By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel is largely bearish on Intel's stock, citing concerns about the profitability and timing of its external foundry business, competition in the CPU market, and the risk of Intel's 18A node becoming a stranded asset.
Risk: The risk of Intel's 18A node becoming a stranded asset if it fails to achieve parity with TSMC's N3, potentially turning the foundry business into a cash incinerator.
Opportunity: The potential for CPUs to gain relevance in an 'agentic' AI era, driving demand for Intel's products.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Jim Cramer, the longtime host of CNBC's Mad Money, recently named Intel (NASDAQ: INTC) 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.
Where to invest $1,000 right now? Our analyst team just revealed what they believe are the 10 best stocks to buy right now, when you join Stock Advisor. See the stocks »
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.
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.
During the earlier stages of the AI revolution, hyperscalers could sequence their chip purchases: first securing massive GPU clusters from Nvidia, 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.
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.
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.
Before you buy stock in Intel, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Intel wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $417,305! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,293,148!
Now, it’s worth noting Stock Advisor’s total average return is 936% — a market-crushing outperformance compared to 209% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of June 23, 2026. *
Adam Spatacco has positions in Nvidia. The Motley Fool has positions in and recommends Intel, Nvidia, and Taiwan Semiconductor Manufacturing. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"Intel faces structural margin and execution risks that may outweigh any upside from agentic-AI-driven CPU demand."
Intel’s case rests on two shifts: CPUs regaining relevance with agentic AI and a foundry that could diversify revenue. The piece correctly flags higher CPU-driven demand and a small but improving external foundry business. However, the bears aren’t wrong to worry: external foundry profitability remains negative (Q1 foundry revenue $5.4B; external revenue only $174M; unit losses persist). AI workloads today remain GPU-dominated, and CPUs as orchestration layers may not translate into durable earnings for years. Add capex intensity, margin pressure, and geopolitical/funding risks around offshore manufacturing, and the upside in a single stock looks unlikely to justify current valuations.
But bulls will argue that agentic AI could make Intel the CPU backbone of data centers, turning foundry headwinds into a scalable growth engine. If Lip-Bu Tan executes well and external customers expand, margins could improve and the optionality may finally pay off.
"Intel's foundry losses and competitive disadvantage in data center CPU efficiency outweigh the theoretical tailwinds from agentic AI workloads in the near term."
The thesis that CPUs gain relevance in an 'agentic' AI era is technically sound but operationally premature for Intel. While the shift toward orchestration-heavy workloads favors CPU-dense architectures, Intel is fighting a two-front war against AMD’s EPYC, which currently holds a superior performance-per-watt advantage, and Nvidia’s Grace Hopper superchips, which integrate CPU and GPU on a single coherent fabric. Intel’s foundry business is a long-term strategic asset for US sovereignty, but it is currently a massive cash incinerator. Until Intel demonstrates consistent 18A process node yields and stops losing market share in the data center to AMD, the stock is a 'show me' story rather than a value play.
If Intel successfully pivots its foundry business to capture the 'sovereign AI' capex wave, the stock could re-rate significantly as it transforms from a struggling chip designer into an essential piece of Western critical infrastructure.
"Intel's CPU thesis is directionally sound for agentic AI, but the stock's 263% rally has already priced in flawless execution on foundry recovery — a bet that contradicts Intel's 20-year track record of missing process node timelines."
The article makes a coherent case for CPU relevance in agentic AI, but conflates three separate theses without stress-testing execution risk. Yes, orchestration layers need CPUs — but AMD's EPYC and custom silicon (Google TPUs, AWS Trainium) already handle this. Intel's foundry business generated $174M external revenue on $5.4B total — a 3.2% attach rate that's still deeply unprofitable. The 263% YTD rally has already priced in a successful turnaround; Intel must now prove it can win design wins against entrenched competitors AND execute manufacturing at leading nodes without delays. The article treats these as probable, not contingent.
Intel's CPU advantage assumes hyperscalers will buy from multiple vendors for orchestration layers, but they've shown willingness to vertically integrate (Google TPUs, Microsoft Maia) or stick with AMD. The foundry business remains a cash drain with no clear path to profitability at the scale needed to matter.
"Intel's AI-driven re-rating is already largely reflected in its 263% YTD move, leaving execution risk dominant."
The article positions Intel (INTC) as an AI beneficiary via rising CPU demand for agentic workloads and foundry stabilization after Nvidia and government funding. Yet INTC has already surged 263% YTD, external foundry revenue remains tiny at $174 million, and the unit is still loss-making. Agentic AI infrastructure rollout is likely slower than hyped, giving AMD and others time to respond, while Intel's history of missed nodes and delays persists. The CPU-GPU multiplier thesis is plausible but requires measurable design wins soon to justify current valuations.
Even if external foundry traction stays weak, Intel's entrenched server CPU sockets could still see accelerated orders as hyperscalers scale agent fleets faster than expected, creating earnings upside the 263% rally has not fully priced.
"Near-term upside hinges on credible 18A yields and external design wins; without that, the rally risks multiple compression."
Gemini is right that 18A yields matter, but the bigger pitfall is timing and profitability, not the potential alone. Even with sovereign AI demand, Intel’s external foundry business must scale and reach breakeven margins for a meaningful upside; otherwise the stock’s move looks like multiple expansion on a thesis that may delay. Watch for tooling/EUV access, wafer capacity, and capex fatigue—any slip pushes the recovery further out and risks a sustained multiple compression.
"Intel's foundry strategy risks becoming a stranded asset if 18A fails to achieve performance parity with TSMC, regardless of hyperscaler diversification needs."
Claude, you correctly identified the 263% YTD rally, but the premise that this is already priced in is flawed. The rally is largely momentum-driven, not valuation-backed. The real risk overlooked here is the 'sunk cost fallacy' regarding Intel's 18A node. If Intel fails to achieve parity with TSMC’s N3, the foundry business isn't just a cash incinerator—it becomes a stranded asset. Hyperscalers won't subsidize Intel’s manufacturing incompetence just for the sake of supply chain diversification.
"Intel's foundry value hinges less on leading-node execution and more on whether geopolitical risk premiums override pure economics."
Gemini's stranded asset risk is real, but underestimates Intel's optionality. Even if 18A fails, Intel retains $5.4B annual foundry revenue (likely mature nodes) and could pivot to trailing-edge sovereign manufacturing—lower margin, but defensible. The sunk cost fallacy cuts both ways: hyperscalers may tolerate subpar yields if geopolitical supply chain diversification becomes non-negotiable. The real question isn't competence parity with TSMC; it's whether US/allied governments will subsidize Intel's survival regardless.
"Subsidy support hinges on 18A parity, not just diversification politics."
Claude's subsidy tolerance argument ignores execution thresholds. Even with geopolitical pressure, US funding via CHIPS Act requires measurable 18A progress and external customer traction to sustain; repeated delays have already triggered congressional scrutiny. If yields trail TSMC N3 by more than one node, hyperscalers will limit design wins to AMD and custom silicon, capping Intel's server socket recovery regardless of sovereign mandates.
The panel is largely bearish on Intel's stock, citing concerns about the profitability and timing of its external foundry business, competition in the CPU market, and the risk of Intel's 18A node becoming a stranded asset.
The potential for CPUs to gain relevance in an 'agentic' AI era, driving demand for Intel's products.
The risk of Intel's 18A node becoming a stranded asset if it fails to achieve parity with TSMC's N3, potentially turning the foundry business into a cash incinerator.