.US) annual technology conference GTC, the chip giant showcased numerous innovative products and strategic collaborations. Wall Street analysts believe that the launch of these new products will further strengthen NVIDIA's position in the AI inference market.

Bernstein maintains an "Outperform" rating on NVIDIA with a target price of $300.

The analyst team led by Stacy Rasgon stated: "We are impressed with NVIDIA's development plans. The company's technology roadmap is extremely solid, and the gap with competitors continues to widen. The new product launches will help replicate its dominance in the inference market, similar to its training market. Based on order conditions, the company's performance still has room for further growth. Considering NVIDIA's market position, the current valuation (the expected price-to-earnings ratio for fiscal years 2026/2027 is about 15 times the expected earnings per share) is quite attractive, and we still recommend buying."

AI inference refers to the process where trained AI models apply learned logic and patterns to new data to make predictions, generate content, or make decisions.

At this year's GTC conference, NVIDIA announced a series of significant news: it expects to create $1 trillion in revenue opportunities by 2027, announced a partnership with OpenClaw, launched the Groq 3 language processing unit (LPU), achieved integration with the Vera Rubin architecture, and showcased multiple new products.

According to analysts, NVIDIA's Chief Financial Officer Colette Kress confirmed that this $1 trillion revenue expectation comes solely from Blackwell and Rubin and their related network products, excluding other product lines such as Groq LPU, CPX, and CPU racks.

The Rasgon team pointed out: "Therefore, we expect the actual performance of the data center business to far exceed this $1 trillion target, significantly surpassing market expectations. Notably, the implied revenue of about $500 billion in the fiscal year 2027 is already higher than the market's general expectation of $438 billion."

Analysts also emphasized that NVIDIA's comprehensive platform strategy's competitive advantages are becoming increasingly apparent. The company continues to deepen its software and hardware ecosystem across multiple product areas, including GPU, CPU, and DPU, and has now expanded into LPU, networking, and storage. Each generation of products reduces Token computing costs by an order of magnitude, which will help the company gain an advantage in the era of exponential growth in inference computing.

The Rasgon team candidly stated: "Frankly, we find it increasingly hard to imagine how other companies can compete with NVIDIA."

Citi maintains a "Buy" rating on NVIDIA with a target price of $300.

The analyst team led by Atif Malik stated: "After listening to the keynote speech, we are even more convinced that NVIDIA's technology roadmap is clear, and its pace of innovation continues to lead competitors."

The team particularly focused on three highlights: First, NVIDIA is approaching a turning point in its inference business, with data center sales expected to reach $1 trillion from 2025 to 2027, aligning with Citi's and investors' expectations, and exceeding the market's expectation of $950 billion. Since this $1 trillion does not include LPU, standalone CPU, and Hopper sales, the actual data could increase by several hundred billion dollars Considering the 9 to 12 months delivery cycle, there is still room for upward adjustment in the expectations for 2027.

Secondly, the product roadmap for 2026-2028 showcases NVIDIA's extreme pursuit of collaborative design, covering GPUs, CPUs, new Co-Packaged Optics (CPO) chips (suitable for horizontal and vertical scaling), and the brand new LPU chips.

Thirdly, Groq IP is used in LPU chips, in conjunction with Rubin, which is said to provide a 35-fold increase in throughput, making NVIDIA more competitive in the fast inference workload domain.

Morgan Stanley maintains an "Overweight" rating on NVIDIA, reiterating its position as a preferred stock in the semiconductor sector, while maintaining a target price of $260.

Analyst Joseph Moore and his team stated: "This year's GTC keynote focused on the inference market, and NVIDIA, with its leading advantages in hardware and software, is driving the development of the next generation of agent-based AI workloads. The core message is clear: inference based on the NVIDIA platform has a significant advantage in terms of cost per token, and the launch of Rubin will further expand this advantage. Our research also confirms this. The financial outlook is positive but not overly aggressive, and this prudent attitude is commendable."

Bank of America maintains a "Buy" rating on NVIDIA, with a target price of $300, and continues to list it as a preferred target in the AI field.

Led by analyst Vivek Arya, the analyst team stated that NVIDIA's full-stack decoupled product line demonstrates a market potential of $1 trillion, further expanding its leading advantage in the AI inference domain