Positive news frequently emerges as institutions reassess NVIDIA: Bank of America sets a target of $235! Goldman Sachs and Citigroup's target prices were just announced

According to Zhitong Finance APP, NVIDIA (NVDA.US) revealed its strategic layout at the GTC conference in Washington, attracting the attention of global financial markets. Citigroup, Goldman Sachs, and Bank of America have successively released research reports, all announcing to maintain NVIDIA's "buy" investment rating, with the core consensus focusing on the certainty of the $500 billion sales scale of the Blackwell and Rubin platforms, multi-domain technological breakthroughs, and ecological expansion potential.

Among the three institutions, Bank of America is the most aggressive, setting a target price of $235, while Goldman Sachs and Citigroup are relatively conservative, setting a target price of $210, although this target has already been reached tonight.

Part.01 Core Consensus: $500 Billion Revenue Anchor

Revenue Certainty: Demand Visibility Far Exceeds Market Expectations

All three institutions recognize the core increment of "Blackwell + Rubin platform cumulative sales reaching $500 billion" disclosed by NVIDIA, and all point out that this scale is significantly higher than market consensus.

Citigroup's analysis believes that this scale implies a potential upward space of over $25 billion in data center sales by January 2027, and NVIDIA has additionally planned to ship 14 million GPUs over the next five quarters. Combined with the 6 million GPUs already shipped, the total will reach 20 million, providing strong validation for demand over the next 18 months.

Goldman Sachs explicitly pointed out that this figure is 10% higher than its previous expectation of $453 billion and 12% higher than Visible Alpha's market consensus of $447 billion.

Bank of America's calculations show that this scale is $50 billion higher than the current industry consensus, equivalent to five times the revenue of the Hopper platform's lifecycle (excluding the Chinese market).

Part.02 Multi-Line Expansion in AI Infrastructure, Quantum Computing, etc.

In the field of supercomputing, all three institutions focus on NVIDIA's deep cooperation with the U.S. Department of Energy, but with different emphases.

Citigroup specifically mentions that the cooperation covers seven systems, including the Solstice supercomputer at Argonne National Laboratory, which is equipped with 100,000 Blackwell GPUs, while Lambda is building a 100MW AI factory with over 100,000 GPUs.

Goldman Sachs focuses on the platform landing in specific laboratories, including the Solstice system at Argonne National Laboratory and the Rubin platform deployment at Los Alamos National Laboratory.

Bank of America highlights the superiority of NVIDIA's collaboration with Oracle on the OCI Zettascale10 system by comparing it with AMD's recent two supercomputing projects (totaling $1 billion in investment).

In terms of communication network layout, Citigroup emphasizes NVIDIA's collaboration with Nokia on the ARC Aerial RAN computer, believing it can support the construction of 5G-A/6G AI-native networks.

Goldman Sachs is more concerned with capital-level actions, namely NVIDIA's $1 billion investment in Nokia at a price of $6.01 per share, and clarifies that the ARC platform is the core cooperation vehicle for both parties Bank of America delves into the technical composition of the ARC platform, pointing out that it integrates Grace CPU, Blackwell GPU, and ConnectX chips, with a core focus on cloud-native base station scenarios.

In the field of quantum computing, the NVQLink technology has become a common focal point. Citigroup quantified the technical performance, noting that it achieves 400 times the GPU-QPU throughput and latency below 4.0ms, with support from 17 vendors and 9 laboratories; Goldman Sachs analyzed it from the application architecture perspective, believing that the technology enables efficient connections between quantum processors and error correction systems, laying the foundation for building accelerated quantum supercomputing; Bank of America emphasized ecological coverage, mentioning that all 17 quantum vendors support the protocol, and clearly stating that GPUs play a core role in error correction.

Part.03 Core Differences in Valuation and Performance

Comparison of Investment Ratings and Valuation Logic

Although all three institutions have given a "Buy" rating, there are significant differences in target prices and valuation logic.

Citigroup set a 12-month target price of $210, based on an estimated profitability of about $7 in 2026, using a 30 times price-to-earnings ratio, which aligns with NVIDIA's 35-year historical average level.

Goldman Sachs also gives a target price of $210, matching Citigroup, but with a different valuation logic, using a 35 times price-to-earnings ratio multiplied by an estimated $6 earnings per share in 2026 to arrive at the target price.

Bank of America's target price is significantly higher than the previous two, set at $235. Its valuation uses a price-to-earnings ratio for the fiscal year 2026 after excluding cash, with a multiple of 37 times, positioned in the middle of NVIDIA's historical price-to-earnings ratio range of 25-56 times, reflecting a higher recognition of its long-term value.

Core Divergence in Performance Forecasts

Citigroup's performance forecast focuses on the "rhythm of demand landing," not disclosing specific EPS figures, but particularly emphasizing NVIDIA's "14 million GPU shipment plan" as a rare clear signal in the industry, and noting that the $500 billion sales scale includes network business; if the data center business is separated, its actual growth elasticity may be higher than the overall estimate.

Goldman Sachs' forecast emphasizes "incremental driving factors," clearly providing specific figures: the EPS for 2026 is expected to be $6.75, rising to $8.26 by 2028. It believes the core of growth comes from three major increments: the deployment progress of OpenAI's Blackwell GPU exceeding expectations, the continuous growth of sovereign government orders, and the first-mover market dividends brought by the release of the Rubin platform.

Bank of America emphasizes "revenue - EPS transmission efficiency," calculating that an excess revenue of $50 billion could increase the 2026 EPS by about $1.15, making the annual EPS expected to reach $8. At the same time, it also provides the corresponding relationship between computing power and revenue: 1GW of computing power corresponds to $2.5-3 billion in revenue, while the target of 20 million GPU shipments by the end of 2026 will further amplify performance elasticity Part.04 Interpretation Differences of the Vera Rubin Platform

Regarding the Vera Rubin platform, which NVIDIA has prominently launched, the interpretations from three institutions have different focuses.

In terms of performance positioning, Citigroup did not analyze it separately but considered it within the overall demand framework of the Rubin platform.

Goldman Sachs clearly defines it as the core platform for the next generation of supercomputing, emphasizing its technological leadership.

Bank of America provides specific performance improvement data, pointing out that its inference model token generation efficiency is 100 times better than the Blackwell platform, highlighting its technological iteration value.

In terms of mass production timeline, only Bank of America provides a clear judgment, believing that the platform will start production in the fourth quarter of 2026; Citigroup did not specify relevant time nodes, and Goldman Sachs did not mention mass production plans in the report