The AI Shift That Nobody Is Pricing For Intel

The AI Shift That Nobody Is Pricing For Intel

Can Intel (NASDAQ:INTC), the legacy CPU giant, remain relevant in a world obsessed with GPUs?

It sure can. In fact, Intel could thrive.

As the artificial intelligence market transitions from the highly compute-intensive training phase to higher-volume inference - which is the practical execution of trained models - Intel's role may expand, not shrink, as CPUs become central to managing AI workloads at scale. While the initial AI boom was defined by GPU clusters designed to build foundational models, the current era is defined by inference. This is where revenue is generated as companies deploy the models they spent on training.

NVIDIA has projected that the demand for its AI chips will reach $1 trillion from 2025 through 2027, driven by an "inflection point of inference." In fact, inference is likely to overtake training workloads around 2027 and considerably exceed training loads by 2030. This shift from "creation" to "usage" directly favors Intel's architectural roadmap and its massive manufacturing scale.

Although Nvidia stock barely moved following its big $1 trillion outlook for top-end chips sales, we believe these Nvidia vendors and customers could rally.

Why Does Inference Drive CPU Demand?

In the training phase, CPUs were often just "traffic cops" of sorts for GPUs. Things change with inference.

Inference requires constant data movement, pre-processing, and post-processing. These tasks are handled by CPUs. As inference volume grows, the bottleneck shifts from raw compute to data orchestration, and the sheer number of CPUs needed to feed data to accelerators is skyrocketing. CPUs are becoming the control layer of AI infrastructure.

The rise of agentic AI - autonomous systems that can plan, reason, use tools, make decisions, and execute multi-step workflows - is accelerating this trend even further.

Besides this, for smaller or medium models (under 20 billion parameters) used in enterprise tasks like document summarization or search, high-end CPUs such as Intel Xeon 6 may often be more cost-effective and energy-efficient than dedicated GPUs. In many real-world deployments, GPUs are overkill.

Moreover, newer AI systems, such as NVIDIA's Rubin platform, explicitly use high-performance CPUs to govern memory access and data orchestration, while offloading security across the entire cluster to dedicated Data Processing Units (DPUs) like BlueField.

There are proof points to support this shift. Intel is sold out of server CPU capacity for 2026 due to AI demand from hyperscalers. It is already capacity-constrained, and there have been reports that average selling prices could rise 10% to 15% this year.

But AMD and Nvidia Arent Sitting Still

Now the CPU landscape isn't what it used to be a decade or so ago. Intel faces a multifaceted competitive environment in catering to the AI space.

AMD has been consistently gaining share in the server market with its high-core-count EPYC "Turin" processors. GPU titan Nvidia is pushing its own ARM-based "Vera" CPUs to create tightly coupled superchips. Furthermore, hyperscalers such as AWS and Google are deploying custom-designed ARM silicon to handle internal inference workloads.

Despite these pressures, Intel's deep history with the x86 instruction set remains an advantage, as most global enterprise software is natively optimized for Intel architecture, making Xeon a strong choice for many data center refreshes. While the AI rollout over the last four years has likely changed the narrative a bit, this software inertia still matters to an extent.

Manufacturing and Foundry Capacity

The most significant tailwind for Intel is its dual role as a designer and a high-volume manufacturer. As inference drives a spike in silicon demand, Intel's massive manufacturing capacity - a large mix of which is located in the U.S. - could prove vital.

Intel is now working with the likes of Nvidia, which has committed $5 billion to a strategic partnership. As part of the deal, Intel will design and build custom x86 CPUs for Nvidia's AI infrastructure, marking a significant shift toward deeper ecosystem integration rather than a pure foundry relationship.

Currently, 90% of advanced AI chips are manufactured by TSMC in Taiwan. Governments and hyperscalers (Microsoft, AWS, Google) are under immense pressure to de-risk their supply chains from potential geopolitical instability in the Taiwan Strait, and this could play to Intel's advantage. Intel has begun high-volume manufacturing of its advanced 18A (1.8nm-class) process node in Arizona, the first leading-edge semiconductor process of its kind to reach volume production on U.S. soil.

Taking Intel To $300 Market Cap

Now Intel stock has roughly doubled since early 2025, trading around $44, with a market cap of about $220 billion, as markets begin to price in its manufacturing optionality, government support, and rising CPU demand. If inference-driven workloads add, say, $10 billion in incremental annual revenue, and Intel can sustain roughly 20% net margins on that business, that translates to roughly $2 billion in additional operating profit. Apply a 40x multiple to that incremental earnings stream and you get $80 billion in incremental equity value. While a 40x multiple is admittedly high for a CPU business, it could be easily justified if Intel captures a recurring role in AI infrastructure rather than a one-time upgrade cycle, allowing the earnings base to compound.

While Intel's opportunity looks compelling, if you want to deploy high-conviction, data-backed strategies across your entire portfolio without managing the day-to-day execution yourself, we can help. Our Trefis HQ strategy has outperformed its market benchmark (a combination of the S&P 500, S&P mid-cap, and Russell 2000) to produce over 105% returns since inception. Readers can see the full HQ performance metrics and 5 reasons why HQ outperformed.