Intel (INTC) is expanding its portfolio of AI data center offerings, unveiling a new data center GPU, codenamed Crescent Island, and revealing a rack-scale reference design based on its Gaudi 3 systems.
Intel is in the midst of building out its AI strategy after falling behind the likes of Nvidia (NVDA) and rival AMD (AMD). During the company’s most recent earnings call, CEO Lip-Bu Tan said he will focus the chipmaker’s efforts on AI inference and agentic AI. Inference refers to running AI models.
Intel stock is up 84% year to date and 57% over the past 12 months.
“We need to start by first understanding emerging and real AI workloads — then work backwards to design software, systems, and silicon to enable the best outcomes for those particular workloads,” Tan said in his prepared remarks during the earnings call.
“We will strive to become the compute platform of choice,” he said. “But we will also work towards a full-stack AI solution. And I look forward to sharing more on our strategy in the coming months.”
While Intel didn’t say whether its Crescent Island data center GPU would be based on its 18A process node, it did say that it will use the company’s Xe3P microarchitecture and include up to 160GB of memory.
Intel’s 18A is one of its most important pieces of processor technology. Last week, the company revealed that its Core Ultra series 3 chips and upcoming Xeon 6+ processors, both based on 18A, will be coming to market by the end of 2025 and early next year, respectively.
“AI is shifting from static training to real-time, everywhere inference — driven by agentic AI,” Intel CTO Sachin Katti said in a statement.
“Scaling these complex workloads requires heterogeneous systems that match the right silicon to the right task, powered by an open software stack. Intel’s Xe architecture data center GPU will provide the efficient headroom customers need — and more value — as token volumes surge,” Katti added.
Intel said Crescent Island chips will begin customer sampling by the second half of 2026.
The company’s Gaudi 3 rack-scale systems, meanwhile, feature up to 64 AI accelerators per rack and are meant for real-time inferencing.
Intel’s moves are part of its broader turnaround strategy focused on gaining a foothold in AI while simultaneously developing a third-party foundry system that will see it build chips for customers similar to rival TSMC (TSMC.BA).
The company is also seeking to regain lost market share in the PC market, its largest business segment.
Intel (INTC) is expanding its portfolio of AI data center offerings, unveiling a new data center GPU, codenamed Crescent Island, and revealing a rack-scale reference design based on its Gaudi 3 systems.
Intel is in the midst of building out its AI strategy after falling behind the likes of Nvidia (NVDA) and rival AMD (AMD). During the company’s most recent earnings call, CEO Lip-Bu Tan said he will focus the chipmaker’s efforts on AI inference and agentic AI. Inference refers to running AI models.
Intel stock is up 84% year to date and 57% over the past 12 months.
“We need to start by first understanding emerging and real AI workloads — then work backwards to design software, systems, and silicon to enable the best outcomes for those particular workloads,” Tan said in his prepared remarks during the earnings call.
“We will strive to become the compute platform of choice,” he said. “But we will also work towards a full-stack AI solution. And I look forward to sharing more on our strategy in the coming months.”
While Intel didn’t say whether its Crescent Island data center GPU would be based on its 18A process node, it did say that it will use the company’s Xe3P microarchitecture and include up to 160GB of memory.
Intel’s 18A is one of its most important pieces of processor technology. Last week, the company revealed that its Core Ultra series 3 chips and upcoming Xeon 6+ processors, both based on 18A, will be coming to market by the end of 2025 and early next year, respectively.
“AI is shifting from static training to real-time, everywhere inference — driven by agentic AI,” Intel CTO Sachin Katti said in a statement.
“Scaling these complex workloads requires heterogeneous systems that match the right silicon to the right task, powered by an open software stack. Intel’s Xe architecture data center GPU will provide the efficient headroom customers need — and more value — as token volumes surge,” Katti added.
Intel said Crescent Island chips will begin customer sampling by the second half of 2026.
The company’s Gaudi 3 rack-scale systems, meanwhile, feature up to 64 AI accelerators per rack and are meant for real-time inferencing.
Intel’s moves are part of its broader turnaround strategy focused on gaining a foothold in AI while simultaneously developing a third-party foundry system that will see it build chips for customers similar to rival TSMC (TSMC.BA).
The company is also seeking to regain lost market share in the PC market, its largest business segment.
