1. GlobalFoundries earnings at a glance
GlobalFoundries earnings for the third quarter of 2026, released on November 4, showed revenue of $1.899 billion and a non-IFRS gross margin of 31.4%, the highest in the company's history. The margin rose 540 basis points from a year earlier, driven by surging demand for optical networking chips used inside artificial intelligence data centers.
The quarter marks a clear turn for the contract chipmaker. For years, a large share of GlobalFoundries' factory time went to components for mobile phones, a market where volumes are high but pricing is tight. The latest results show that a different kind of customer is now setting the pace: companies building the networks that link thousands of AI processors together.
Revenue increased 6% from the second quarter and 13% from the same period in 2025. Those growth rates are solid for a manufacturer of mature and specialty chips, but the more telling figure is the margin. A 540 basis point improvement, equal to 5.4 percentage points, means the company is keeping substantially more of every dollar it brings in, a sign that its product mix has shifted toward higher-value work.
2. Where the record margin came from
One business line did most of the heavy lifting. Revenue in the communications infrastructure and data center segment climbed 81% year over year, and that segment now accounts for roughly 17% of GlobalFoundries' total sales.
The growth traces back to two specialty technologies: silicon photonics and silicon germanium, often abbreviated as SiGe. Both are central to the optical links that move data inside and between AI server racks, and both have been in heavy demand as cloud providers and AI developers expand their computing clusters.
The distinction between gross margin and revenue matters here. Revenue measures how much the company sells; gross margin measures how much is left after the direct cost of making those products. A foundry that fills its lines with commodity parts can grow sales without improving profitability. GlobalFoundries is showing the opposite pattern: modest top-line growth paired with a large jump in the share of revenue it keeps, which points to customers paying more for capabilities that are harder to source elsewhere.
The reported margin is a non-IFRS figure. GlobalFoundries prepares its primary financial statements under International Financial Reporting Standards, and its non-IFRS measures adjust those results to exclude certain items management considers outside normal operations. Companies use such figures to show underlying performance, and investors typically compare them alongside the standard reported numbers.
3. What is silicon photonics?
Silicon photonics is a technology that builds light-based data transmission components directly onto silicon chips, allowing information to travel as pulses of light rather than electrical signals. The approach lets data move faster and with greater energy efficiency between servers and across racks.
That efficiency has become critical in the AI era. Training and running large AI models requires thousands of processors to share information continuously. When those processors sit in different servers or different racks, the connections between them can become a bottleneck. Copper wiring loses signal strength and consumes more power as speeds rise and distances grow, so data center operators increasingly rely on optical connections to keep the processors fed.
Making optical components on silicon offers a practical advantage: it allows them to be produced in the same kind of semiconductor factories that make conventional chips, using established manufacturing techniques. That opens the door to higher volumes and tighter integration than older approaches to building optical hardware.
Silicon germanium is a related specialty material. By combining silicon with germanium, chipmakers can produce transistors that handle very high-speed signals, which makes SiGe well suited to networking equipment and to the transceivers that convert electrical signals into light and back again. In an AI cluster, silicon photonics and SiGe often work side by side within the same optical modules.
4. A customer base built around optical transceivers
GlobalFoundries now counts more than 50 silicon photonics customers. Among them are seven of the ten largest suppliers of optical transceivers, the modules that plug into networking switches and servers to carry data over fiber.
That level of penetration among the leading transceiver makers suggests the company has become a mainstream manufacturing option for this part of the AI supply chain rather than a niche supplier. Transceiver vendors sell into the same hyperscale data center operators that are spending heavily on AI infrastructure, so their production plans translate fairly directly into orders for the chips inside their modules.
The company also has a partnership with Marvell covering silicon germanium capacity at GlobalFoundries' facility in Vermont. Marvell is a major designer of data infrastructure chips, including components used in optical networking, and securing dedicated SiGe capacity with a manufacturing partner gives it supply for products aimed at the same AI build-out.
For a foundry, arrangements of this kind provide visibility. When a large customer commits to capacity at a specific site, the manufacturer can plan investment and staffing with more confidence, and the customer gains assurance that it will have access to production when demand peaks.
5. A $1 billion photonics target pulled forward by a year
GlobalFoundries raised its outlook for silicon photonics revenue in 2026 to $500 million. It also said it expects the business to reach an annualized revenue run rate of about $1 billion by the end of 2027.
That $1 billion milestone now sits a full year ahead of the company's earlier estimate. Bringing forward a target of that size indicates that orders and customer commitments have arrived faster than management originally planned.
An annualized run rate takes revenue from a recent period, typically a quarter, and scales it to a full year. Reaching a $1 billion run rate by the end of 2027 would mean the photonics business was generating roughly $250 million in a single quarter by that point, about twice the quarterly pace implied by the $500 million full-year figure for 2026.
For context, GlobalFoundries' total revenue in the third quarter was just under $1.9 billion. A photonics business at a $1 billion annual pace would therefore become a meaningful part of the company, rather than a side project, and one that carries the kind of pricing that helped lift margins to a record this quarter.
6. Moving away from smartphones toward data centers and cars
Management is deliberately pulling manufacturing capacity away from chips for mobile devices, a business that typically earns lower margins. The freed-up capacity is being redirected to two priorities: optical connectivity for data centers and custom silicon for the automotive industry.
The logic is straightforward for a manufacturer whose main constraint is factory capacity. Every production slot used for a low-margin smartphone component is a slot that cannot serve a higher-paying optical or automotive customer. Reallocating those slots changes the company's earnings profile even if total output stays roughly the same.
Smartphone chips have long been a volume business for contract manufacturers, and the market is crowded with competitors able to produce similar parts. Optical networking and automotive chips tend to involve longer design cycles, more specialized processes and closer collaboration with customers, which can make those relationships more durable and less exposed to price competition.
Automotive silicon also offers a different demand pattern from AI. Modern vehicles rely on a growing number of chips for power management, sensors, driver assistance and in-car connectivity, and automakers often want custom designs produced on proven, reliable manufacturing processes. By building positions in both AI data centers and cars, GlobalFoundries is spreading its bets across two markets with distinct drivers.
7. Tim Breen on AI momentum
Chief Executive Officer Tim Breen attributed the quarter's results to strong momentum in AI applications and to the company's focus on its chosen end markets.
His explanation fits the numbers. The segment tied most directly to AI infrastructure grew 81% while the company's overall revenue rose 13%, meaning the data center business expanded at roughly six times the rate of the company as a whole. That gap is what allowed a relatively small segment to have an outsized effect on the margin.
The comments also reflect the broader strategy behind the capacity shift. Rather than chasing volume across every possible chip category, GlobalFoundries is concentrating on areas where it believes its specialty technologies give it an edge, and where customers are willing to pay for that edge.
8. What's next: fourth-quarter guidance
Management expects the momentum to carry into the final quarter of 2026. GlobalFoundries guided fourth-quarter revenue to about $2.0 billion, with a non-IFRS gross margin of approximately 31.5%.
If the company meets that guidance, fourth-quarter revenue would rise from the $1.899 billion reported in the third quarter, and the margin would edge slightly above the record just set. That would indicate the third quarter's profitability was not a one-off, but part of a trend tied to the ongoing change in product mix.
Investors will watch several markers in the coming quarters: whether silicon photonics revenue tracks toward the $500 million target for 2026, whether the business stays on course for its $1 billion run rate by the end of 2027, and how quickly the capacity freed from mobile work is filled by data center and automotive customers.
9. Why the GlobalFoundries results matter beyond chipmaking
The AI build-out is often discussed in terms of the most advanced processors, but the GlobalFoundries results show how the spending is spreading through the rest of the hardware stack. Every large cluster of AI accelerators needs a network to tie it together, and that network depends on optical components made with specialty manufacturing processes.
For markets that track AI infrastructure demand, including the technology stocks and digital assets that often trade on the same risk appetite, results from suppliers further down the chain offer an additional read on whether data center investment is holding up. In GlobalFoundries' case, the latest quarter points to continued demand for the connections that link AI systems, along with a company reshaping its factories to serve that demand.