Citi Raises AMD Price Target to $800, Forecasts CPU Market to Reach $300 Billion by 2030
nashnova research
Citi lifted its AMD price target from $575 to $800, driven by analyst Atif Malik's forecast that the CPU market will swell to $300 billion by 2030 — a bet that AI agents are rewriting the compute-demand curve for the entire chip industry.
Why did Citi raise the target by 39% in one move?
Analyst Atif Malik raised AMD's price target from $575 to $800, maintaining a buy rating.
The catalyst: after a wave of personal AI agent launches — Muse, Dots, Instinct, GrokBot, and GeminiSpark — he remodeled CPU demand and concluded the market will reach $300 billion by 2030.
This means → Citi is not betting on any single AMD product. The thesis is that the entire CPU market expands because of AI agents — and AMD, as a leading player, captures a natural share of that growth.
How much more compute do AI agents consume than chatbots?
Traditional chatbots sit mostly idle between user interactions. AI agents — programs that autonomously execute multi-step tasks — can run continuously for hours or persist in the background indefinitely.
In plain terms = a chatbot is like a customer-service rep who waits for your question. An AI agent is a full-time assistant that works around the clock — constantly reasoning, constantly computing. The difference in compute demand is orders of magnitude.
This reflects a fundamental shift: AI is moving from "passive response" to "active execution," steepening the demand curve for inference chips.
How many GPUs will Meta's Muse need?
Malik singled out Meta's Muse as one of the first AI agents deployed at true social-network scale, but called its compute footprint "one of the most important unknowns in AI infrastructure."
His framework: in a base-case scenario, each user requires roughly 0.0020 GB200-class GPUs. At 100 million daily active users, that translates to 200,000–390,000 Blackwell-class GPUs, 2,800–5,500 NVL72 racks, and approximately $7–19 billion in one-time Nvidia revenue (including networking equipment).
Per-user math: every additional 1 million agent DAUs is worth roughly $70–190 million to Nvidia on new server builds.
Is Citi more bullish than Nvidia itself?
Nvidia previously forecast the CPU market at $200 billion by 2030. Citi's new estimate is $300 billion — a gap of 50%.
This means → the disagreement boils down to one question: after AI agents scale, does compute demand grow linearly or exponentially? Citi is betting on the latter.
The debate will ultimately be settled by actual procurement data from cloud providers and social platforms — whoever is placing large-scale GPU orders will reveal whose forecast is closer to reality.
What does AMD's CEO say? And what about Nvidia?
AMD CEO Lisa Su said the same day that she expects chip demand to remain at extremely high levels for years — echoing Citi's long-term expansion thesis.
Malik simultaneously maintained his buy rating and $315 target on Nvidia. Put simply = Citi sees this AI-agent wave as not a zero-sum game — AMD and Nvidia can both benefit.
AMD's pre-market stock rose 1.7% on the upgrade — an early sign the market is buying in.
市场有风险，内容仅供研究参考，不构成投资建议。