Initially, people said Intel was America's national fortune stock, and I was dismissive—how could stock trading turn into MEME?

Later, it soared astonishingly, from 20 to 80, and people kept saying it was underestimated.
No way, missing out is not scary, what's scary is not knowing why you missed out.
Having missed out all the way, I TMD just kept researching:
Intel is actually two companies:
One is Intel Products—selling x86 CPUs(servers + PCs + AI PCs)—a mature cash flow machine. By 2025, total revenue is about $46 billion, gross margin 40-45%, a typical "long slope, not thin snow" business in the sense of Chang Yongping.
The other is Intel Foundry—burning money to gamble on U.S. semiconductor sovereignty with advanced process manufacturing. External revenue in 2025 is only a few hundred million dollars, with annual operating losses of $8.5-13.4 billion, but it has secured U.S. government 10% equity stake, SoftBank $2 billion, Nvidia $5 billion, Musk’s Terafab 14A first order—this is a typical 0→1 monopoly candidate.
The valuation logic, growth logic, and risk structure of these two companies are completely different. Combining them into a single P/E score results in a complete muddle.
To see Intel clearly, you must look at them separately.
1. Look at the trend: three structural trends, where Intel stands
Trend 1: AI infrastructure is a decade-long structural theme
This is basically undisputed. Gartner, McKinsey, Goldman Sachs all say similar things—by 2028, global inference computing power consumption will exceed training by more than 3 times, possibly 4:1 in China. The AI inference market in 2025 is about $106 billion, projected to reach $255 billion by 2030, with a CAGR of 19%.
More importantly, the economic structures of inference and training are different. Training is a one-time investment, with GPUs as the main players; inference involves continuous consumption of billions of requests daily, with CPUs as the indispensable scheduling hub. When Agentic AI moves from concept to enterprise deployment, serial logic processing, context switching, tool invocation, API orchestration—these workloads are primarily handled by CPUs.
This trend favors Intel, but Intel is not the biggest beneficiary. The biggest beneficiaries are Nvidia( for training), TSMC( for manufacturing), SK Hynix( for HBM). Intel is a participant, not a main player. It can only benefit from the CPU's "control plane" in AI systems—an important position, but not the top position.
Trend 2: U.S. semiconductor sovereignty is a 10-year irreversible geopolitical strategy
This is the truly unique trend for Intel.
In August 2025, the Trump administration used funds from the CHIPS Act that were not yet disbursed to acquire a 10% stake in Intel. This is not a one-time event but a paradigm shift in national industrial policy. Previously, CHIPS was "government funding, companies doing their own thing"; now, the model is "government equity, government endorsement, government backing." The possibility of extending this model to other strategic industries is being discussed—nuclear power, key materials, defense chips, etc.
Intel’s unique position in this trend is: the only option for advanced process foundry manufacturing in the U.S. TSMC’s factory in Arizona, even if it starts production of 2nm in 2028, the U.S. Department of Defense’s Secure Enclave project will always give priority to Intel. The Pentagon will not hand over its most confidential chips to a company headquartered in Taipei, regardless of where its capacity is.
This is a Thiel-style "secret"—a rarely agreed but potentially extremely important truth: if the situation worsens, U.S. AI hegemony cannot exist without Intel. This secret may not necessarily need to be realized, but it defines Intel’s "lower limit"—the U.S. government will not allow Intel to disappear.
This trend manifests in Intel as an exclusive monopoly.
Trend 3: Pressure for diversification in supply chains for mega-scale clients
AWS, Microsoft, Google, Meta, OpenAI—none of these clients are willing to rely solely on TSMC for all AI infrastructure. It’s not that they don’t trust TSMC, but no board of directors can accept a "single point of failure" in the supply chain.
This creates a structural window—only Intel in the U.S. can provide truly advanced process foundry manufacturing. Microsoft’s Maia2, AWS custom AI chips, Google’s IPU collaborations—all are based on this logic.
But the benefits of this trend for Intel are conditional. The condition is that Intel can handle the orders—good yield, large capacity. The brutal reality revealed by institutional analysis is: Intel’s 18A monthly capacity is about 10,000 to 15k wafers, while TSMC’s N2 is about 1.3-1.4 million wafers—10 times the gap. Even if all mega-scale clients want to allocate 20% of their orders to Intel, it cannot handle it.
The upper limit of benefits from this trend is constrained by Intel’s own capacity ramp-up speed.
Conclusions from the three trends combined:
Intel’s track is real—the AI infrastructure is upward for 10 years, U.S. semiconductor sovereignty is irreversible for 10 years, and supply chain diversification will continue for 10 years. But Intel’s position within these trends is:
AI infrastructure: participant(main force is Nvidia)
U.S. semiconductor sovereignty: sole main force(no substitute)
Supply chain diversification: potential beneficiaries(limited by capacity)
This is a "trend but not the biggest beneficiary" target. This position has its advantages—its lower limit is supported by national strategy, making it less likely to disappear. But it also has limitations—the upper limit is dominated by Nvidia and TSMC’s respective dominance, making explosive growth difficult.
2. Look at the business model: two companies, two perspectives
(Please refer to the better-formatted link)
"Intel INTC Research Report: Is it America’s National Fortune Stock?"