The combination of surging data center CPU demand and continuously improving profitability prospects for Intel’s 18A advanced chip process node is jointly fueling and advancing Intel’s super-bull narrative.
Against the backdrop of explosive growth in data center CPU demand and Intel’s 18A advanced chip manufacturing process entering a growth trajectory, Wall Street financial giants have recently grown increasingly bullish on the x86 architecture CPU powerhouse—$Intel (INTC.US)$Citigroup significantly raised its price target for Intel from $95 to $130, while another prominent investment firm, Melius Research, increased its target from $100 to $150, underscoring how ‘surging data center CPU demand plus continuously improving profitability prospects for the 18A advanced chip process’ are jointly igniting and propelling Intel’s super-bull narrative.
According to reports, as Intel has asked PC manufacturers to adopt chips built on its 18A advanced semiconductor manufacturing process, Wedbush Securities, a major Wall Street investment firm, views this as a positive signal that the semiconductor manufacturing giant is prioritizing margin expansion.
In a note to clients on Wednesday, Matt Bryson, an analyst at Wedbush Securities, wrote: “In our view, this strategy is highly rational, as Intel should prioritize using its older, capacity-expanded process nodes for higher-margin Xeon data center CPUs. Intel’s ability to bring online new capacity—thanks to existing cleanroom space—is a strategic advantage that enables it to leverage this strength.” He added, “The real question, in our view, is how strong the actual performance of the 18A node and chips built on this process will be (CEO Lip-Bu Tan recently indicated that yields are improving rapidly).”
At JPMorgan’s 54th Annual Global Technology, Media and Communications Conference,$Intel (INTC.US)$CEO Lip-Bu Tan stated that Intel 18A (an advanced chip process node below 2nm, equivalent to 1.8nm) is already supporting Panther Lake volume production, with yields improving by approximately 7% per month—exceeding Intel’s internal expectations.
According to the latest updates disclosed by Tan, Intel has released the 0.5 PDK for 14A and plans to roll out the 0.9 PDK to external customers in October. The team is already engaged in long-term advanced process planning for the 10A and 7A nodes. Tan also noted that as AI computing infrastructure shifts its focus from training to inference, CPUs are becoming increasingly critical and indispensable in the AI era, with the CPU-to-GPU deployment ratio accelerating from 1:8 toward 1:1—and potentially even reaching 4:1. Additionally, Intel’s business strategy indicates it is actively pursuing ASIC opportunities, offering customized AI CPU or AI GPU chip solutions.
CPU Renaissance
As Anthropic launches its highly anticipated Claude Cowork and super AI agents like OpenClaw—capable of autonomously executing complex tasks—reach critical mass in 2026, the wave of AI agents is rapidly sweeping the globe. The bottleneck in AI computing architecture is decisively shifting from GPUs, which are centered on matrix multiply-accumulate throughput, to data center CPUs focused on control flow, task orchestration, and memory/IO coordination. High-performance CPUs tailored for hyperscale AI data centers are now facing severe supply shortages.
Wall Street analysts are now broadening the AI compute infrastructure narrative beyond ‘GPU dominance/single-engine driven’ to a full-stack revaluation encompassing ‘AI GPUs/ASICs + CPUs + HBM/DRAM/NAND memory chips + optical interconnect-driven high-speed data center connectivity systems.’
Amid the global proliferation of AI agents, investment focus in AI computing power is transitioning from a ‘single-point compute race centered on GPUs’ to a ‘full-stack AI computing system driven by AI agents.’ The next wave of alpha returns will no longer be confined solely to leading names in AI GPUs or AI ASICs but will systematically spread across the entire stack of AI infrastructure—including CPUs, memory, PCBs, liquid cooling systems, ABF substrates, and broad-based wafer foundry services. Within this evolving AI investment narrative, CPUs, optical interconnects, and memory chips are poised to emerge as the biggest beneficiaries.
Over the past two years, the narrative around AI has been almost entirely dominated by GPUs, with CPUs playing what seemed like a supporting role in the AI arms race. However, with the rise of open-source, agent-based AI workflows (i.e., AI agents) such as OpenClaw, which have driven comprehensive growth in inference workloads, data orchestration, task scheduling, memory access, network communication, and multi-tool invocation, the market has come to fully realize that without a powerful CPU acting as the system’s central hub, GPU clusters cannot operate efficiently. This marks the CPU’s return from being an ‘underrated infrastructure’ to center stage in the chip industry, carrying a distinct ‘Renaissance’-like retro wave.
Early large-model inference primarily followed a ‘single request–single generation’ paradigm, where CPUs mainly handled data movement, request routing, and basic scheduling—typical auxiliary control-plane functions. However, with the advent of AI agents and reinforcement learning, system workloads have evolved from simple forward inference into complex closed-loop processes involving task planning, tool invocation, sub-agent coordination, environment interaction, state management, and result validation. The orchestration layer is inherently CPU-intensive, characterized by strong control flow, extensive branching logic, frequent system calls, and intensive memory access—tasks that GPUs cannot efficiently replace. Consequently, CPUs are transitioning from their historical supporting role to become a new bottleneck determining system throughput, latency, and resource utilization.
Surging CPU demand in data centers + rising momentum in advanced chip foundry capabilities
The 18A node serves as a manufacturing-side validation point for the narrative of ‘surging data center CPU demand combined with continuously improving profitability prospects from Intel’s 18A advanced process technology.’ Wedbush interprets Intel’s requirement for PC OEMs to adopt 18A chips as a ‘gross margin protection’ strategy. The core rationale is that Intel’s management is actively prioritizing its limited legacy-node capacity for high-margin Xeon, server, and industrial customers, while using 18A to support new client products—thereby optimizing both capacity allocation and gross margin structure.
Almost simultaneously, Tom’s Hardware cited a Nikkei report stating that$Intel (INTC.US)$Intel has already redirected its constrained Intel 7 capacity toward server and industrial customers, as these segments offer higher margins, amid persistently rising AI-driven data center CPU demand expected through 2025.
At JPMorgan’s 54th Annual Global Technology, Media and Communications Conference, Lip-Bu Tan’s core message to the market was clear: 18A is not merely a node on a technology roadmap—it has already entered the Panther Lake volume production phase, with yields improving by approximately 7%–8% per month, shifting from an ‘engineering risk’ to a ‘commercially verifiable’ asset. If 18A achieves stable volume production, it will not only support Panther Lake for PCs but also lay a foundation of trust for future server CPUs, AI head-node chips, ASIC foundry engagements, and customer adoption of the 14A node. This forms the central pillar of a broader narrative: Intel’s repositioning from a ‘lagging-edge process company’ to a ‘vanguard U.S. advanced chip manufacturing asset’ in the eyes of the market.
Meanwhile, as AI shifts from training to inference and Agentic AI, the strategic importance of CPUs within AI data centers will rise significantly. While GPUs handle large-scale matrix computations, CPUs manage scheduling, I/O, memory management, task orchestration, security isolation, database access, network stacks, and multi-agent workflow execution. As AI applications evolve from ‘one-time inference’ to ‘continuously operating software labor,’ CPU demand will escalate beyond traditional server refresh cycles into the AI infrastructure expansion cycle. Citi’s latest model echoes this view: it forecasts the total addressable market (TAM) for data center server CPUs to grow from USD 29.3 billion in 2025 to USD 131.5 billion by 2030—a compound annual growth rate (CAGR) of approximately 35%—and has accordingly raised its price target for Intel from USD 95 to USD 130.
The highest Wall Street price target of USD 150 comes from Ben Reitzes, star analyst at Melius Research. Melius has incorporated Intel into its revaluation framework for ‘bottleneck assets’ in AI semiconductors. Reitzes argues that as demand for AI compute infrastructure continues to create supply bottlenecks,$Intel (INTC.US)$semiconductor companies focused on AI compute bottlenecks stand to gain more market capitalization or upside potential compared to traditional software firms and non-semiconductor members of the Mag 7. Specifically regarding Intel’s fundamentals, Melius’ thesis rests on two key points: First, Agentic AI is reigniting acceleration in x86 server CPU demand—as AI transitions from training to inference and agent execution, more CPUs are required for scheduling, I/O, memory management, workflow orchestration, and security control. Second, under Lip-Bu Tan’s leadership, Intel Foundry (Intel’s advanced-process semiconductor manufacturing business) holds significant potential to unlock substantial growth value.
Editor/melody