Bank of America sees Nvidia's push into AI CPUs as its next major growth engine, but entrenched rivals and shifting customer priorities could complicate the company's path to dominating the data center market
For years, Nvidia has been synonymous with high-performance GPUs powering artificial intelligence and cloud computing. But as demand for AI infrastructure accelerates, the company is setting its sights on a new frontier: the server CPU market, a space long dominated by Advanced Micro Devices (AMD) and Intel. According to Bank of America, Nvidia's expansion into CPUs could unlock a $170 billion opportunity by 2030, driven by the rise of agentic AI workloads that require faster, more integrated processing.
Recent earnings from Alphabet, Google's parent company, reinforced the intensity of AI investment. The company raised its 2026 capital expenditure forecast to between $195 billion and $205 billion, signaling that hyperscale cloud providers remain committed to building out next-generation data centers. Nvidia's stock has reflected this momentum, closing at $212.06 on July 22, up 2.3% for the day and delivering a 14.7% gain over the past six months-outpacing the S&P 500's 8.5% rise over the same period, based on data from Seeking Alpha.
AI CPUs: Nvidia's Next Growth Engine
Bank of America analysts, led by Vivek Arya, argue that Nvidia's Vera CPU is designed specifically for agentic AI, where each computational step depends on the previous one. This sequential nature makes single-threaded performance critical, and Vera's architecture-featuring 88 custom ARM-based cores, 1.2 TB/s of memory bandwidth, and a monolithic compute die-aims to minimize latency and data bottlenecks. CEO Jensen Huang has positioned Vera as the first CPU built for AI agents, betting that future workloads will prioritize rapid, step-by-step execution over raw parallel throughput.
Nvidia's approach goes beyond hardware. The company is integrating Vera CPUs with its Rubin GPUs, Groq LPX processors, and Spectrum and BlueField networking products, creating a tightly coupled system designed to maximize performance and raise switching costs for customers. This system-level integration could help Nvidia establish its architecture as the default for large-scale AI factories, making it harder for rivals to compete on individual component performance alone.
Valuation and Risks
Bank of America maintains a Buy rating on Nvidia, with a $350 price target-implying nearly 69% upside from the report's reference price of $207.29. This target is based on a multiple of 26 times estimated 2027 earnings, near the low end of Nvidia's historical forward price-to-earnings range. Despite the premium, analysts caution that the stock's valuation is tempered by three main risks: the unpredictable timing of large AI projects, the cyclical nature of Nvidia's gaming business, and potential power constraints as data center demand outpaces infrastructure capacity.
As of July 2026, Nvidia trades at 23.6 times forward non-GAAP earnings, which is 46% below its five-year average, according to Seeking Alpha. The company's leadership in both AI computing and data center networking supports the case for a premium valuation, but investors should be mindful of the volatility inherent in these fast-evolving markets.
AMD and Intel: Compatibility and Throughput
While Nvidia is betting on single-threaded speed and tight integration, AMD and Intel are emphasizing compatibility and rack-level throughput. AMD claims its EPYC 9965 Turin processor can deliver 2.4 times the throughput of Nvidia's Vera baseline in a modeled 100-kilowatt deployment, with its next-generation Venice platform expected to widen that gap. The debate also extends to chip architecture: Nvidia's ARM-based CPUs face an uphill battle against the entrenched x86 ecosystem, which underpins most enterprise software and databases.
For many businesses, the cost and complexity of migrating away from x86 may outweigh the performance benefits of a new platform. As AI agents become more deeply embedded in corporate systems, compatibility with existing software stacks could give AMD and Intel a practical edge, especially in environments where parallel processing and scale matter more than latency.
What's at Stake for Investors
Bank of America stops short of declaring a winner in the CPU battle. The agentic AI market is still in its early stages, and it remains unclear whether customers will prioritize lower latency and integration or stick with familiar, scalable platforms. For Nvidia, success hinges on whether its full-system approach and Vera CPU can redefine what matters in AI data centers. For AMD and Intel, the strength of the x86 ecosystem and rack-level efficiency may help them defend their turf.
Investors should recognize that the market may not produce a single winner. Nvidia could dominate latency-sensitive AI factories, while AMD and Intel remain competitive in enterprise environments built around compatibility and parallel workloads. As real-world deployments unfold, the balance of power-and the value already priced into these stocks-will become clearer. For a look at how Bank of America is evaluating other tech giants facing AI-driven shifts, see this analysis of Microsoft's upcoming earnings here.
Agentic AI refers to artificial intelligence systems that operate as autonomous agents, making decisions and taking actions based on sequential reasoning. Unlike traditional AI models that process large batches of data in parallel, agentic AI often requires rapid, step-by-step computation, making CPU latency a critical factor. This shift in workload characteristics is driving chipmakers to rethink processor design, balancing the need for compatibility, throughput, and integration. For investors and IT decision-makers, understanding these trade-offs is essential as the next generation of data center infrastructure takes shape.
