What is inference, and why does it matter for this prediction?

Inference is the step where an AI generates an answer after processing a prompt. According to Bloomberg Intelligence projections, the inference market is expected to grow at a 32% compound annual growth rate through 2032 and become nearly double the size of the AI training market, hitting annual spending of $1.3 billion, making it more important to revenue than training.

What has AMD done to compete in inference?

AMD has improved its ROCm software stack, leveraged its chiplet design to reduce latency, acquired memory optimization company MEXT and inference chip start-up Taalas, and partnered with Cerebras on a disaggregated inference system where AMD handles the pre-fill phase and Cerebras handles the decode phase.

What is AMD's opportunity in agentic AI?

AMD is a leader in server CPUs and has been steadily taking share from Intel. With agentic AI increasing the ratio of CPUs to GPUs from 8:1 (for training) to 1:1, AMD sees this becoming a $220 billion market in the next few years.

AI Business & IndustryAI Stocks & MarketsYahoo Finance AIAug 22, 2026Aug 22, 2026, 22:00 JST3 min read

AMD poised to outperform Nvidia in AI chips over 3 years, analyst says

Send on LINE

Key takeaway

AMD is forecast to outperform Nvidia in artificial intelligence semiconductors over three years.
The inference market—where AI generates responses—grows at 32% annually through 2032 and will nearly double AI training's size to $1.3 billion annually.
AMD leads in inference and server CPUs for autonomous AI agents, a market expected to reach $220 billion soon.

3 Key Points

What happened
A Motley Fool analyst predicts Advanced Micro Devices (AMD) will outperform Nvidia over the next three years, citing AMD's strength in the faster-growing inference market and agentic AI, where it holds competitive advantages through software improvements, chiplet design, strategic acquisitions (MEXT and Taalas), and a partnership with Cerebras.
Why it matters
The inference market—where AI systems generate answers after processing prompts—is projected to grow at a 32% compound annual growth rate through 2032 and become nearly double the size of the AI model training market, hitting annual spending of $1.3 billion, according to Bloomberg Intelligence. While Nvidia dominates training, AMD is positioned as a serious competitor in this larger, faster-growing segment. Additionally, agentic AI (AI systems that act autonomously) is driving demand for server CPUs, a market AMD sees becoming $220 billion in the next few years—an area where AMD leads in server chips and has been gaining share from Intel.
What to watch
AMD's revenue growth is expected to accelerate as it executes on its inference and agentic AI strategies. The company's partnerships (particularly with Cerebras for disaggregated inference systems) and recent acquisitions will be critical to measuring whether it can translate its technical positioning into market share gains against Nvidia.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Get news like this every morning — free

Context & Analysis

The analyst's prediction hinges on a structural shift in AI spending: while Nvidia remains dominant in model training—the initial phase where large language models are created—the article emphasizes that inference (generating responses) is both technically less demanding and far larger in total market size. Bloomberg Intelligence projects inference will grow at 32% annually through 2032 and eventually represent nearly twice the spending of training, reaching $1.3 billion annually. This shift favors AMD because inference is memory-bound rather than compute-intensive, allowing AMD's chiplet architecture and software improvements to compete effectively where Nvidia's GPU dominance in training is less relevant.

AMD has also made targeted acquisitions and partnerships to close technical gaps. The purchase of MEXT (memory optimization) and Taalas (inference-specific chips) directly address inference workloads, while the Cerebras partnership creates a hybrid solution for the two phases of inference. Beyond inference, agentic AI—systems that autonomously execute tasks—demand far more CPUs relative to GPUs (a 1:1 ratio versus 8:1 for training), and AMD holds significant advantages here as the second-largest server CPU vendor with momentum against Intel. The analyst identifies this $220 billion opportunity as critical to AMD's growth trajectory.

FAQ

What is inference, and why does it matter for this prediction?: Inference is the step where an AI generates an answer after processing a prompt. According to Bloomberg Intelligence projections, the inference market is expected to grow at a 32% compound annual growth rate through 2032 and become nearly double the size of the AI training market, hitting annual spending of $1.3 billion, making it more important to revenue than training.
What has AMD done to compete in inference?: AMD has improved its ROCm software stack, leveraged its chiplet design to reduce latency, acquired memory optimization company MEXT and inference chip start-up Taalas, and partnered with Cerebras on a disaggregated inference system where AMD handles the pre-fill phase and Cerebras handles the decode phase.
What is AMD's opportunity in agentic AI?: AMD is a leader in server CPUs and has been steadily taking share from Intel. With agentic AI increasing the ratio of CPUs to GPUs from 8:1 (for training) to 1:1, AMD sees this becoming a $220 billion market in the next few years.

Yahoo Finance AIRead Original Article

Send on LINE

Get the latest AI Business & Industry news every morning

For example, today's edition would include:

DataAgent launches with $10M to auto-fix Kubernetes faultsSiliconANGLE AI · 46m ago
Taoyuan pitches northern AI data center hubDIGITIMES Asia · 46m ago
SK Hynix custom HBM boosts inference up to 5.15xDIGITIMES Asia · 46m ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Subscribe free →

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →