Aehr Test Systems landed a major AI ASIC production forecast from a lead customer, with shipments scheduled to begin Q1 fiscal 2027 starting May 30, 2026. The company received multiple orders for new high-power Sonoma configurations capable of testing devices at up to 2,000 watts per unit.
Aehr forecasts $60M to $80M in bookings for H2 FY2026, driven primarily by AI wafer-level and packaged-part burn-in testing. The company expanded its partnership with ISE Labs and ASE for wafer-level and packaged-part testing services targeting top-tier semiconductor customers in HPC and AI applications.
The AI hardware sector is deploying next-generation architectures including 4-nanometer PCIe 6 chips and NVIDIA's Blackwell and Hopper GPU platforms. Advanced packaging facilities are coming online to support high-bandwidth memory technologies required for AI workloads.
Specialized AI accelerators are proliferating beyond general-purpose GPUs. Google's TPUs and reconfigurable dataflow units address specific computational patterns in machine learning training and inference. These architectures optimize for tensor operations and reduce data movement overhead compared to traditional processors.
Aehr reported Q2 FY2026 revenue of $9.9M, down 27% year-over-year from $13.5M, with non-GAAP gross margin at 29.8% due to lower WaferPak volumes. The company maintains $31M cash after raising $10M through its ATM program by selling 384,000 shares.
Production capacity exceeds 20 systems per month for both wafer-level and packaged-part testing, according to CEO statements. The Silicon Valley test lab received orders for Sonoma systems totaling $5.5M in Q3 to date, exceeding the entire Q2 total.
Credo Technology Group projects GAAP gross margins between 63.8% and 65.8% for Q3 FY2026, reflecting healthy economics in AI connectivity infrastructure. Ensurge Micropower is developing advanced microbattery technology for AI-enabled edge devices requiring compact power solutions.
Applications span enterprise AI data centers, autonomous systems requiring real-time inference, and edge computing devices with constrained power budgets. The infrastructure expansion addresses computational bottlenecks as model sizes and training datasets continue scaling.
Delayed WaferPak shipments worth approximately $2M for gallium nitride customers shifted from Q2 to Q3 due to high-voltage faults requiring protection circuit redesigns, highlighting technical challenges in next-generation power semiconductors.