This Is the Top Artificial Intelligence (AI) Chip Stock to Buy Right Now, According to Jim Cramer (Hint: It's Not Nvidia) |

Jim Cramer, the longtime host of CNBC’s Mad Money, recently named Intel (INTC +0.65%) his top artificial intelligence (AI) chip stock. This was a pretty bold move considering that the stock has already rallied by 263% so far this year.

Indeed, Cramer commands one of the more durable audiences in retail investing. His rapid-fire delivery and unfiltered opinions have resulted in countless soundbites featuring actionable investment ideas amid market noise. With that said, his visibility can be polarizing, and detractors often label his calls hyperbolic — noting the many instances where his enthusiasm has outpaced important nuance or his timing has proven inaccurate.

Nevertheless, his Intel bull thesis centers on two underappreciated dynamics: the company’s CPU heritage as the artificial intelligence revolution heads towards its agentic AI era, and the tangible signs that its chip foundry operation is stabilizing. These points deserve scrutiny rather than a simple echo of pundit commentary. Let’s dig in to see if Cramer is right.

Move over, GPUs — CPUs are making a comeback

When given a specific objective to accomplish, agentic AI systems can plan out a set of steps, gather data, and follow through with multistep actions to complete it with minimal human oversight. These software models are changing the nature of the accelerated computing equation, moving it beyond its prior focus on parallel processing power. When it comes to training generative models and basic inference deployments, the complex matrix operations involved need to be handled by GPUs or other types of parallel processing chips. But when users are deploying fleets of autonomous agents, that introduces orchestration layers that CPUs handle more efficiently.

Duringthe earlier stages of the AI revolution, hyperscalers could sequence their chip purchases: first securing massive GPU clusters fromNvidia, and then retrofitting their servers or expanding CPU capacity later as their utilization needs became clearer. This tactic worked when AI workloads were dominated by generic training jobs or simple inference serving.

However, the rise in agentic workloads is inverting the old logic. GPU servers already connect each accelerator with a host CPU to manage traffic, memory coherency, and virtualization. The growth of agentic deployments exponentially multiplies the volume of CPUs required. Because each agent instance can create its own dynamic sub-tasks by querying external APIs and maintaining persistent context, the CPU architectures to support the whole system must now be procured and installed earlier in the process.

Intel’s long history in server CPU production positions it to capture incremental socket demand that pure-play GPU designers will struggle to meet. The result is not a zero-sum displacement of GPUs, but a multiplier effect whereby each new tranche of AI accelerators sold results in orders for the CPUs that will make those clusters usable at scale.

Intel’s foundry recovery has been gradual, but respectable

Throughout most of the AI revolution, Intel struggled with advanced-node chip manufacturing. Recent capital investments from both Nvidia and the U.S. government, as well as the hiring of Lip-Bu Tan as CEO last year, have helped the company make rapid improvements in the foundry operation.

During the first quarter, Intel’s foundry business generated $5.4 billion in revenue — an increase of 16% year over year. While this may look impressive on the surface, external foundry revenue — sales that are not attributed to Intel’s own products — was only $174 million. Meanwhile, the foundry unit is still operating at a hefty loss.

Nevertheless, I think that a credible turnaround of Intel’s foundry operation actually matters less for its own chips than for the broader AI infrastructure ecosystem. What I mean by that is that the chip sector’s concentrated reliance on a single offshore manufacturer (Taiwan Semiconductor Manufacturing) introduces a number of potential points of failure — geopolitical, logistical, or capacity-related.

Sophisticated buyers are going to increasingly price these factors into their capex plans. Against this backdrop, Intel’s ability to secure more external customers for its leading-edge process nodes would validate its recovery and help it diversify its revenue sources away from its legacy integrated devices. While its external foundry business is still small, it has grown nearly sixfold year over year. I’m cautiously optimistic the company can capitalize on the demand tailwinds going forward.

Is Intel stock still a buy?

Intel stock’s massive upward moves this year have already priced in considerable optimism about AI tailwinds. To achieve sustained share-price appreciation from here will require Intel to convert the CPU demand thesis into measurable design wins and achieve foundry milestones without the multiyear delays that have previously plagued it.

Furthermore, it’s important to realize that we are early in the agentic AI era. The infrastructure build-out required to support mass adoption of these applications will likely unfold more gradually than many pundits have predicted. Ultimately, this will give Intel’s competitors in the chip design space some time to respond.

Nevertheless, the combination of the resurgent relevance of CPUs and Intel’s recent validation as a third-party foundry gives it a degree of optionality that GPU-centric companies lack. Investors evaluating Intel are effectively betting that the next phase of the data center infrastructure build-out will reward balance across the AI chip stack over specialized products.

While Cramer’s endorsement amplifies Intel’s visibility, the underlying buy case should rest on more observable shifts in AI workload composition and supply chain choices. Whether this translates into durable earnings growth will depend on management’s execution, which is never guaranteed. With that said, the directional logic of paired CPU-GPU demand and chip designers’ desire to reduce the reliance on overseas foundry partners is enough to at least justify paying close attention to Intel’s fundamentals rather than dismissing Cramer’s commentary as mere market theater.