GlobalFoundries and Monolithic Power: A Manufacturing Deal That Reveals Where the Constraint Has Moved

GlobalFoundries announced a long-term manufacturing partnership with Monolithic Power Systems on September 9, 2026, to produce Monolithic Power's proprietary power management chips at GlobalFoundries' 300mm facility in Singapore, with volume production expected in early 2027. The press release uses the standard language of these announcements — "expand capacity," "high-growth markets," "supply assurance." The headline reads like a routine foundry customer win.

The actual story is more structural. This deal maps onto a constraint that has quietly moved through the foundry industry. Monolithic Power is growing at 26% a year, much of it from AI data center power modules. It does not own fabs. And the foundries it has traditionally relied on are cutting the very mature-node capacity it needs to scale. GlobalFoundries is the alternative — a mature-node foundry with new Singapore capacity sitting at the intersection of a supplier who needs more wafers and a fab that needs more revenue.

The capacity constraint most investors are not tracking

The prevailing assumption among semiconductor investors is that foundry capacity is abundant. The conversation centers on TSMC building advanced nodes for NVIDIA. What receives far less attention is what is happening at the other end of the process spectrum — the 8-inch and 12-inch mature-node capacity where power management chips are made.

TrendForce reported in May 2026 that 8-inch utilization among the world's top 10 foundries is approaching 90% this year, up from roughly 80% in 2025, global 8-inch capacity is in negative growth through the first half of 2027, and TSMC is planning to reduce its 12-inch mature-node capacity gradually over the next one to three years. Foundries that still have mature-node slots are reallocating them away from display driver chips and CMOS image sensors and toward power management ICs, BCD processes, and power discretes — because power-related processes at 90nm and above command higher prices and better margins.

The order spillover from TSMC's cuts has already reached UMC and Vanguard, which raised prices in the first half of 2026. The full redistribution effect is expected to intensify after the second half of 2027, as new product tape-outs take nearly a year to reach mass production.

This is not a cycle that is about to turn. It is a structural tightening at the mature node.

What Monolithic Power is selling into

Monolithic Power Systems operates a "fabless-lite" model. It designs its own power management ICs using proprietary Bipolar-CMOS-DMOS process technology and installs that process on foundry partners' equipment. It does not use standard foundry offerings. This gives it control over performance and cost, but it also means it cannot simply place an order at any foundry — the process has to be qualified.

The company's revenue trajectory shows why capacity matters. Full-year 2025 revenue was $2.8 billion, up 26% from $2.2 billion in 2024. The second quarter of 2026 brought $981 million in revenue, beating consensus estimates of $903 million. Gross margins sit at 55%. Return on invested capital is 21%. The stock trades at roughly 74 times trailing earnings, priced as a growth story.

The growth is real and it is multi-market. Automotive revenue grew 43% in 2025, storage and computing grew 46%, and industrial grew 35%. But the highest-margin, highest-growth engine is the AI data center power supply chain — Monolithic Power's Intelli-Phase and Intelli-Module series that deliver power to GPU racks at 98% efficiency, supporting the transition from 12V to 48V power distribution in AI infrastructure.

A company with this growth profile, this margin profile, and a fabless model cannot scale without foundry capacity. Monolithic Power has been diversifying for years — it signed with Vanguard in 2022, and has operations in Malaysia and South Korea to reduce China exposure. The GlobalFoundries deal adds a third major foundry anchor, this time at 300mm scale in Singapore.

Why GlobalFoundries matters here

GlobalFoundries does not compete at the leading edge. It abandoned the 7nm race years ago and has built a business around "feature-rich" mature and specialty nodes — 22nm through 130nm CMOS, RF-SOI, power discretes, silicon photonics. The company holds roughly 4% of the global foundry market, behind TSMC at 73%, Samsung at 7%, and UMC at 4%.

The financial profile shows why it needs deals like this one. Fiscal 2025 revenue grew roughly 1.4% year-over-year to about $6.9 billion. Gross margin is 26%. Return on invested capital is 5%. The stock trades at 35 times trailing earnings with a market capitalization of roughly $25 billion.

But the company has been building capacity. The Singapore facility — a $4 billion expansion that opened in 2023 — represents one of the most significant mature-node capacity additions in the non-Chinese supply chain over the past three years. That fab needs volume. Utilization on new capacity is the difference between an investment that works and one that drags on margins.

The Monolithic Power deal places proprietary, high-performance power technology onto a fab that was built for exactly this type of work. Monolithic Power brings the technology and the revenue commitment. GlobalFoundries brings the wafers and the scale. The economics work because power management processes at GlobalFoundries' nodes are higher-margin than commodity CMOS, and Monolithic Power's revenue trajectory provides a revenue floor that improves the fab's utilization outlook.

The two companies this connects

This is not a symmetric deal. Monolithic Power needs GlobalFoundries more than GlobalFoundries needs Monolithic Power. The power designer has 26% revenue growth, 55% gross margins, and a supply chain risk that accelerates every quarter. The foundry has flat revenue growth, 26% gross margins, and utilization risk on a multi-billion-dollar Singapore expansion.

But the structural alignment is real. Monolithic Power's proprietary BCD process qualifies GlobalFoundries' Singapore fab as a long-term production node. GlobalFoundries gains a customer whose revenue is growing fast enough to matter at the foundry level — $2.8 billion in 2025, with each percentage point of Monolithic Power's foundry spend at GlobalFoundries representing meaningful incremental volume for a $6.9 billion foundry.

The valuation gap between the two stocks reflects their positions in the value chain. Monolithic Power at 74x earnings is priced for the growth and margin quality that a proprietary design business delivers. GlobalFoundries at 35x earnings is priced as a capital-intensive manufacturer with slower growth — though 35x for a foundry with 5% ROIC is not cheap either. The market has bid GlobalFoundries up 31% year-to-date, in part because of deals like this one that demonstrate demand for non-Chinese mature-node capacity.

The forward condition

Volume production begins in early 2027. The question for both investors is not whether the deal is signed — it is. The question is whether the capacity constraint tightens enough over the next twelve months to make this partnership a durable structural allocation, or whether the foundry industry's broader capex cycle produces enough new mature-node capacity to ease the squeeze.

If 8-inch utilization stays above 85% and TSMC's 12-inch mature-node cuts proceed as TrendForce projects, Monolithic Power's diversification into GlobalFoundries becomes a multi-year supply chain anchor rather than a one-quarter headline. If new capacity from Chinese foundries or other Tier 2 players absorbs the spillover demand, the urgency dissipates.

The key issue is not the partnership itself. It is whether the mature-node capacity constraint that created the partnership persists long enough for the capital investment in GlobalFoundries' Singapore fab to compound into real utilization and margin expansion. The deal makes sense for both companies on today's facts. Whether it matters for their financial trajectories depends on what the rest of the foundry industry does over the next two years.