Citi forecasts server CPU market to hit $132B by 2030

Citi now models the server CPU market reaching $132 billion by 2030, up from $29.3 billion in 2025, implying a 35% CAGR, with agentic CPUs expected to grow fastest to $59.4 billion. The firm raised Intel's price target to $130 and AMD's to $460, citing higher data center sales estimates and potential gains from Intel's ASIC business and AMD's AI accelerator pipeline. Citi also expects Intel to hold 47% server CPU share by 2030, versus 34% for AMD and 19% for ARM and others.

Analysis

The key takeaway is not the headline TAM expansion, but the implied redistribution of spend inside the server stack: if agentic workloads become the dominant growth vector, CPU relevance rises because orchestration, inference control, and data movement shift back toward general-purpose compute rather than pure accelerator spend. That is structurally supportive for AMD and, to a lesser extent, Intel, while ARM’s share gains in servers may be slower than the market has been pricing because the near-term incremental dollar pool is in x86-compatible datacenter refresh and adjacent ASIC/IPU content.

AMD appears better positioned than Intel to monetize the cycle because it has both product performance and a credible attachment point to scarce TSMC capacity; that combination tends to capture the highest-margin share of an upcycle first. The second-order implication is tighter supply for leading-edge wafer starts and advanced packaging, which could constrain upside for smaller CPU challengers and keep pricing discipline intact into 2026. Intel’s upside is more leverage-to-estimate revisions than proof of share gain, so the stock can react strongly to any data-center mix improvement but remains more fragile if execution slips.

The most important risk is timing: this is a 12-24 month earnings revision story, not an immediate revenue inflection. Consensus may be overemphasizing the long-dated TAM and underestimating how much of the benefit is already crowded into AMD’s premium multiple; if July’s AI day lacks a concrete customer name or volume commitment, the stock could give back a material part of the move even if the strategic thesis stays intact. Conversely, a tangible Anthropic announcement would validate the “CPU renaissance” framing and likely force sell-side upward revisions across both CPU and adjacent networking/IPU names.

The contrarian angle is that the market may be too willing to extrapolate agentic workloads as CPU-led when a meaningful share could still be absorbed by ASICs, custom interconnect, and memory bandwidth rather than merchant CPUs. If that mix tilts away from standardized x86, Intel’s and AMD’s share assumptions would prove too aggressive, and suppliers like TSMC would be the more durable beneficiary than either chip designer.

Key Decisions for Investors

• Long AMD vs short INTC on a 3-6 month horizon: best risk/reward is riding relative estimate revisions and execution credibility while fading Intel’s more speculative re-rating; target 1.5-2.0x upside on the spread if July confirms a named AI customer.
• Buy AMD July/August call spreads into any pre-event pullback: defined-risk way to express catalyst upside with a hard cap on premium decay if the AI day disappoints; size for event volatility, not structural holding.
• Initiate a tactical long TSM vs short ARM pair: if server CPU spend expands faster than the market expects, TSM captures the manufacturing bottleneck economics while ARM is more vulnerable to a delayed share translation in data center.
• Avoid chasing INTC strength above post-target-step-up levels until there is evidence of sequential data-center acceleration; this is a show-me story and likely offers better entry on a post-event fade or on any guide-down in legacy segments.