Susquehanna raised its price target on AMD this week, arguing that the company’s next-generation server processor could double its share of the data-center market. The report points to a structural shift that AMD has been betting on for years: the movement of AI inference work from specialized chips back to general-purpose processors.
The analyst’s case rests on Turin, AMD’s upcoming server CPU, which the firm expects to win sockets in data centers that have historically belonged to Intel. AMD has taken share from Intel steadily over the past half decade, and the new chip generation is designed to accelerate that trend.
The inference shift is the more interesting part of the argument. AI workloads split into two phases: training, where models learn from data, and inference, where they respond to users. Training runs almost entirely on Nvidia’s GPUs. Inference, the argument goes, is a different game, and a growing share of it will run on CPUs, the chips that already sit in every server.
AMD’s position in that transition is strong. Its server processors have closed the performance gap with Intel, and its technology for accelerating AI operations on CPUs has matured. The company also sells its own GPUs, giving it a foothold in both camps and a hedge if the inference shift is slower than expected.
The report lands at a favorable moment. The broader chip rally this week, driven by Microsoft’s record data-center commitments, lifted the whole sector, and AMD shares have outperformed the group over the past year on the strength of its AI progress. The company’s data-center revenue has grown faster than any segment of its business.
The competition is the counterweight. Nvidia has been developing its own CPU designs, and its dominance of the AI software stack gives it a moat that hardware performance alone cannot cross. Intel, meanwhile, has been fighting to protect its server franchise, and its latest processors have closed some of the gap with AMD.
The market share math is the heart of the bull case. AMD holds a minority share of the server CPU market, and every point of share it gains translates into billions of dollars of revenue. If Turin delivers, and if the inference shift accelerates, the company’s data-center business could double within a few years, Susquehanna argued.
The risks are visible in the stock’s valuation. AMD trades at a premium to most of its peers, a multiple that already assumes a substantial share gain. A Turin misstep, or an inference market that stays on GPUs longer than expected, would compress the multiple quickly.
The company’s own guidance has been cautious. AMD executives have described the inference opportunity in measured terms, warning that the transition will take years and that GPUs will remain the primary engine of AI computing for the foreseeable future. That caution has been a feature for investors who remember AMD’s history of overpromising.
AMD’s rise in servers has been a decade in the making. The company’s share of the server CPU market was in the single digits when it introduced its current chip architecture in 2017, and it has climbed steadily since, taking most of its gains from Intel. Turin is designed to continue that march, with more cores, faster memory and built-in AI acceleration.
The company’s GPU business gives the thesis a second leg. AMD’s Instinct line of accelerators has won contracts with hyperscalers and cloud providers, and the company has said its AI revenue is growing faster than the overall market. The combination of CPUs and GPUs positions AMD as the only major supplier with credible products in both halves of the AI compute market.
The competitive threats are real. Nvidia has announced its own server CPU designs, and its software ecosystem, which developers learn first, gives it an advantage that hardware benchmarks do not capture. Intel, meanwhile, has been fighting to hold its server franchise, and its recent processor launches have narrowed the performance gap with AMD.
The valuation question will resolve itself in the data. AMD’s stock already trades at a premium, and the Susquehanna target assumes the share gains materialize on schedule. If Turin ramps into the data centers that Microsoft’s record lease commitments are now funding, the numbers will do the talking. If it slips, the premium will deflate faster than the chip ships.
The immediate question is execution. Turin needs to ship on time, win the marquee customers that anchor server platforms, and hold its advantage until the next generation. If it does, the analyst’s doubled-share forecast starts to look conservative. If it stumbles, AMD’s AI story will have to wait for the chip after next.
