Commerce's 4% Plan Tests GlobalFoundries' Independence

Four percent. That is the stake the United States reportedly plans to take in GlobalFoundries. It is too small to establish majority control. It is not too small to raise a question that a grant never did: what happens when the government owns part of a company whose competitors and customers it also wants to influence?

Federal semiconductor policy has become portfolio governance. Commerce is moving beyond grants toward minority claims across chip companies. The test is whether Washington can set clear rules for ownership rights, valuation and exit while leaving customers free to choose capacity on commercial grounds. Building the fabs is only one part of that test.

Commerce added ownership to the grant program

The CHIPS Act began with awards for specified capacity. The first grant, announced in December 2023, gave BAE Systems $35 million to expand production of chips used in fighter jets. In 2024, Commerce awarded GlobalFoundries $1.5 billion to build a Malta, New York, fab and expand an existing plant, alongside $1.6 billion in loans. The government could judge those commitments against facilities and production plans without holding the recipient’s shares.

By January 2025, officials had put 85% of CHIPS funding into binding deals. Binding is not the same as disbursed, but it described the program’s unit of progress: a funding commitment attached to an industrial project.

Commerce changed that unit in July 2026 when it signed letters of intent for $874 million in CHIPS funding tied to minority stakes in GlobalFoundries, Kepler Computing and five other companies. Seven letters are not seven completed share purchases. The roughly $125 million per company obtained by dividing $874 million by seven is not a disclosed allocation, either. The relevant change is contractual: public support would carry a prospective claim on company value as well as a claim on project delivery.

Intel made the ownership incentive visible

The United States’ Intel arrangement supplied a larger example. In August 2025, Intel said the government would buy 433.3 million newly issued shares for $8.9 billion, at $20.47 a share. Intel’s CFO later said the agreement was designed to deter a sale of Intel Foundry. If Intel’s interest in that unit fell below 51%, the government would have a right to buy an additional 5% of Intel at $20 a share.

That right is not a general veto over Intel’s operations. It does show why equity cannot be assessed like a construction grant. Washington acquired an interest in the company’s capital structure and in whether the foundry stayed attached to it. By April 2026, reporting put the government stake’s value near $36 billion, about four times the original $8.9 billion investment. That was a reported market value, not a realized return or proof that the foundry had won independent customer demand.

Share appreciation and foundry competitiveness answer different questions. A shareholder benefits when the stock rises. A buyer needs a process and a price that serve its product. Intel’s ownership terms made the first interest explicit; customer contracts remain the evidence needed for the second.

TSMC gave GlobalFoundries a customer test

On October 9, GlobalFoundries signed a five-year, $2 billion agreement to supply TSMC with silicon-interposer capacity from Malta. The contract preceded the October 15 report of the planned U.S. 4% stake. It is a concrete counterweight to the idea that public sponsorship has displaced customer choice. TSMC committed to capacity; the record does not show Washington directing that decision.

The agreement does not settle every commercial question. Five years of contracted capacity is not five years of completed deliveries or a disclosed profit margin. GlobalFoundries had reported first-quarter 2026 revenue of $1.63 billion, up 3% year over year, and forecast second-quarter revenue and adjusted earnings above estimates. Those operating results establish a business baseline against which new work can be judged.

Commerce also planned a $300 million CHIPS award for GlobalFoundries’ silicon-photonics research aimed at more efficient AI data centers. Research support and TSMC’s interposer order serve different purposes. One finances technical development; the other identifies a buyer for capacity. Combining them into a single measure of “demand” would make the public program look more commercially validated than the evidence permits.

Japan’s backing makes the New York project international

The October GlobalFoundries expansion report paired two planned Malta plants with up to $21 billion in backing from Japan and the planned 4% U.S. stake. The Japanese figure is an upper bound on proposed backing, not money already spent. Nor does the available evidence establish operational control or customer-direction rights for the United States through its minority position.

Other foundries are making cross-border capacity decisions of their own. Tower Semiconductor plans a $4 billion investment to establish its largest optical-communications-chip production hub in Japan. Taiwan’s planned U.S. semiconductor industrial parks likewise treat the surrounding supplier ecosystem, not just an individual factory, as an object of policy.

GlobalFoundries must reconcile Japanese support for plants in New York, a U.S. ownership interest and TSMC’s supply requirements. Those parties need not value the same outcome in the same order. A government may prize available domestic capacity during a disruption. TSMC must assess the capacity it contracted for. GlobalFoundries must execute both the buildout and the order. State backing can help finance scale-intensive manufacturing; it cannot substitute for the buyer’s technical judgment.

Minority positions still need majority-grade disclosure

Seven prospective smaller positions spread exposure beyond Intel. They also multiply the decisions Commerce may have to explain. The July letters of intent do not, by themselves, disclose the final share allocations, voting or purchase rights, valuation methods, conflict procedures or conditions for selling a stake. Lawyers and bankers have raised a prior legal question about whether the CHIPS Act authorizes converting grants into equity. Commerce cannot settle that question by making each holding a minority one.

Commerce can make the distinction between industrial policy and preferential treatment observable. It can disclose the rights attached to each holding, how it values shares received for public funds, who recuses themselves when funding and ownership decisions overlap, and what would trigger an exit. The same rules should be intelligible across recipients. Otherwise, a rival foundry cannot tell whether a public endorsement reflects a capacity judgment or an owner’s interest.

Four percent tells investors how much of GlobalFoundries Washington proposes to own. It does not tell TSMC, another customer or a rival foundry what Washington would do if commercial choice ran against its stake.