For nearly three years, investors judged the AI race by one metric: GPU performance.
Bank of America thinks the next winner may depend on something far less discussed: server CPUs.
The firm says Nvidia Corp. (NASDAQ:NVDA) and Advanced Micro Devices Inc. (NASDAQ:AMD) now promote two competing visions for running AI agents, and that debate could shape a server CPU market worth about $170 billion by 2030.
The question isn’t which company builds the faster processor. It’s which metric customers decide matters most.
One Company Wants Faster AI, The Other Wants More AI
Bank of America says Nvidia measures success by how quickly a single AI agent completes its work.
“Nvidia introduced a new framework: max single-threaded performance at scale,” analyst Vivek Arya said.
The approach focuses on latency. AI agents execute tasks one step at a time, with CPUs feeding data to GPUs throughout the process. Faster CPU cores shorten each cycle and keep expensive GPUs busy.
AMD sees the problem differently.
The company argues future AI systems will run thousands of agents simultaneously across large data centers. That shifts the focus from speed to throughput.
“Production AI resembles a distributed software platform,” Arya said.
That architecture favors CPUs that handle more concurrent workloads rather than completing one task faster.
According to Bank of America, AMD estimates its upcoming EPYC Venice platform could deliver about 3.3 times the rack-level throughput of Nvidia’s Vera reference system under its modeling assumptions.
The Metric Could Decide Where Billions Flow
For investors, the biggest takeaway isn’t today’s benchmark. It’s that hyperscalers may soon adopt a new way to measure AI infrastructure.
“The key question is time-to-complete an agent or number-of-agents-per-rack,” Arya said.
If cloud providers prioritize latency, Nvidia’s strategy gains support.
If they prioritize concurrency and utilization, AMD could strengthen its position.
Bank of America expects AMD’s AI event to focus on that distinction rather than traditional benchmark comparisons.
The Debate Goes Beyond Nvidia And AMD
The discussion also extends to processor architecture.
Nvidia’s Vera uses Arm-based designs, reinforcing the industry’s shift toward custom Arm CPUs.
AMD and Intel continue to back x86, arguing enterprise software, databases and middleware remain heavily optimized for that ecosystem.
The outcome may influence far more than CPU market share.
If AI infrastructure spending increasingly depends on CPU architecture and workload optimization, investors may need to look beyond GPUs to identify the next winners in AI hardware.
What Are Analysts Saying Now?
According to Benzinga Analyst Ratings, Nvidia holds a Buy consensus with an average price target of $309.75, implying roughly 50% upside from Wednesday’s $207.16, with individual targets running from a low of $215 to a Street-high $500.
The recent flow has leaned higher.
| Date | Firm | Price Target | Action | Rating |
|---|---|---|---|---|
| Jul 14, 2026 | Keybanc | $310 → $330 | Maintains | Overweight |
| Jun 5, 2026 | China Renaissance | New → $319 | Initiates | Buy |
| Jun 2, 2026 | Needham | $270 → $270 | Reiterates | Buy |
| Jun 1, 2026 | DA Davidson | $300 → $300 | Maintains | Buy |
| May 27, 2026 | Tigress Financial | $360 → $425 | Maintains | Strong Buy |
| May 21, 2026 | UBS | $275 → $280 | Maintains | Buy |
Bank of America sits inside that consensus.
In the note, Arya reiterated a Buy and a $350 price objective — roughly 69% above the $207.29 level cited in the report — arguing Nvidia’s lead in AI compute and networking still justifies a premium even as the CPU front opens.
The GPU made Nvidia the most valuable company on earth.
The CPU is where it now has to prove the story isn’t over.
Photo: Shutterstock
Login to comment