Wall Street’s sleeping on Nvidia, AMD, and Broadcom? A major investment bank says yes
Nvidia, AMD, and Broadcom are the most talked-about companies in AI. JPMorgan still believes investors don’t grasp how much these three enterprises are worth, even after many big mergers, Wall Street upgrades, and values in the trillions of dollars.
On a new note, analyst Harlan Sur says that the AI revenue potential for these three chip titans is far higher than current estimates, even if targets are going up, thanks to these companies’ substantial progress.
“Simply put, though out-year Street numbers for AVGO, NVDA and AMD have moved sharply higher over the past couple of weeks, there is still understandably some conservatism being baked into estimates,” Sur wrote.
We anticipate a period of sustained positive revisions to AI revenue estimates as visibility improves and additional capacity deals are announced.
JPMorgan thinks visibility will emerge when tech companies build next-gen data centers driven by GPUs and XPUs, which are the gear that all three businesses need.
They recently struck big agreements with OpenAI and other top AI companies. This is what Sur calls “a growing AI pie,” and it may bring Nvidia’s earnings up to $35 billion per gigawatt, with Broadcom and AMD not far behind.
Nvidia, AMD, and Broadcom are key players fueling the AI data center boom.
AI has already changed what people expect from Nvidia, AMD, and Broadcom. But JPMorgan sees something more fundamental happening: an AI hardware supercycle that even recent upgrades from analysts haven’t completely taken into account.
More Nvidia:
Power is at the center of everything. In particular, the wattage and cost of next-generation data centers. JPMorgan says that each gigawatt of AI computation, which is around the same amount of electricity required to operate these facilities, brings in tens of billions of dollars for the chipmakers who create them.
-
Nvidia: Could generate $35B–$40B per gigawatt, especially with the upcoming Rubin and Rubin Ultra platforms.
-
AMD: New Helios platform expected to hit $20B per gigawatt; OpenAI deal alone could approach $30B+ in annual revenue.
-
Broadcom: Estimated at ~$27B per gigawatt with XPUs; full deployment from OpenAI could mean $70B–$90B over three years.
Related: StubHub just landed on a big-time analyst’s buy list
JPMorgan says that “some conservatism being baked into estimates,” even if those numbers are high. This is partially because there are still worries about getting the money and building the infrastructure.
The company thinks that AI revenue predictions will go up again shortly, however, as deal visibility improves and additional capacity is presumably on the way.
JPMorgan’s call isn’t only good news for Nvidia, AMD, and Broadcom shareholders; it’s also bad news. If the company’s revenue calculation is right, conventional value models may not be in line with reality.
For example, Wall Street has already priced in Nvidia’s supremacy throughout the Blackwell era. The Rubin Ultra platform, which is projected to be released by 2026, might bring in $70,000 to $80,000 per chip, which is about three times as much as Blackwell’s income per die.
Related: Nvidia just scored a massive AI win, but Jensen Huang has regrets
The same thing happened with AMD’s OpenAI agreement. The Street thinks that data center GPU sales will be approximately $31 billion in 2027, while JPMorgan thinks that the OpenAI alliance might bring in $30 to $35 billion a year on its own. That means AMD could already have enough orders to outperform projections, not including other customers.
Broadcom could be the greatest shock of all. The company thinks that Broadcom’s AI revenue might reach $100 billion by 2027, up from just $21 billion in 2025 and 50–70% more than the average estimate.
JPMorgan isn’t providing official price targets here. The message is clear, though: the next step for AI is hardware-driven and power-limited, and these three names remain at its core.
If JPMorgan is accurate, and the AI gold rush is just in its second inning, then Nvidia, AMD, and Broadcom may all be sitting on underestimated growth curves.
That doesn’t mean there aren’t any hazards. Execution, infrastructure problems, and changing regulatory winds may all slow down growth.
But with tech companies pouring billions on bespoke silicon and new data center designs, chipmakers that are connected to AI acceleration seem to be in a good position — maybe even underrated.
Related: Samsung, Apple AI take different approaches to AI surge
This story was originally reported by TheStreet on Oct 22, 2025, where it first appeared in the Investing section. Add TheStreet as a Preferred Source by clicking here.
