GlobalFoundries sees strong Q1 on data center demand

GlobalFoundries expects first-quarter revenue of around US$1.6 billion, exceeding Wall Street estimates, driven by demand for chips used in data centers.

It also announced a US$500 million share repurchase program, sending its shares up more than 7% in premarket trading.

The company has shifted focus toward AI infrastructure, producing specialized chips for high-speed data transfer between data centers.

In November, GlobalFoundries acquired Singapore-based Advanced Micro Foundry, which specializes in silicon photonics.

The company reported revenue of US$1.8 billion in Q4 2025, beating estimates, with adjusted profit of 55 cents per share, above the expected 47 cents.

It doubled silicon photonics revenue to over US$200 million last year and expects that figure to nearly double again in 2026.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

GlobalFoundries’ bet on silicon photonics targets a chokepoint in AI data centers

GlobalFoundries is changing course with its acquisition of Advanced Micro Foundry, aiming to ease limits on AI system growth.
As AI models expand, copper cables lose signal and draw more power at very high speeds such as 400 Gbps and above, which slows data movement 1.
Silicon photonics (SiPho) sends data with light pulses. LightCounting expects silicon photonics chip shipments to rise from 9.6 million units in 2024 to 45.5 million by 2029 1.
GF’s announcement and related coverage say the deal makes GlobalFoundries the world’s largest pure-play silicon photonics foundry (a contract chip manufacturer focused on this technology rather than designing its own chips) by revenue 2.

Beyond the node wars, photonics is changing what counts as a leading-edge fab

GlobalFoundries’ results fit an industry shift where gains come from more than transistor density alone.
After stepping back from leading-edge logic (the smallest and most advanced transistor manufacturing processes), the company is leaning into “specialty” technologies such as silicon photonics. Research on AI data centers says interconnect bandwidth (how much data can move at once), latency (delay), and power now limit scale more often 3.
That opens room for foundries outside the newest logic nodes to compete in AI infrastructure through photonics platforms plus integration work. The draft’s line about STMicroelectronics and “mature 300mm wafers” lacks support in the provided sources.
It also leaves space for the U.S. and Europe to lead in AI-adjacent technologies, backed by government incentives referenced in silicon photonics market analysis, while Asia keeps an edge in cutting-edge logic fabrication 4.