Goldman Sachs Maintains Buy on NVIDIA: Vera Rubin Nears Mass Production, Driven by AI PC and Agent Ecosystem

Overview

This report is Goldman Sachs' analysis of key takeaways from NVIDIA CEO Jensen Huang's GTC keynote at Computex Taipei in 2026. It highlights three core investment theses: next-generation Vera Rubin platform entering mass production optimized for Agentic AI, collaboration with Microsoft targeting high-end AI PC market, and accelerating enterprise-level AI Agent ecosystem through software toolchains. Based on these developments and improving visibility into hyperscale cloud capital expenditures, Goldman Sachs maintains its Buy rating and $285 target price, seeing upside catalysts for the stock in the coming months.

Core views

Product Iteration and Data Center Advantage Consolidation: NVIDIA announced the Vera Rubin platform has entered full mass production, comprising rack-level system components including NVL72 GPU, Vera CPU, Groq 3 LPU, BlueField storage, and Spectrum-X networking. The Vera GPU is specifically designed for Agentic AI scenarios, delivering up to 1.8x performance of x86 systems and 10x Agent throughput improvement over the previous Blackwell generation. Goldman Sachs expects the revenue ramp from Rubin starting in Q3 to be steeper than Blackwell due to significant manufacturing efficiency gains and larger total capacity. Additionally, the company's DSX AI Factory reference platform helps customers deploy AI data centers faster and optimize power consumption and runtime, further strengthening its dominant competitive position in markets beyond the largest hyperscale clouds. PC Market Expansion and AI Endpoint Deployment: NVIDIA is more actively entering the traditional PC market through collaboration with Microsoft and MediaTek. The newly launched RTX Spark platform combines Blackwell RTX GPU with a 20-core Grace CPU (co-designed with MediaTek), connected via NVLink to deliver high-performance experience optimized for AI applications, primarily targeting high-end segments. OEM partners including ASUS, Dell, HP, Lenovo, Microsoft, MSI, Acer, and Gigabyte will release notebooks, desktops, and workstations starting this fall. Goldman Sachs believes this could inject new momentum into the slow-moving Windows on ARM ecosystem. Software Ecosystem and Physical AI Strategy: In enterprise AI, NVIDIA released new software tools including NemoClaw, Nemotron 3 Ultra, OpenShell, and CUDA-X Agent Skills tailored for enterprise Agentic AI use cases. In Physical AI, the company introduced Cosmos v3, an open-source frontier model for multimodal reasoning; Alpamayo v2 as an autonomous driving reference platform; and the first open reference design for humanoid robots based on Isaac Groot and Jetson Thor hardware. These initiatives aim to accelerate developer and ecosystem partner adoption of Agentic AI through full-stack hardware and software capabilities.

Analysis framework

Goldman Sachs' analytical framework revolves around three themes: 'event-driven + product cycle + ecosystem validation.' First, using the specific event of the GTC Taipei keynote as a trigger to quickly extract key signals from management. Second, positioning new products (Vera Rubin, RTX Spark) within NVIDIA's own product iteration cycle, assessing commercial potential through performance comparisons with the previous Blackwell generation and production timelines. Lastly, shifting from pure hardware sales to a full-stack ecosystem perspective encompassing 'hardware + software + reference designs' to evaluate NVIDIA's moat depth in emerging areas like Agentic AI and Physical AI. For valuation, a standardized P/E multiple approach is used, translating product cycle expectations into specific earnings forecasts and target price.

Methodology notes

• Valuation MethodPE/PEG valuation

  Price-to-earnings valuation based on Normalized EPS

  The report applies a 30x P/E multiple to a Normalized EPS of $9.50 to derive the $285 target price. This method excludes non-recurring items, making valuation better reflect sustainable core business profitability, commonly used as a pricing anchor for growth technology companies.

• Industry/Sector Analysis FrameworkProduct life cycle

  Comparative analysis of revenue ramp curves between new generation and previous generation products

  The report compares the revenue ramp slope of Vera Rubin versus Blackwell to gauge new product commercialization speed. In the semiconductor industry, manufacturing yield, capacity readiness, and customer adoption pace for new architectures determine revenue realization timeline, making this cross-generation comparison key for predicting inflection points.

• Competition & Strategy FrameworkMoat / competitive advantage

  Data center full-stack performance and cost leadership as differentiated barriers

  The report highlights NVIDIA's advantage lies not only in single chips but in overall performance and cost efficiency of rack-level systems (GPU+CPU+networking+storage). This system-level competitiveness makes it difficult for competitors to break through with single components, constituting a structural moat in markets beyond the largest hyperscale clouds.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• NVIDIA Corp. (NVDA.US)

  Direct Beneficiary: Vera Rubin mass production, AI PC platform launch, and Agentic AI software toolchain all point to core business growth

  Strengths

  Data center full-stack performance and cost leadership; Vera Rubin manufacturing efficiency and capacity exceed previous generation; deep partnerships with Microsoft, MediaTek, etc., expanding PC market

  Comparison

  Maintains competitive dominance in markets beyond the largest hyperscale cloud providers compared to competitors

  Risks

  Slowdown in AI infrastructure spending; increased competition leading to market share or margin erosion; supply constraints

Key data

• 12-Month Target Price$285Based on 30x P/E and Normalized EPS of $9.50, implying 35% upside
• Vera GPU Agent Throughput10x BlackwellDesigned for Agentic AI, delivering up to 1.8x performance of x86 systems
• FY2027E EPS Forecast$9.50Normalized EPS estimate used for target price calculation
• FY2027E Forward P/E Multiple22.2xBased on current share price of $211.14
• FY2028E Revenue Forecast$635.1 billionReflecting growth expectations post Vera Rubin ramp-up

Impact & implications

The report posits that Vera Rubin's rapid mass production and AI PC's high-end expansion will open new growth avenues for NVIDIA, while enterprise Agent toolchain refinement helps lock in long-term software ecosystem revenue. For investors, this means NVIDIA's growth drivers are diversifying from pure data center training demand to inference, edge AI, and endpoint devices. Improved visibility into hyperscale cloud capital expenditure plans through 2027 and beyond also supports medium-term performance.

Risks

• AI infrastructure spending growth slowdown
• Intensified competition leading to market share loss
• Competitive pressure causing margin erosion
• Supply chain constraints affecting product delivery

What to watch

• Further clarity on hyperscale cloud capital expenditure plans for 2027 and beyond
• Actual revenue ramp pace of Vera Rubin platform starting Q3
• Market acceptance of Windows on ARM ecosystem driven by RTX Spark
• Developer adoption and commercialization progress of enterprise Agentic AI toolchain