Phison Posts First-Half Revenue Exceeding TWD 100 Billion: Strong AI Inference Demand Drives Order Visibility to 2027

Phison Electronics has released its latest financial results for the second quarter and the first half of 2026. Data shows that, driven by the transition of AI applications from model training to large-scale inference, the company's cumulative revenue for the first half of the year surpassed the TWD 100 billion mark for the first time, setting a new all-time high.

First-Half Revenue Breaks TWD 100 Billion Barrier, Marking Seven Consecutive Quarters of Record Highs

According to Phison's financial disclosures, the company achieved consolidated revenue of TWD 24.853 billion (approximately RMB 5.27 billion) in June 2026. This represented an 8.87% increase month-on-month and a substantial 301% surge year-on-year, marking seven consecutive months of record-high monthly revenue.
In the second quarter, Phison's single-quarter revenue reached TWD 67.888 billion (approximately RMB 14.406 billion), up 65.71% quarter-on-quarter and 279.47% year-on-year, extending its streak to five consecutive quarters of record quarterly revenue. Overall, Phison's cumulative revenue for the first half of 2026 totaled TWD 108.855 billion (approximately RMB 23.1 billion), a massive year-on-year increase of 243.08%, officially ushering the company into a new milestone of surpassing the "TWD 100 billion revenue" threshold.

AI Inference Drives Demand, Order Visibility Extends to 2027

Commenting on the robust revenue performance, Phison CEO K.S. Pua pointed out that the core driver is the continuous expansion of AI infrastructure. Currently, generative AI is gradually shifting from the early model training phase to the large-scale AI inference application phase. This trend has kept global demand for NAND Flash in a state of high growth, with market supply and demand remaining tight and showing no obvious signs of easing.

Pua further stated that order demand from Cloud Service Providers (CSPs) and AI-related clients remains steady. At present, the order visibility for some AI-specific projects has already extended into the first half of 2027. Against the backdrop of the AI era, the continuous accumulation of data is driving long-term growth in memory demand. Looking ahead, enterprise-grade memory infrastructure featuring high performance and high reliability will become a critical element in supporting the development of the entire AI ecosystem.