Intel Bets on Manufacturing for Rivals and the Market Celebrates Prematurely
Last week, Intel's stock saw a notable upswing after the company announced a shift in its manufacturing strategy. The central message: Intel will not only produce its own chips but will open its industrial capacity to external clients, competing directly against TSMC in the contract manufacturing business. Wall Street interpreted this news as a move towards discipline and profitability. However, the situation is more complex than it appears.
The Logic Behind the Model Change
For years, Intel operated as an integrated model: it designed and manufactured its own processors under one roof. This was a competitive advantage while it maintained technological leadership. When that leadership began to erode against TSMC and fabless designers like AMD and Qualcomm, the integrated model became a burden: enormous fixed assets, monumental capital costs, and a structure that couldn’t adapt to the market's speed in demanding new manufacturing nodes.
Intel's response was to launch Intel Foundry Services, which positions the company as a chip manufacturer for third parties. The financial logic appears solid at first glance: if you already have the factories built, amortizing those assets with external clients improves return on capital invested. Fixed costs are distributed over more units, theoretically enhancing margins.
But this logic carries a critical assumption: that external clients will choose Intel over TSMC. That assumption is not trivial. TSMC has spent over a decade building not just manufacturing capacity but institutional trust with chip designers. Companies like Apple, NVIDIA, and Qualcomm don’t partner with TSMC because it’s the only option; they do so because TSMC's model is structured for their success. Intel, on the other hand, designs chips that compete with its clients.
This structural tension is the real blind spot in the strategy. A client bringing their designs to Intel Foundry is, at some level, sharing sensitive information with a company that also competes in the processor market. The sector has a long memory: the relationship between manufacturer and client in semiconductors relies on a categorical separation of interests. Intel has yet to resolve how to credibly eliminate this ambiguity for the market.
The Value Distribution That Nobody Is Calculating
This is where the market narrative becomes inaccurate. The stock rally reflects expectations of operational efficiency but does not adequately incorporate the power dynamics between Intel and the clients it needs to attract.
For Intel Foundry to operate successfully as a business, it needs clients that bring volume. The most attractive designers—those generating the highest revenue per wafer—are precisely the ones with the most negotiating power: they can demand lower prices, guaranteed timelines, and prioritized access to capacity. If Intel competes for those clients with aggressive discounts to capture market share from TSMC, it would effectively subsidize the manufacturing of its own competitors in the design market. Value would shift to the clients, not to Intel’s shareholders.
The alternative scenario isn't any more comfortable: if Intel does not lower prices and provide comparable conditions to TSMC, large clients will not switch. The only ones that might consider Intel are medium-sized designers with limited negotiating power or projects subsidized by the government under the CHIPS Act in the United States. This latter group is real and relevant, especially since federal financing reduces capital risk for Intel. But government subsidies are not a sustainable competitive advantage; they are a bridge. The question is, where does that bridge lead?
The unit economics of a foundry do not improve simply because you have clients. They improve when you have clients who pay enough to cover the capital costs of your assets, plus the expected return for shareholders. In cutting-edge semiconductors, building a new factory costs between $10 billion and $20 billion. The utilization rate of that capacity needs to be extremely high and consistent for the numbers to work. TSMC accomplishes this because, for decades, it has built a model where the success of the client is a condition for its own success. Intel is still trying to prove it can achieve the same without sacrificing its design business.
Manufacturing for Others Requires an Identity That Intel Has Yet to Consolidate
Intel's strategic move is not inherently wrong. Contract manufacturing makes sense as part of the industry's future, especially with geopolitical pressures to diversify the global semiconductor supply chain beyond Asia. The CHIPS Act, European incentives, and genuine governmental interest in having domestic manufacturing capability create a favorable context for Intel to capitalize on.
But a gap exists between having the opportunity and being positioned to capture it. Intel arrives at this pivot with a partially tarnished technical reputation after years of delays in its manufacturing nodes, with an organizational structure that still does not clearly separate the incentives of its design unit from those of its manufacturing unit, and with potential clients who have multi-year contracts with TSMC they won’t abandon for a transitional bet.
The market applauded the announcement because it simplified the narrative: Intel ceases to be a company losing to AMD and becomes a company selling infrastructure. This narrative is appealing. It is also incomplete. Infrastructure only generates value when the actors using it have structural reasons to prefer it over alternatives, and those reasons are not built through press releases but through months or years of impeccable technical execution and commercial conditions that make every client feel that the deal with Intel is better than any other available option.
The rally was a reading of the attempt. Real value will materialize, if at all, when Intel demonstrates that its manufacturing clients can achieve greater success within its network than outside of it. Until that moment, shareholders are purchasing the narrative, not the model.
