Intel 2.0
A Plan to Rebuild American Exellence, One Wafer at a Time
The Cliff‑Edge Moment
Eighteen months ago Intel looked like the next Nokia: process‑tech lagging, x86 moats eroding, and a cash‑firehose foundry draining profits. In August 2023 I wrote “Intel: The Nokia in the AI Era” making exactly that case. Today—thanks to new leadership, High‑NA EUV audacity, and Washington’s industrial‑policy tailwinds—Intel has a narrow but real path to become America’s first sub‑2 nm foundry and an AI‑accelerator contender. This piece lays out that path, Intel 2.0, and stress‑tests whether it can work.
How We Got Here: A 24‑Month Whiplash
Links to the above events, in the order: 1 — 2— 3— 4—5
Glossary — Abbreviations at a Glance
Term Meaning
CHIPS US $280 B semiconductor‑subsidy law
High‑NA EUV Next‑gen extreme‑ultra‑violet lithography with 0.55 NA optics
IFS Intel Foundry Services (spin‑out target)
Turbo Cells Double‑height GAA standard‑cell library for extra GHz
SASF Secure AI Silicon Fund (draft US grant/pre‑buy program)
Spin the Foundry, Save the Products
Intel’s financial bleed comes from one place: Intel Foundry Services (IFS) lost $13.4 B in 2024 on minuscule external sales. Every other business—PC, server, networking, automotive—was profitable.
The fix is obvious but wrenching:
Spin IFS into a stand‑alone “FoundryCo,” capitalise it with government & private money, and let Intel Corp become a fab‑lite design house that can finally chase AI products without subsidising empty megafabs.
Deal Structure — Who Owns FoundryCo?
Table 2: Equity split & rationale
FoundryCo recruits an external CEO (think Samsung or TSMC alum);
Lip‑Bu Tan focuses on Intel Corp’s product slate.
Funding the 2025‑27 Buildout
Table 3: Cash needs vs sources
KPIs to Declare Victory by 2027
18A defect density < 0.1/cm² (Q4 ’25)
External wafer share ≥ 30 % (’27)
Break‑even at $10 B external sales (’27)
Valuing Intel Corp & FoundryCo
Intel Product‑Side Valuation (FY 24 pro‑forma)
Peer EV/EBITDA multiples (May ’25): Qualcomm 16×, TI 17×, AMD 29× → Anchor 18×.
$17 B × 18 = $306 B EV – $35 B net‑debt = $271 B equity (≈3× today’s cap) before any FoundryCo stake.
What 45 % of FoundryCo Could Be Worth (2027)
Table 4: Stake valuation scenarios
Add the base‑case $45 B stake to the $271 B product valuation and Intel’s sum‑of‑parts potential ≈ $316 B—about 3.6× today’s market cap. A bull‑case stake lifts total upside beyond $330 B.
2028‑30 Growth Optionality
Table 5: AI‑wave upside
Combined upside SOTP ≈ $495 B—about 5.7× today’s market cap—if Intel converts Gaudi into a double‑digit accelerator share and FoundryCo scales to $25 B external sales by 2030.
¹TAM: Mercury Research “AI Accelerators 2024‑30”.
Key execution levers:
Triton‑native Gaudi compiler & Hugging Face endpoints to erode CUDA lock‑in.
3‑D Foveros Direct packaging enabling sub‑600 W 8‑GPU tiles.
High‑NA wafer cost ≤ $1.5 K and gross margin ≥ 35 % by 2030.
Can FoundryCo Win AI Business?
Valuation upside only matters if Intel secures real AI customers—here’s why the pitch resonates.
1. Supply choke → Dedicated U.S. wafers
NVIDIA’s 2025 HBM backlog means >30 week lead‑times. FoundryCo can pre‑allocate entire High‑NA EUV lines to a single anchor customer. OpenAI’s rumored $7 B silicon budget equals ~25 k 14A wafers—about 4 % of a fab year—well within FoundryCo’s first‑site capacity.
2. Software moat erosion
CUDA is sticky because of kernel libraries and tooling. Gaudi already uses standard PyTorch ops; moving to Triton‑native lowers the “hello world” gap. Intel funds port bounties for top open‑source models (Llama, Mixtral). The goal: same script, change device = run.
3. Geopolitical insurance
Taiwan’s defense ministry openly warns that mass migration of leading‑edge nodes could invite coercion. U.S. clouds need a risk‑diversified foundry. A DoD‑brokered advance‑purchase agreement (part of SASF) would make FoundryCo the de‑facto secure node for classified AI workloads—commercial players then piggy‑back the volume.
4. Power & TCO edge
At 350‑400 W per GPU, a 20 k‑GPU cluster is >7 MW before cooling. 14A’s power‑delivery plus 2.5‑D EMIB‑T packaging reduce board losses and enable 800 Gb E coherence links. Early thermal sims show air‑cooled 700 W envelopes—no rear‑door heat exchangers needed. OPEX math swings ~$0.02/kWh in FoundryCo’s favor, a non‑trivial margin at scale.
Bottom line: a single anchor deals eliminates 50 % of FoundryCo’s ramp risk; each of these four levers is designed to make signing that deal a no‑brainer.
Swing Factors & KPIs
Why This Isn’t 2009 All Over Again
Skeptics often compare this spin‑off to AMD’s 2009 decision to separate its manufacturing arm into GlobalFoundries—an entity that failed to keep pace with TSMC. But Intel’s strategy with FoundryCo is fundamentally different in two critical ways:
1. Technological Starting Line.
GlobalFoundries launched with aging fabs and never caught up on FinFET or EUV. FoundryCo, by contrast, begins with a leapfrog asset: the first commercially assembled High‑NA EUV tool outside ASML. This positions Intel’s 14A and 10A nodes to lead rather than chase the pack. FoundryCo starts at the bleeding edge—exactly where GF never managed to get.
2. Strategic Policy Tailwinds.
GlobalFoundries had no industrial-policy tailwind at launch. FoundryCo benefits from massive public co-investment via the CHIPS Act and the prospective Secure AI Silicon Fund (SASF). These provide not just capital, but pre-committed demand from U.S. government buyers. FoundryCo isn’t just subsidized—it’s structurally de-risked.
Bottom Line
Intel walked to the edge of the Nokia cliff, peered over—and stepped back just in time. Intel 2.0 is high‑beta and high‑stakes, but it’s also the most credible turnaround blueprint the company has had since the original x86 boom. If Tan and an independent FoundryCo pull it off, Intel becomes a $300 B‑plus product company plus a strategic U.S. foundry stake. If they miss, the Nokia analogy will look prophetic. Either way, the next 24 months will decide whether America finally has a home‑grown answer to TSMC—and whether hyperscalers escape their GPU choke.
Afterword
This is an important and widely discussed topic—and for good reason. The passion for this legendary American tech giant resonates through countless articles. Spinning off the foundry business may be an act of tough love, but I believe it’s the right move. To truly rebuild America’s semiconductor ecosystem, we need an independent, world-class foundry—and at the same time, a revitalized, fabless Intel that can compete at the forefront of the AI era. Without separating these two ambitions, neither goal is likely to be achieved.
Update June 2nd, 2025: Intel eyes memory comeback with SoftBank-backed AI DRAM venture
Update July 8th, 2025: Intel's new CEO considers big chipmaking shift focus from 18A to 14A. This is likely a better plan than the one I outlined above for 18A, and it seems Tan is targeting the most advanced technology for the longer term. Clearly, when 18A is “losing its appeal to new customers,” those customers must also be at the cutting-edge frontier—my speculation being AI training and inferencing, the future generation of iPhones, or even I/O hardware for OpenAI.
Update July 25th, 2025: Intel Stock Tumbles After CEO Says ‘No More Blank Checks’ for Chip Foundry . This actually is a step towards spinning off its foundry business and it’s POSITIVE!
more links to the same guess about this spin-off:
6/13/2025, Monexa.ai: …”Adding to this strategic evolution is the ongoing speculation surrounding a potential spinoff of INTC's chip manufacturing arm, Intel Foundry Services (IFS). This move, widely discussed among analysts, is seen as a mechanism to unlock shareholder value and streamline operations by allowing INTC to focus on its core strengths: high-performance CPUs and AI chips Wall Street Journal. Such a separation could also make IFS a more attractive partner for external customers, including hyperscalers like Amazon (AMZN and Microsoft (MSFT, who are increasingly seeking diverse and reliable foundry services.”
12/16/2024, the CubeResearch : Time to Spin Out Intel’s Foundry Business
12/02/2024, Investopedia: An Intel Break-Up Could Follow Its CEO Shake-Up, Bank of America Says
Disclaimer
Financial data sourced from Intel’s FY 2024 Form 10‑K, Q1 2025 Form 10‑Q, and other public filings. Market capitalization and peer multiple data are as of May 1, 2025.
Although I’m a recent shareholder in INTC, this article is not intended as investment advice, analysis, or a recommendation. Please do not base any investment decisions on the content presented here.
Acknowledgments
Special thanks to friends for their thoughtful feedback—especially to Sam, whose honest disagreement helped sharpen the argument.
With the help of ChatGPT-o3.