GlobalFoundries is a specialty semiconductor foundry that has quietly repositioned itself from a mature node laggard into the supplier of the physical layer for AI infrastructure, and the latest quarterly results are the moment that makes the repositioning visible in the numbers.
The most important recent development is the communications infrastructure and data center end market, which grew 62 percent year over year in the second quarter, the seventh consecutive quarter of double digit growth in that segment and the fastest since 2022. The mechanism is silicon photonics and silicon germanium, two high margin technologies that carry the optical interconnects of AI data centers and that GF manufactures at a node and process maturity where the pure play foundry giants have ceded the ground. Management raised its full year growth outlook for that end market to 50 to 60 percent, up from the high 30s, a raise that reframes the entire growth argument.
The tension sits in the mix shift that is simultaneously lifting the company. The dividend, the buyback, and the 45 percent gross margin target all rest on the assumption that the optical and photonics growth can scale fast enough to outrun the memory price squeeze that is now dragging smart mobile devices, the largest single end market at roughly 36 percent of revenue, into a low teens decline. If the data center ramp is a one time step change rather than a durable mix shift, the margin expansion thesis and the new capital return framework both lose their footing.
The catalyst is the long term model presented at the Investor Day in spring, paired with the third quarter print in early November. That print should show whether the guided gross margin holds as a floor rather than a peak. The model itself is the framework against which every number in this report should be measured, built around a 10 to 12 percent revenue CAGR and a long term gross margin target.
GlobalFoundries operates as a foundry for specialty and mature process nodes, manufacturing chips that the leading edge giants have walked away from. The company runs fabs across the United States, Europe, and Asia, with manufacturing services accounting for roughly 89 percent of revenue and technology services, a high margin segment built from IP licensing, RISC V processor IP, and the Synopsys ARC acquisition, making up the remaining 11 percent. The strategic pivot over the past two years has been to move up the value chain from a commodity wafer supplier into a partner that co designs and delivers software to silicon solutions for physical AI, automotive power, and data center interconnect.
The end market mix is the defining feature of the business. Smart mobile devices, historically the anchor, now represent about 36 percent of revenue, and the company has guided it to decline in the low teens on the back of memory pricing pressure. Communications infrastructure and data center, the fastest growing end market, reached roughly 16 percent of second quarter revenue. Automotive is about 19 percent, and home and industrial IoT another 19 percent. The mix shift away from the phone and toward the data center, automotive, and edge is the entire strategic thesis in one sentence.
The strategic context is further shaped by the CHIPS and Science Act funding that underwrites the company's expansion, including a 300 million letter of intent with the Department of Commerce to accelerate U.S. silicon photonics leadership and a separate expected 375 million grant to support a quantum technology solutions launch. These government awards are not incidental revenue; they are capital that de risks the build out of the photonics capacity that is the engine of the data center growth story, and they position GF as a trusted U.S. foundry in a period of escalating export controls and supply chain fragmentation.
The competitive context matters because the specialty foundry space is crowded. GF competes against UMC, Hua Hong, and TSMC's mature node capacity, and the differentiator is not cost but the depth of the technology stack and the design win pipeline. The company has taped out an ADAS radar for Bosch on its FinFET platform and secured three strategic chiplet design wins with Lockheed Martin, both of which signal that the company is winning work that requires a partner rather than a commodity supplier. The U.S. footprint and the defense and aerospace design wins are a moat that a purely offshore competitor cannot replicate under the current export control regime.
The core product is a wafer of silicon fabricated to a customer specification on a specialty process node, and the moat is the breadth of the technology stack combined with the depth of the design win pipeline. GF manufactures on FinFET nodes down to 22 nanometers and on a wide array of mature nodes, and the differentiating processes are silicon photonics, silicon germanium, bulk CMOS for RF and analog, BCD for power management, and GaN for high frequency power. Each of these is a process where the competitive field is thin and where the customer lock in is deep, because moving a design from one foundry process to another is a multi year re qualification effort. The company's 144 million second quarter R&D spend is a signal that the technology investment is being directed at extending that stack, not at competing with the leading edge logic nodes.
The silicon photonics and silicon germanium platforms are the products that matter most to the thesis. Silicon photonics integrates optical transceivers onto the same silicon as the compute, cutting the power and cost of the optical interconnects that bind AI accelerators together. Silicon germanium is the high frequency analog platform that enables the radio frequency front ends of the same data center infrastructure. GF is ramping capacity in both technologies and making the investments necessary to unlock the demand that customers have indicated, and the 300 million U.S. Department of Commerce letter of intent is a direct subsidy to that capacity build. The mechanism is that the photonics and SiGe processes are high margin, and they are the processes where GF has a real competitive position against the foundry giants, who have ceded the specialty ground to focus on the 3 nanometer logic race.
The RISC V processor IP stack, built from the MIPS acquisition and the Synopsys ARC Processor IP acquisition completed in June, is the other moat that is not a wafer. The two acquisitions together bring RISC V processor IP, software tools, custom design, and advanced manufacturing into a single offering, and the company now positions itself as a software to silicon partner for physical AI. The strategic significance is that GF is no longer selling wafers to design houses; it is selling a reference design flow that includes the processor IP, the software, and the fab, which is a far stickier relationship and one that carries a materially higher gross margin profile than the manufacturing services base.
The BCD and power platform is the third product family with real growth momentum. GF closed a first of a kind design win for smart power stage gate drivers on its BCD platform, a step into the rapidly evolving market for data center power delivery, and it completed the acquisition of the Photeon Technologies integrated voltage regulator business in July, adding IVR technology and specialized engineering talent to the power delivery stack. The automotive power design wins, including a significant 5 V and 10 V power management win on the BCD platform and an ADAS radar tape out for Bosch on the FinFET platform, are evidence that the power and sensing stack is extending into the physical AI and electric vehicle markets where GF has a credible process position.
The income statement in the second quarter is the cleanest expression of the mix shift thesis. Revenue came in at 1.786 billion, up 9 percent sequentially. The non IFRS gross margin expanded to 29.9 percent, a 470 basis point year over year increase that cleared the management gross margin target ahead of schedule. The operating margin on a non IFRS basis reached 16.7 percent. The non IFRS adjusted EBITDA came in at 587 million, a 32.9 percent margin that represents a meaningful improvement from the cycle low point. The net income print of 167 million is lower year over year, but that is an IFRS artifact driven by a compression in IFRS operating margin from higher R&D and SG&A, not a sign of deteriorating operating leverage.
The operating expense dynamics tell a more layered story. Research and development for the quarter was 144 million, up from a lower base a year earlier, and SG&A was 92 million. Total operating expenses of 236 million were up 16 percent sequentially, a step change that management attributes to the acceleration of core R&D investments to support the AI data center, physical AI, quantum computing, and advanced packaging growth opportunities. The 16 percent sequential opex step is the number to watch, because it signals that the company is choosing to invest ahead of revenue growth in the near term, a decision that is margin dilutive in the near term but is the cost of the long term model that the Investor Day framework is built on.
The balance sheet is the strongest part of the financial picture. Cash, cash equivalents, and marketable securities stood at approximately 3.3 billion at the end of the second quarter. Total debt was 1.1 billion, with a 1 billion revolving credit facility remaining undrawn. Net cash provided by operating activities was 405 million, and the non IFRS adjusted free cash flow was negative 3 million. That figure reflects the 408 million capex spend for the quarter, roughly 23 percent of revenue. The negative adjusted FCF is not a red flag; it is the planned consequence of the capex cycle that is building the photonics and SiGe capacity, and the 3.3 billion cash position gives the company the runway to sustain that investment without taking on debt.
The capital allocation framework is the newest financial development and the one that carries the most forward implication. The company paid its first ever quarterly cash dividend of 0.12 per share in mid July, and the board approved a second dividend of the same amount payable in October. The framework targets returning up to 50 percent of trailing twelve month non IFRS adjusted free cash flow after investments to shareholders through dividends and buybacks. Approximately 100 million remains under the share repurchase authorization. The dividend is a signal that management believes the cash generation is durable, and it is a commitment that constrains future capital allocation, because a company that has started a dividend is under pressure to maintain it even in a downturn. The full year adjusted free cash flow margin target of approximately 10 percent is the number that the dividend framework is anchored to, and it is the target that the second half capex spend has to clear for the framework to work as designed.
The forward path hinges on a single variable: the durability of the communications infrastructure and data center growth. Management raised its full year growth outlook for that end market to 50 to 60 percent, up from the high 30s. The third quarter revenue guide of 1.885 billion implies a sequential step that depends on the optical interconnect demand holding up. The mechanism is that the photonics and SiGe capacity is being built in response to customer demand that has been accelerating, and the execution risk is whether the capacity comes online fast enough to capture that demand, or whether the lead time on the photonics process leaves a gap between customer orders and actual shipments. A gap of even one quarter would show up as a sequential revenue miss and would reset the margin expansion narrative.
The second execution variable is the opex step. The 16 percent sequential increase in operating expenses in the second quarter is a deliberate choice to invest ahead of revenue growth, and management has indicated that second half operating expenses should be consistent with the third quarter guidance as the company accelerates R&D across AI data center, physical AI, quantum computing, and advanced packaging. The execution risk is that the opex step outruns the revenue growth, which would compress the operating margin even as the gross margin expands. The company is effectively trading near term margin expansion for a larger and more durable revenue base, and the Investor Day long term model is only credible if that trade works out over a full cycle rather than a single quarter.
The third execution variable is the smart mobile device mix. The company has guided the phone end market to decline in the low teens, driven by the memory pricing squeeze, and the phone is still the largest single end market at roughly 36 percent of revenue. The execution risk is that the phone decline is deeper or more persistent than the low teens assumption, which would pull total revenue growth below the 10 to 12 percent CAGR target and put pressure on the utilization numbers that the gross margin expansion depends on. The positive counterweight is the design win momentum in next generation smart mobile devices, including a notable BCD power management design win with MediaTek, which is a validation that the power platform is extending into the phone even as the overall phone mix softens.
The fourth execution variable is the government funding pipeline. The 300 million letter of intent for U.S. silicon photonics leadership and the expected 375 million grant for the quantum technology solutions launch are both letters of intent, not signed awards. The risk factors in the filings specifically flag the possibility that CHIPS and Science Act funding is delayed or withheld.
The primary risk is the data center demand ramp. The 62 percent year over year growth in communications infrastructure and data center in the second quarter is the single number that the entire repositioning thesis rests on, and the mechanism that produces that growth is the optical interconnect build out of AI data centers. If the AI data center capex cycle slows, or if the optical interconnect demand is met by competing silicon photonics suppliers, the end market growth rate decelerates and the mix shift that is driving the gross margin expansion stalls.
The second major risk is the memory pricing squeeze on the phone end market. The low teens decline guidance for smart mobile devices is a conservative framing of what is a structurally harder market, and the mechanism is that the memory pricing inflation is raising the bill of materials for handsets, which is reducing the number of units that customers are shipping and the number of wafers that GF is shipping. The downside risk is that the phone decline is a low teens percentage in the first half but a high teens or low twenties decline in the second half, which would pull total revenue growth below the 10 to 12 percent CAGR target and put the utilization numbers under pressure at a point when the company has just stepped up its opex. The phone is still the largest single end market, and a deterioration in the phone mix is the risk that the data center growth cannot fully offset.
The third risk is the opex step and the margin math. The 16 percent sequential opex increase is a near term margin headwind, and the execution risk is that the R&D and talent ramp for the AI data center, physical AI, quantum, and advanced packaging opportunities takes longer to convert into revenue than the Investor Day model assumes. In a downside scenario where the opex step holds at the second half level through the following year, the revenue growth comes in at the low end of the growth range, and the non IFRS operating margin could compress below the 15 percent level. That compression would put pressure on the dividend framework that is anchored to the adjusted free cash flow margin target. The dividend is the most visible commitment the company has made, and a dividend that is funded by capex rather than free cash flow is a dividend that is at risk in a downturn.
The fourth risk is the competitive and geopolitical backdrop. The specialty foundry space is competitive, with UMC and TSMC mature node capacity as the primary threats, and the export control regime is a tailwind that could shift. The mechanism is that the U.S. footprint and the defense and aerospace design wins are a moat under the current export control regime, but a relaxation of export controls or a shift in the U.S. industrial policy could reduce the strategic premium on the U.S. manufacturing footprint. The geopolitical risk is two sided: the CHIPS and Science Act funding is a tailwind, but the same political environment that is funding the photonics build is also the environment that is imposing the export controls that drive the data center customers to a U.S. foundry. A shift in that political environment could simultaneously reduce the government funding and reduce the strategic demand for U.S. manufactured wafers.
The valuation framework starts from the earnings, not the price. The stock trades at roughly 47 per share with a market cap near 26 billion, and the multiple to work from is the forward non IFRS earnings multiple rather than the trailing print, because the trailing IFRS EPS is distorted by the impairment charge and the non recurring items that have not yet cleared. On a forward non IFRS EPS basis, the stock trades at a multiple in the mid teens, a level that sits below the peer group for a company that is guiding to low double digit revenue growth and a gross margin that is expanding at a pace that the foundry peers are not matching. The anchor for the analysis is the revenue CAGR and the long term gross margin target, because those two numbers define the earnings path that the current multiple is pricing.
The bear case is a data center growth stall. If the communications infrastructure and data center end market growth decelerates to high single digits and the phone mix deteriorates faster than the low teens assumption, revenue growth comes in at the low end of the growth range and the gross margin stalls below the 30 percent level. In that scenario the non IFRS EPS trajectory flattens, and the stock de rates to a multiple in the low teens on a depressed earnings base. The bear case implies a share price in the high 30s, a drawdown of roughly 25 percent from the current level, and it assumes that the dividend is maintained but the buyback authorization is not deployed because the cash is needed for the photonics capacity.
The base case is the Investor Day framework coming in on plan. Revenue grows at the midpoint of the growth range, the non IFRS gross margin reaches the 30 percent level for the full year and continues expanding toward the long term target, and the adjusted free cash flow margin target is achieved. In that scenario the non IFRS EPS grows at a low double digit rate and the stock re rates modestly to a multiple in the mid to high teens on the stronger earnings base. The base case implies a share price in the mid 50s, a gain of roughly 15 to 20 percent from the current level, and it assumes that the dividend is raised modestly as the free cash flow margin improves.
The bull case is the photonics re rating. The scenario assumes that the silicon photonics and silicon germanium capacity ramps ahead of the current guide, that the communications infrastructure and data center end market growth extends above the current range, and that the gross margin expansion accelerates toward the long term target faster than the framework assumes. In that scenario the non IFRS EPS trajectory steepens, the adjusted free cash flow margin exceeds the target, and the stock re rates to a multiple in the high teens to low 20s on the basis that the company is a pure play AI infrastructure supplier rather than a diversified specialty foundry. The bull case implies a share price in the high 50s to low 60s, a gain of roughly 30 to 40 percent from the current level. The valuation conclusion is that the stock is fairly priced for the base case and carries asymmetric upside to the bull case, with the bear case providing a defined floor in the high 30s, and the dividend is the floor under the multiple because a company that has started a quarterly dividend and a 50 percent free cash flow return framework is under pressure to maintain that return even in a downturn.
The investment case for GlobalFoundries rests on a single question: is the mix shift from the phone to the data center durable, and is the market pricing it in fast enough. The second quarter provided the clearest evidence yet that the answer to the first question is yes, with the communications infrastructure and data center end market growing 62 percent year over year and the non IFRS gross margin clearing the 30 percent target ahead of schedule. The answer to the second question is less clear. The stock trades at a forward multiple in the mid teens, a level that is in line with the peer group for a company guiding to low double digit growth, which suggests the market has not yet fully repriced the photonics re rating that the Investor Day framework is built on.
The strategic shift toward the data center is the most underappreciated aspect of the story. The silicon photonics and silicon germanium platforms are the products where GF has a real competitive position against the foundry giants, and the 300 million U.S. government letter of intent and the design win pipeline for optical interconnects are evidence that the capacity build is being directed at demand that is real and accelerating. The RISC V IP stack, built from the MIPS and Synopsys ARC acquisitions, is the other underappreciated asset, because it transforms GF from a wafer supplier into a software to silicon partner and carries a materially higher gross margin profile than the manufacturing base. Together, these assets are the foundation of the 45 percent gross margin target, and they are not fully reflected in the current multiple.
The risks are real and material. The data center demand ramp is the single number the thesis rests on, the phone mix is a structural headwind that the data center growth cannot fully offset, and the opex step is a near term margin headwind that the Investor Day model assumes converts to revenue over a full cycle. The government funding pipeline is the most exposed variable to political and policy risk, and a delay in the CHIPS and Science Act awards would force the company to fund the photonics capacity out of its own cash flow. The balance sheet, with its 3.3 billion cash position and 1.1 billion debt load, provides the flexibility to navigate a downturn without making distressed decisions, and the dividend provides a cushion for investors who are not convinced by the photonics argument.
The judgment is that the stock is fairly valued for the base case and offers asymmetric upside to the bull case, with the bear case providing a defined floor in the high 30s. The company is a quality business with a credible repositioning story and a strategic shift that is not fully priced in. The primary monitoring items are the data center end market growth trajectory through the second half, the non IFRS gross margin trajectory relative to the 30 percent level, the opex step relative to revenue growth, and the status of the government funding pipeline. The third quarter print in early November is the next data point that should confirm or challenge the framework, and the full year 2026 adjusted free cash flow margin of approximately 10 percent is the number that the dividend framework is anchored to.