NVDA
Not just a GPU party anymore
For the last few years, AI investing has been basically a GPU beauty contest. Faster chip, bigger hype, repeat. But Bank of America says the next big AI fight may come down to something more boring-sounding and arguably more important: server CPUs.
Nvidia and AMD are now pitching two different dreams for how AI agents should run inside data centers. Nvidia’s pitch is all about speed — finish one task faster, keep the expensive GPU busy, don’t let the whole machine sit around twiddling its thumbs. AMD’s version is more like a factory floor: if the future is thousands of AI agents running at once, then what matters most is throughput, not just raw speed.
Why investors should care
That’s not just nerdy architecture trivia. Bank of America says the server CPU market could be worth about $170 billion by 2030, which is a pretty chunky prize for something most investors probably haven’t daydreamed about at dinner.
Here’s the investor takeaway:
- If hyperscalers decide latency wins, Nvidia’s approach gets a nice tailwind.
- If they care more about concurrency and rack-level utilization, AMD could get more love.
- If enterprise software keeps leaning on x86, AMD and Intel still have a case to make against Arm-based designs.
The Street still likes Nvidia
Even with that bigger CPU debate bubbling up, BofA didn’t exactly walk away from Nvidia with a sad little shrug. Analyst Vivek Arya reiterated a Buy and a $350 price target, saying Nvidia’s lead in AI compute and networking still deserves a premium.
That matters because Nvidia’s stock has already been priced like the AI prom king. When a top bank says the story still has room to run — even while acknowledging a new battlefield — that’s the kind of thing traders tend to notice.
Big picture: Nvidia may have won the GPU chapter, but Wall Street is now asking whether the CPU sequel is where the plot gets interesting.