GlobalFoundries (GFS): Company Dossier

GlobalFoundries was assembled by acquisition. Abu Dhabi’s Mubadala funded AMD’s exit from manufacturing, then bought Singapore’s Chartered Semiconductor and took over IBM’s chip plants, creating a foundry with sites on three continents.

The company spent its first decade chasing the leading edge against TSMC and Samsung. In 2018, under new CEO Tom Caulfield, it abandoned 7nm development and refocused on what it calls differentiated, “pervasive” technologies: mature-node platforms with features such as RF, embedded memory, power and photonics that customers design into long-lived products. It sold non-core fabs, returned to profitability, and listed on Nasdaq in October 2021.

Leadership passed to Tim Breen in April 2025 after a multi-year succession plan; Tom Caulfield, CEO since 2018, became Executive Chairman. Breen joined GF in 2018 and was COO before becoming CEO. He previously held senior roles at Mubadala and was a McKinsey partner. Under him GF has leaned on acquisitions, IP licensing and government partnerships. Niels Anderskouv, named President and COO in the same transition, left on 2 March 2026; his duties were absorbed by Breen and the executive team.

GF is a pure-play foundry: customers design chips and GF manufactures the wafers. It serves more than 200 customers, from fabless chip designers to automakers. Since January 2026 it reports two revenue types. Manufacturing Services (about 89% of revenue) is volume wafer production. Technology Services covers engineering (NRE), masks, qualification, bump, test and packaging, and IP licences and royalties, a line that grows as the MIPS and ARC businesses scale.

New lines of business since 2025

Processor IP. MIPS ($215M cash, closed Aug 2025) plus Synopsys’ ARC line ($440M, closed Jun 2026) give GF RISC-V cores, NPUs, DSPs and processor-design tools to license, aimed at “physical AI” in cars, robots and industrial gear.²

Silicon photonics at scale. The Advanced Micro Foundry purchase ($453M, Nov 2025) adds a dedicated Singapore photonics fab; InfiniLink ($48M) adds an optical-transceiver design team in Egypt.^2,3

Power for AI data centers. A long-term GaN partnership with Navitas (Nov 2025) and the Photeon integrated-voltage-regulator purchase (Jul 2026) build on GF’s BCD and GaN processes.^1,3

Quantum. Quantum Technology Solutions (May 2026) manufactures parts for several qubit types, with Google, Microsoft, NVIDIA, PsiQuantum and Quantinuum among the named supporters.¹⁵

GF runs four manufacturing sites with more than 7,000 tools and about 2.8 million 300mm-equivalent wafers of installed annual capacity at end-2025. It also has design, engineering and support teams in India, Japan, Taiwan, Europe and the U.S., and a partner-operated line in China. About 14,000 people work for the company, roughly 70% of them engineers or technicians.⁴

U.S. investment program

In June 2025 GF raised its U.S. plan to $16B over ten-plus years: $13B to expand and modernize Malta and Vermont, plus $3B of R&D in packaging, silicon photonics and GaN.¹⁸ Public support behind it includes up to $1.5B in CHIPS direct funding (plus $75M for packaging), $570M from New York State, the 25% investment tax credit, and letters of intent for $300M (photonics) and $375M (quantum).^4,1 The October 2026 Japan initiative adds up to $21B of third-party capital for two more Malta fabs, with $955M of the CHIPS award redirected.¹⁶

Revenue peaked in 2022 during the chip shortage, fell through the 2023–24 inventory correction, and is now growing again on data-center demand while margins recover.

Gross margin rose from 15% in 2021 to 28% in 2023 as long-term agreements reset prices. It dipped to about 25% in 2024–25 as utilization fell and prices softened, then climbed back to 28.3% IFRS (29.9% non-IFRS) in Q2 2026 on a richer mix, lower depreciation and cost cuts. 2024’s net loss reflects a $935M impairment; 2025 returned to $888M of net income. Annual free cash flow has stayed positive, which has funded debt repayment, acquisitions, buybacks and, from July 2026, a dividend.

Foundry is in a two-speed upcycle. AI accelerators keep leading-edge capacity sold out and lift total foundry revenue about 25% in 2026. Mature nodes, GF’s market, are only now tightening after a two-year inventory correction, and they face a wave of new Chinese capacity from 2027.

The leading edge is where the money has gone. Nodes of 28nm and larger fell from 54% of pure-play foundry revenue in 2021 to about 36% in 2025, while their dollar revenue barely moved. GF competes almost entirely in that flat pool and the 12–20nm band just above it. Its growth case therefore rests on taking a larger share of the specialty niches inside it (RF, silicon photonics, SiGe, power, embedded memory) rather than on the pool itself growing quickly.

The valuation spread is the most telling part of this table. Tower, the peer most exposed to silicon photonics, trades at about 13x sales and 38x EBITDA; GF trades at 3.6x and 13x. Investors pay up for photonics exposure but still value GF mainly as a utilization-driven mature-node foundry. Closing part of that gap is the core of the bull case.

After four flat years (2021–25 revenue CAGR +1%), GF’s growth now comes from a narrow set of AI-linked businesses while smartphones, still a third of revenue, shrink.

Why management anchored these numbers

Revenue at 10–12% sits just above the market. GF’s served market is projected to grow ~10% a year to 2030. Management needs only modest share gains, concentrated in optical networking where its served market grows ~40%, while it models smartphones flat. Leaving its weakest segment out of the growth case makes the headline more defensible.

The margin bridge leans on levers GF controls. Roughly 5 of the 10 points from ~30% to ~40% come from mix alone. The rest comes from filling ~10 points of idle capacity in existing fabs, higher-margin services and IP, and productivity. Pricing and falling depreciation (D&A dropped from $1.57B in 2024 to $1.31B in 2025) are upside not explicitly counted.

“Exit 2028” is a run-rate, and it is ambitious. $4.00 of annualized EPS works out to roughly a $10.4B revenue run-rate at a 25% operating margin and a 15% tax rate on ~560M shares. That implies roughly 13–15% annual growth from 2025 to late 2028, faster than the 10–12% headline. Our Base case reaches about $4 of EPS in 2029, a year late.

Capital intensity was set before the Japan deal. The ~20% long-term capex target predates the October agreement, under which the U.S.–Japan vehicle owns the new fabs. If GF’s own capex eases, free cash flow could beat the 10% floor; we keep capex at 15–18% of revenue to stay cautious.

How the model is built

Revenue is forecast by end market from the FY2025 base, with 2026 anchored to first-half actuals and Q3 guidance. Gross margin, R&D and SG&A are scenario inputs calibrated against management’s targets; Japan-fab fees are a separate high-margin line. A full balance sheet and cash flow statement tie out every year, including the unwinding of customer prepayments, the dividend, buybacks at 50% of free cash flow, and the ~28M shares expected to be issued to the Commerce Department. The DCF runs to 2035, fading growth to a terminal rate and margins to a terminal level, and treats share-based pay as a real cost.

What the price implies

At $48.76, GF trades at about 18.5x 2027 consensus EPS, in line with UMC and TSMC and far below Tower. On management’s $4.00 exit-2028 EPS it is about 12x. Our Base DCF of $44.38 is 9% below the price, and the probability-weighted value is $43.69. The market is therefore paying for our Base case plus a partial chance that the Investor Day plan lands on time, which our Bull case values at $66.14.

Peer multiples give a higher range ($47.88–$79.80 on 18–30x 2027E EPS) because peers carry richer growth expectations, and the analyst mean target is $74.68. The gap between those and our DCF comes mostly from the discount rate and from capex staying high through 2030.

What would change the view: evidence that gross margin is on the 40%-by-2028 path (Q3 results on 4 Nov 2026 and the 2027 pricing reset); a co-packaged-optics production win; definitive Japan agreements confirming fee economics; or, on the downside, Chinese mature-node prices turning down again in 2027.