Skip to content
All case studies
Case 04 of 4Technology · Semiconductors2021–2026

Intel

Intel: can a chip designer also be the world’s chip factory?

Intel bet its future on manufacturing chips for others. The bet produced one of the largest annual losses in corporate history, a CEO change and a US government stake. Now AI demand is lifting its products. Should it keep the foundry?

Suggested for: MBA strategy · Corporate finance · PhD: vertical integration · Facts as of 16 November 2026

2024 net loss$18.8bnOn $53.1bn revenue
2025 revenue$52.9bnNet loss narrowed to $0.3bn
Q3 2026 revenue$16.5bn, +21%Data Center & AI +78%
Q3 2026 Foundry−$1.85bnOperating loss on $6.7bn revenue
01 · The situation

What happened

For decades Intel designed and manufactured its own processors, and its factories were the industry’s best. Over the 2010s it fell behind Taiwan’s TSMC in manufacturing technology, while rivals that only design chips — AMD, Nvidia, Apple — used TSMC’s factories to overtake it.

Intel’s response under Pat Gelsinger was to double down: spend heavily to regain manufacturing leadership and open its factories to outside customers as a foundry. The investment arrived before the customers did. Foundry losses ballooned, the company posted an $18.8bn loss in 2024, and the board replaced its CEO.

Since 2025 Intel has cut costs, taken new investors including the US government, and benefited from AI data centres buying more server processors. Its products business is recovering. The foundry, however, is still losing billions a quarter and has yet to win a large outside customer for its leading process — the central strategic question of the case.

02 · Timeline

Key moments

  1. Pat Gelsinger returns as CEO and launches IDM 2.0: catch up in process technology and build a foundry business for outside customers.

  2. Foundry operating losses reach $13.4bn; Intel reports an $18.8bn net loss. Gelsinger leaves in December.

  3. Lip-Bu Tan becomes CEO, cuts costs and the workforce sharply.

  4. The US government takes a ~9.9% stake for $8.9bn (converted CHIPS grants); SoftBank invests $2bn and Nvidia $5bn.

  5. Revenue $52.9bn, flat; net loss narrows to $0.3bn as costs fall.

  6. Q3 revenue +21% to $16.5bn on AI-driven server demand. Foundry still loses $1.85bn in the quarter; the Intel 14A design kit is released and customer commitments are expected in 2027.

03 · Exhibit
Exhibit 1 · Intel net income (loss)
$bn+1.72023-18.82024-0.32025

GAAP net income attributable to Intel. Q3 2026 alone: net income $3.3bn. Source: Intel Q4 and full-year 2024 and 2025 results; Q3 2026 results.

04 · Core tensions

The trade-offs at the heart of the case

Integration vs. focus

Owning factories can be a strategic advantage, but customers may not trust a foundry that also competes with them.

Capex timing vs. cash

Leading-edge fabs require years of spending before customers commit; AI demand competes for the same capital.

National interest vs. shareholder interest

Government ownership and subsidies support domestic manufacturing but can constrain options such as separating or selling the foundry.

05 · Discussion questions

For your seminar, paper or interview

  1. Q1

    Should Intel separate Foundry from its Products business through a spin-off, partial sale or joint venture? When, and why?

  2. Q2

    What would it take to win an external anchor customer for Intel 14A? Design the offer.

  3. Q3

    Capital allocation: AI data-centre demand needs supply now, while 14A needs capex for later. How would you split the next three years of investment?

  4. Q4

    What does a ~10% US government shareholding change about Intel’s governance, risk and strategic options?

  5. Q5

    Compare Intel’s path with TSMC’s pure-play model and AMD’s fabless model. Which theory of the firm best explains the outcomes?

Frameworks worth applying
  • Vertical integration and transaction-cost economics
  • Learning curves and scale economies
  • Real options in capital budgeting
  • Platform trust and neutrality
  • Stakeholder and state-capitalism perspectives
06 · Our view

How we would approach it

Simplifying Finance’s opinion, offered as a starting point for debate, not as fact or investment advice.

Intel’s products business is earning the right to fund a decision, not a forever bet. Our view is that Intel should give the foundry a clear, public milestone — a committed external anchor customer on 14A — and prepare the structural separation it will need either way.

  1. Set a customer gate

    Commit large-scale 14A capacity only after binding commitments from one or more external anchor customers, and say so publicly.

  2. Build the firewall now

    Run Foundry with separate governance, reporting and customer-data protections so outside customers can trust it, whether or not it is later spun off.

  3. Prioritise supply for AI servers

    Allocate near-term capacity and working capital to data-centre processors, where demand currently exceeds supply.

  4. Use partners for risk, not rescue

    Seek co-investment in specific fabs from customers and governments in exchange for capacity rights, rather than broad equity raises.

What this plan gives up

A customer gate could slow 14A and cede more ground to TSMC if customers wait to see the node work first. Separation planning may unsettle employees and complicate the government’s stated priorities.

The strongest counter-argument

The opposite view is that leading-edge manufacturing is a national-scale, all-or-nothing game: hesitation guarantees failure, so Intel should build ahead of demand and rely on public and strategic partners to bridge the losses.

Metrics to watch

  • Named external 14A customers and committed volumes
  • Foundry operating loss trajectory
  • Data Center and AI revenue and supply
  • Gross margin
  • Capital expenditure vs. operating cash flow
07 · Your turn

Propose your solution

Pick a discussion question and set out what you would do, why, and what it would cost. Strong, well-argued answers are reviewed and published below with your first name and programme.

Submissions are moderated and must relate to this case. Your name, programme and text are stored on our hosting provider and may be published; no email address is collected. Please don’t include personal contact details or confidential information.

Published solutions

Loading published solutions…

Sources

Figures are as reported by the companies or cited outlets as of 16 November 2026. Currency-neutral and reported figures differ; check the source before quoting. Corrections: let us know through the solution form.