Written by: Rita

Trends Guide

Nvidia and AMD are competing for standard definition rights in the $170 billion intelligent CPU market.

Under the wave of intelligent AI, there is a divergence in server CPU architecture routes. Nvidia advocates for faster cores, believing that single-core performance determines the system's upper limit. AMD, on the other hand, focuses on more cores, emphasizing that concurrent throughput is key.

Nvidia released the Vera CPU architecture last week, equipped with 88 custom ARM cores, 1.2TB/s memory bandwidth, utilizing a monolithic compute die design. The rationale is that intelligent AI requires repeated interaction between the CPU and GPU, with each round relying on the completion of the previous step; thus, single-core performance directly affects overall response speed. AMD will respond at AI Day on Thursday, with market expectations that it will emphasize the "more cores" route. Currently, the EPYC 9965 has achieved a rack-level throughput 2.4 times that of Vera in a 100kW deployment scenario.

Bank of America has a clear judgment on this route dispute: whoever can define the next generation of measurement standards in the industry will win.

Two Technical Routes, Two Design Philosophies

Nvidia's Vera CPU differs from traditional "core stacking" server processors. 88 cores are not particularly remarkable in the server CPU field, but Nvidia focuses on "maximum single-thread performance." Vera's memory bandwidth reaches 1.2TB/s, and its on-chip interconnect bandwidth is 3.4TB/s, with all designs aimed at enhancing single-core operating efficiency.

Nvidia's logic is that intelligent AI differs from one-time large-scale parallel computing and more reflects repeated interactive loops between CPU and GPU. Involvements like tool calling, code execution, retrieval, and orchestration depend on each previous step. Insufficient single-core performance will slow down the entire AI's response speed, leading to GPU waits, and thereby reducing the overall utilization of the AI factory.

AMD's design philosophy, however, is radically different. AMD believes production-grade AI is closer to a distributed software platform, with components like databases, APIs, vector storage, orchestration engines, caching, and middleware running in parallel. In this scenario, the system bottleneck lies in the number of concurrent workflows that can be sustained within a fixed power budget. The rack-level throughput of EPYC 9965 in a 100kW deployment is 2.4 times that of Vera, with the next-generation EPYC 6 expected to reach 3.3 times.

x86 vs ARM: The Subtle Game of Software Ecosystem

Apart from the core number debate of "faster vs more," there is a subtler yet equally important front, which is the choice of instruction set and the ecological game between x86 and ARM.

Nvidia's Vera is based on the ARM architecture. Nvidia believes that as long as the microarchitecture is excellent, the difference in instruction sets is not important. Can ARM handle AI workloads? The answer is affirmative. Can ARM run enterprise software? If its performance surpasses x86 comprehensively, the software ecosystem will migrate accordingly.

AMD and Intel clearly hold a different view. Their argument is straightforward: intelligent AI is extending from model inference to enterprise-level workflows, with scenarios like databases, middleware, security platforms, and enterprise applications optimized based on the x86 architecture for decades. The feasibility of ARM replacing x86 cannot be proven by just a few AI test results.

With AI Day Approaching, AMD Welcomes a Response Window

The AI 2026 Day held by AMD on Thursday will be the first public response point in this route dispute.

Bank of America expects AMD will not simply compare speed through benchmark data, as that would fall into the "single-core performance" evaluation framework set by Nvidia. AMD needs to redefine the competition dimensions, shifting from "single-core performance" to "the intelligent agent's capacity in real production environments."

Bank of America believes this is the core of the debate. The essence of winning and losing is not a comparison of technical superiority, but rather which definition of measurement standards the industry ultimately accepts.

Trends Perspective

The insight of this Bank of America report lies in reverting a CPU technology route dispute to a competition for "standard-setting rights" within the industry.

The computational power demands of intelligent AI differ from traditional AI training. Traditional training is mainly large-scale parallelism, while intelligent AI is a hybrid of serial loops and concurrent scheduling. Which architecture is superior depends on the definition of the evaluation standard. If the industry uses "individual intelligent agent response latency" as the core metric, Nvidia's route is more reasonable. If the industry focuses on "single machine intelligent agent capacity," AMD holds the advantage.

Thursday's AMD AI Day will be an important node in this debate. However, Bank of America implies that the route dispute will not be resolved in the short term. The ultimate winner may not necessarily be the one with higher benchmark scores but rather the vendor capable of driving the industry to accept its defined measurement standards.

For investors, the investment value of this debate is not to determine the correctness of routes but to understand each company's strategic bets. Nvidia bets that intelligent AI is extremely sensitive to latency, while AMD bets that the core need in production environments is concurrency density. Both routes have the potential to succeed, and it ultimately depends on the real evolution direction of AI applications.

Bank of America gives both companies a buy rating, with a target price of $350 for Nvidia and $620 for AMD. This suggests that Bank of America believes this competition is not a zero-sum game, and both companies can achieve growth through differentiated routes, just with different growth paths.

Disclaimer

This article is a compilation and interpretation of a third-party brokerage research report (Bank of America Securities, July 22, 2026) by Trends Research. The ratings, target prices, profit forecasts, and related judgments quoted in the article are the opinions of the brokerage's analysts and represent their respective institutions' positions, not representing Trends Research's viewpoint, nor constituting any investment advice.

The market carries risks, and decisions should be made independently. This article should not be used as the basis for buying or selling any securities.