Intel's stock has surged over 20% recently as investors anticipate a possible company split amidst a rumor suggesting that the Trump administration might push Intel and TSMC to form a joint venture that would take over Intel's manufacturing capacity. Four former directors of Intel wrote a column in the Fortune magazine explaining that this is a terrible idea and proposed spinning off Intel manufacturing to a separate company that would be owned by American investors instead.
Speculation has arisen that the new U.S. government might pressure Intel and TSMC to form a joint venture that would take over Intel's semiconductor production facilities, which are valued at approximately $108 billion. These facilities span multiple U.S. locations, including Arizona, New Mexico, and Oregon, with an additional site under development in Ohio. In addition, Intel has fabs in Israel and Ireland. The Trump administration has denied its involvement and said that it would not welcome a Taiwanese company taking over Intel fabs. TSMC's management also denied interest in taking control over Intel's manufacturing capacity.
The old plan is bad
Former Intel directors David B. Yoffie (a professor at Harvard Business School, has an extensive background in high-tech business strategy), Reed Hundt (a former chairman of the FCC), Charlene Barshefsky (a lawyer and a former U.S. Trade Representative), and James Plummer (a professor of electrical engineering and former dean of the School of Engineering at Stanford University, has a rich semiconductor background and holds 20 appropriate patents), wrote a column for Fortune magazine condemning the idea of handling Intel manufacturing capacity to TSMC and calling to spin it off and sell to a group of Western investors.
A TSMC-controlled semiconductor industry poses significant risks. Concentrating leading-edge semiconductor production in the U.S. under a foreign entity could weaken American technology firms by creating a near-monopoly, the former directors believe. While companies like Apple, AMD, and Nvidia rely on TSMC today, they still benefit from competitive pressure on the market that is created by Samsung Foundry and will be supported by Intel Foundry if the latter is a success. Intel presents little competition for TSMC as it has only two production nodes to offer. However, if Intel disappears completely, American firms could face reduced bargaining power, the former directors suggest.
A proposed plan
The former Intel directors think that a more strategic approach would be for Intel to separate its manufacturing division from its design business. In fact, they go as far as saying that the U.S. government should require the company to spin off its production operations and sell them to a Western private investor group. To make this viable, Washington should provide $10 billion in capital as non-voting equity, similar to the 2008 bank bailouts, ensuring taxpayers benefit if the venture succeeds. Additionally, major U.S. semiconductor firms, including Intel's design division, must commit to placing orders to guarantee profitability. The former members of the Intel board of directors believe these measures would make the business attractive to investors while preserving American semiconductor production capabilities.
On paper, spinning off Intel's manufacturing operations to a separate company sounds like a plausible idea. However, significant technological and business barriers will likely prevent this from happening smoothly, if at all.
Intel's semiconductor production fab chain costs around $108 billion (maybe a bit less if we subtract the cost of office buildings and other non-production assets). Finding an investor willing to buy these manufacturing assets for their value will be hard, to put it mildly, as $100 billion M&A transactions are not common.
Of course, it would be easier to persuade a consortium of investors to invest in Intel Foundry if a group of leading American companies would commit to using its services for years to come. However, committing to an unknown Intel Foundry is a huge risk for companies like AMD, Apple, Intel, and Nvidia, as this might slow down their progress and hinder their growth. Something that happened to AMD's commitment to GlobalFoundries back in the day.
Analysis
The proposal to handle Intel's manufacturing facilities to TSMC completely omits that Intel has its own process technologies — Intel 14nm (and older), Intel 10SF/10ESF, Intel 4, and Intel 3 — and already makes chips in high volumes. Also, the company is about to launch high-volume manufacturing of its Panther Lake and Clearwater Forest, with its 18A process technology, and get third-party 18A tape-ins in mid-2025. Discussions about spinning off Intel's fabs to a separate entity run by TSMC would hurt Intel's negotiations with potential 18A customers.
But let's assume that Intel and TSMC agreed to form a joint venture to run Intel's fabs., and that Intel keeps making chips on its Intel 4, Intel 3, and 18A process technologies, but TSMC would like to also make chips on Intel's capacity as well. Transferring TSMC's fabrication technology to Intel's U.S. factories with EUV capabilities would be highly complex. Although both companies use EUV lithography, their equipment setups, fab configurations, and vendor-specific modifications differ significantly. ASML's lithography tools, for instance, are custom-calibrated for each company. Intel and TSMC also have distinct etching, deposition, and process control methods, all of which influence production tools.
A joint venture would hurt TSMC's profit margins. The only reason TSMC might consider working with Intel is if a Trump administration introduced tariffs forcing it to manufacture its advanced 3nm and 2nm chips in the U.S. However, Intel lacks sufficient EUV production capacity in the U.S. to meet demand for both its own products and those of TSMC's clients. Instead, TSMC is accelerating its own Arizona fabs to support 3nm and 2nm production, which is a more viable business strategy, as industry analysts have suggested.
Even if TSMC was compelled to take control of Intel's capacity, its most advanced research would remain in Taiwan, and over time, it might close Intel's factories and lay off most of its workforce.
It should be kept in mind that Intel also has non-EUV capacity designed to produce Intel's own products. TSMC will unlikely be able to port its nodes to production lines meant for Intel 14nm (and older) and Intel 10SF/10ESF (at least, not in a financially viable way). These production capacities are worth billions for Intel as they produce sellable products. For an external company (whether TSMC or a group of investors), they will unlikely pose that much value as Intel would remain the only customer for those nodes and fabs. Meanwhile, at some point, those fabs will either have to be upgraded (which means additional investments), or shut down (something that the former members of the board mentioned above), which means writing off billions of dollars.
