INTC has seen its market price decline from its 2021 peak of $64 down to $19, marking a decline of more than 70%. Of course, this decline happened for good reason. The key reasons behind INTC's decline are many, including its sector rival NVDA running away with the most profitable chip-supply deals, and AMD gaining an ever-increasing market share in areas such as server-grade chips and PC processors. Even more, INTC has been a cash-burning machine, having announced roughly $100 billion in capital commitments across its multi-year Foundry buildout: a third-party semiconductor manufacturing service for other chip designers. That budget has kept climbing for years, with little sign of slowing.
Intel closed out the 2024 fiscal year with a headline $18.8 billion net loss on $53.1 billion of revenue — the worst year for the company in more than 4 decades. That is represented in its stock price, which currently sits at just $19 per share, valuing the company at merely $82 billion. However, has the company exhausted its streak of bad luck, and is it due for an upside surprise?
My thesis for INTC is threefold:
- The legacy CPU business on its own is worth double what INTC is trading at today.
- AI compute demand is shifting away from training, a GPU-heavy practice, and towards inference, a CPU-friendly process. Such a transition would heavily benefit INTC's core franchise, even more so than what the market is forecasting.
- INTC is the only credible path to leading-edge chip manufacturing on US soil. Washington cannot afford to let that capability slip from its hands, and that should artificially put a floor on the stock.
Pillar 1: The legacy business alone is worth double
Currently, the market is pricing INTC as if the Foundry's losses are permanent and the core business is impaired right along with it. Fiscal 2024 shows that "Intel Products" produced an Operating Income of $13.1B on $48.9B of Revenue — an operating margin of 27%. However, once we include the Foundry business, the company sits at a consolidated operating loss of $11.7B. That entire gap is accounted for by the Foundry's cash burn plus additional impairment and restructuring charges booked below the operating line. In other words, INTC's underlying business model is not broken — it is buried under the headline numbers.
If the board halted all incremental R&D and Capex on the Foundry and the restructuring charges finally stopped, a 12x EV/EBIT multiple on that $13.1B of operating income puts Enterprise Value at $157B. Net of Intel's ~$28B net debt, and adding back the ~$4B implied by Intel's retained 49% stake in Altera (from the Silver Lake deal), that is an equity value of roughly $133B — just over $31 per share on 4.3B shares. That 12x multiple is deliberately conservative: AMD and Texas Instruments both trade well above it. At 14x, the same math gets to roughly $37 per share; at 16x, roughly $43. So even using multiples below what comparable, profitable chip franchises actually command, the legacy Products business alone brackets $31-43 per share — call the upper end of that range $40-43 — before we underwrite a single dollar for the Foundry business.
Pillar 2: AI demand is shifting from training to inference
Market watchers generally agree that the last two years have been all about spending on AI infrastructure, laying the groundwork for AI usage. As a result, most of the money spent by hyperscalers such as Google, Amazon, and Microsoft has gone toward building data centers and training AI models. Training these models has been enormously GPU-intensive, with the rule of thumb being something around 8 GPUs for every 1 CPU, with the CPU mostly there to feed and orchestrate the GPUs. That ratio makes sense for training, but it does not hold for inference. The industry will shift away from training a handful of frontier models and toward running inference on hundreds of millions of devices, which will change the compute mix.
CPUs are much better suited to inference than to training, and agentic workloads that rely on tool-calling, retrieval, and sequential logic lean heavily on CPU-side execution — much more so than training does. Even more, as frontier models become more efficient, an ever-larger share of real-world inference will run perfectly well on CPUs, and vendors will start optimizing their software for that by design as they focus more on cost than raw throughput. That is exactly how INTC benefits: every dollar that moves from AI training to inference is a dollar that falls within the scope of INTC's products. My own view is that the effective GPU:CPU ratio within the AI industry will compress meaningfully from the current 8:1 ratio down to a more reasonable 3:1 or even 2:1. I am not aware of anyone else putting a hard number on this yet — it is my own directional call, and it is the main reason I believe analysts' 2025 and 2026 revenue estimates for INTC look understated.
Pillar 3: The US cannot afford to lose its only domestic path to leading-edge chips
TSMC is, practically speaking, the only company in the world currently able to manufacture the most advanced semiconductor chips at a scale that can satisfy worldwide needs, and that capability sits almost entirely in Taiwan. That geographic concentration is a national security risk for the United States — not because TSMC or Taiwan are adversarial, but because of what could cut off that supply. This is not hypothetical either. The Chinese government has, on multiple occasions, publicly declared its intent to bring Taiwan under its control. So, consider this: what would happen if China cut the US and its military off from its semiconductor supply? Modern US military hardware such as guidance systems and radar depends entirely on leading-edge semiconductors. That means a meaningful share of the equipment the US military relies on would become difficult or impossible to resupply or upgrade, with no other company able to fill TSMC's shoes.
Of course, TSMC is currently building fabs in Arizona; however, that is not sufficient. Those fabs are still under construction and will need several more years before they are fully operational, and even then, they will not run the most advanced nodes. That 4-5 year gap is exactly what will benefit INTC, the only other US company able to compete and produce chips domestically. This is not a theoretical argument either — the US government has already awarded INTC up to $3 billion under the "Secure Enclave" program (confirmed November 2024), designed specifically to incentivize leading-edge chip manufacturing domestically for defense and intelligence use. That is on top of billions more in direct CHIPS Act funding.
On the other hand, Intel Foundry lost $13.4B in 2024 and another $2.3B in Q1 2025 alone. Annualizing that Q1 run rate would put full-year 2025 Foundry losses at roughly $9.2B — a meaningful improvement on 2024, though still a large number — against a gross Capex guidance for 2025 of $18B. This is a real, ongoing cash commitment, not a rounding error, and it is the reason the stock trades below tangible book value. Because of that, I am not setting a hard price target, but rather laying out two scenarios that I believe can play out.
Base case — Foundry is unsuccessful
18A does not manage to land external customers, and the Foundry continues to burn between $10-14B per year indefinitely, funded by asset sales, government subsidies and grants, and an ever-growing debt pile. In that world, the Foundry business is worth close to nothing to INTC shareholders. However, as shown above, the legacy Products business is still worth roughly $40-43 per share — more than double the current market price. Eventually, either the board recognizes that continued Foundry investment is destroying shareholder value, or shareholder litigation forces the issue; either path unlocks the value sitting in the legacy business.
Bull case — Foundry is successful
If 18A ramps and Panther Lake ships in large volumes in H2 2025 as guided, and INTC lands external customers — even one or two — the Foundry stops being a liability and becomes an asset. In that case, on top of the $40-43 price floor from the legacy business, INTC could also see multiple expansion rather than just earnings normalization.
The asymmetry
What is worth noting is the asymmetry of this position. I believe the downside is currently limited, cushioned by a business that's already worth double its current price. On the other hand, we are paying next to nothing for a call option on one of the highest-profile, most-needed, and most government-backed industrial turnarounds in America. For these reasons, I believe INTC is a more reasonable way to play the currently frothy AI and semiconductor industry. That is why I am making it one of my largest positions.
Data as of this post: figures above are drawn from Intel's full-year 2024 earnings release (reported January 2025), Q1 2025 earnings release (reported April 24, 2025), and the Altera/Silver Lake transaction announced April 14, 2025. Entry price and date are my own trade record.