Wall Street has spent much of 2026 debating whether AMD is an Nvidia alternative or a genuinely different bet. On Aug. 25, one of the Street’s higher-ranked chip analysts made his position clear.
AMD shares jumped roughly 4% on the day. The upgrade came with a specific argument about where server chip revenue is headed that goes well beyond the usual AI trade.
Raymond James upgrades AMD to Strong Buy, raises price target
Simon Leopold, Raymond James’ semiconductor analyst, upgraded AMD to Strong Buy from Outperform and raised his price target to $641 from $565 on Aug. 25, implying roughly 40% upside from AMD’s closing price that day, CNBC reported.
Leopold ranks 120th out of 12,496 analysts tracked by TipRanks, with a 60% success rate and an average return of 30% per rating. His $641 target sits above the consensus average of roughly $613 across all analysts covering AMD.
Read more AMD:
- AMD filling reveals unexpected SpaceX and Nutanix bet
- Nvidia dominates AI chips, but BOFA sees AMD closing in
“AMD offers the strongest combination of direct earnings leverage, datacenter positioning, and market-share gains,” Leopold wrote in his note. Then he added a sentence that landed harder than the price target itself: “AMD’s growth should enable it to overtake Intel during 2027.”
AMD is up roughly 113% year to date heading into the upgrade. The stock had already been one of the better-performing semiconductor names in 2026 before the Aug. 25 move.
AMD server CPU revenue forecasted to hit estimated $201 billion or more by 2030
The center of Leopold’s argument is not AMD’s AI accelerator business. It is server CPUs, a part of the market that tends to get less attention than graphics chips but which he thinks is about to get much more profitable for AMD.
Raymond James forecasts server CPU revenue growing at a 44% compound annual rate to roughly $201 billion by 2030. Leopold’s own forecast is slightly below AMD’s internal estimate of $220 billion. He noted that his number could close that gap if agentic AI is adopted faster than he currently models, according to Benzinga.
The logic is straightforward. AI factories need accelerators to train and run models. But those accelerators still need powerful CPUs to coordinate tasks and manage data.
As enterprises deploy more agentic systems, those multi-step workflows need even more CPU coordination. That is where AMD’s EPYC processors come in.
BMO Capital recently said AMD is on the “verge of becoming a complete AI infrastructure provider,” citing the Helios AI rack as a product that could help it gain share against Nvidia, as StockTwits reported.
Microsoft will deploy Helios across Azure AI services beginning in the second half of 2026. Anthropic has also signed a major deal with AMD that significantly expands the chipmaker’s push into AI infrastructure.
Caroline/Getty Images
AMD data center revenue up, EPYC gains market share against rival Intel Xeon
The upgrade also rests on AMD’s recent execution. Data center revenue surged in the second quarter to $6.7 billion, up 107% year over year. That accounted for more than half of AMD’s total quarterly revenue of $11.54 billion, which itself rose 50% year over year.
AMD’s server CPU market share has been moving in one direction. Its overall x86 processor shipment share reached 30.7% in the second quarter, while Intel’s fell to 69.3%.
AMD’s server-specific share rose to 34.5%, up 7.3 percentage points from a year earlier. Intel’s server share fell to 65.5%, according to Tom’s Hardware.
When comparing EPYC against Intel’s Xeon directly, AMD’s server share rises as high as 46.4%, Mercury Research President Dean McCarron noted, according to Tom’s Hardware.
Intel is still the larger supplier. But every percentage point AMD takes is revenue that did not exist in AMD’s model a year ago.
AMD stock risks valuation and what investors should watch in 2026
Leopold’s bull case has a long list of moving parts. SpaceX chose Nvidia exclusively for its orbital AI infrastructure. Custom silicon from Microsoft and Google keeps getting better. Intel is not standing still.
And agentic AI, the part of the thesis that pushes the server CPU market to $201 billion, could roll out slower than anyone’s model assumes right now.
At current prices, AMD trades at roughly 41 times projected earnings. Getting to $641 means crossing a trillion-dollar market cap. That requires earnings growth to outrun revenue growth for years.
Leopold thinks that will happen. The next few earnings reports will say whether he is right.
The three numbers worth tracking are data center revenue, EPYC server processor revenue, and gross margins. Data center revenue is where AMD has been winning, but the pace needs to hold.
EPYC is the specific product driving Intel share losses. Any sign the gap is narrowing would put the 2027 overtake prediction in question. Gross margins will show whether AMD is growing profitably or buying share at the expense of earnings quality.
Nvidia reports on Aug. 26. Any commentary from Jensen Huang on AI infrastructure demand or server CPU competition will land directly on Leopold’s thesis.
A strong Nvidia quarter reinforces the bull case for the whole sector. A cautious outlook does the opposite.
