ARM Holdings (ARM.US) recently held its fiscal year 2026 Q2 earnings call, highlighting power efficiency as a critical industry bottleneck. In this context, the company emphasized that every player seeks the most energy-efficient computing platform. Arm’s architecture delivers approximately 50% higher energy efficiency than competing solutions, a claim validated by real-world performance, attracting major adopters like NVIDIA, Amazon, Google, Microsoft, and Tesla.

The surge in unprecedented compute demand has driven all publicly announced new capacity expansions to adopt Arm-based solutions, creating significant growth opportunities. This momentum contributed to a more than 100% year-over-year increase in its Neoverse business. Additionally, China delivered robust performance this quarter, with demand hitting record highs. Growth was primarily fueled by licensing revenue—including a major authorization deal—while royalties maintained steady growth. The company noted a healthy pipeline of pending licenses but deferred specific Q4 guidance to the next quarter. Large licensing deals typically take 6–9 months to finalize, with timing dependent on customer needs. Management expressed confidence in the outlook, citing sustained AI-driven capital expenditure.

**Q&A Highlights**

**Q: Arm’s strategic positioning in new AI data center projects?** *A:* Power constraints make energy efficiency paramount. Arm’s 50% advantage has secured adoption by industry leaders. All new compute capacity announcements are Arm-based, fueling growth and Neoverse’s doubling YoY.

**Q: Rationale for acquiring DreamBig Semiconductor?** *A:* DreamBig’s Ethernet and Arm DMA controller IP will expand networking capabilities and enhance end-customer product offerings.

**Q: Increased related-party revenue and SoftBank’s Stargate project collaboration?** *A:* Arm is deeply involved in Stargate, providing data center technologies (compute, networking, power distribution, assembly), unlocking substantial opportunities.

**Q: Opex growth and ROI timeline for new tech investments?** *A:* R&D-intensive solutions are being carefully managed, with revenue growth outpacing opex. Arm is evaluating all opportunities amid unprecedented demand. Product timelines will be disclosed post-silicon tape-out, sampling, and firm customer orders—a new business frontier for Arm.

**Q: SoftBank-related revenue details and impact on licensing?** *A:* Q2 revenue from SoftBank rose to $178M (+$52M QoQ), serving as a future benchmark. This includes lower-margin design services alongside IP licensing. The revenue stream may transition to chip royalties after ~1 year. While potential chip sales could displace some licensing income, iterative products may layer royalties. Stargate, involving OpenAI, is viewed as recurring revenue.

**Q: Stargate’s 1/3/5-year revenue potential and Lumex CSS royalties?** *A:* Compute demand has exceeded initial projections since January, with power infrastructure as the key constraint. Accelerated investments aim to capture this opportunity. Lumex CSS royalties emerged swiftly post-September launch due to prior adoption of Gen1 CSS, showcasing its time-to-market advantage.

**Q: Cloud/networking’s share of Arm royalties?** *A:* Infrastructure now grows twice as fast as other segments, likely reaching 15–20% of royalties (up from 10% last fiscal year). Arm’s penetration into DPUs, switches (e.g., Bluefield, Tomahawk, Arista) leverages its energy efficiency.

**Q: Chip demand and edge AI inference impact?** *A:* While data centers currently focus on training, inference will dominate. Strong edge demand for CPUs and Lumex (with scalable matrix extensions) supports efficient AI workloads. Arm’s edge strengths—battery-powered devices and cloud-edge synergy (e.g., Meta collaboration)—position it well for rising AI-driven edge demand.

**Q: China growth drivers (licensing vs. royalties) and H2 licensing outlook?** *A:* China’s record demand was led by licensing (including a major deal), with steady royalty growth. The licensing pipeline is robust, but Q4 guidance awaits next quarter. Deal cycles (6–9 months) hinge on customer timing. Confidence remains high given AI-driven capex.