I'm of the opinion that Cohu's recent upward price is driven by it recent (2024) entry into high band width memory testing done by its Neon tester for a physical inspection PLUS other ATE for electrical testing. This puts Cohu new system offerings in multiple footprints in the HBM fabs which as Morgan Stanley calls it a SUPER CYCLE IN MEMORY!
Morgan Stanley thinks the memory sector is in for a bigger and longer lasting high margin chip "Super Cycle" - this bodes well for growth in multiple new systems for Cohu.
Morgan Stanley pointed out that a new memory “super cycle” driven by AI has already arrived, and its intensity and logic are completely different from any previous cycle.
Morgan Stanley stated that, unlike in the past, this cycle is led by AI data centers and cloud service providers, whose customers have low price sensitivity, and inference workloads have become the main driver of general-purpose memory demand. Meanwhile, the latest channel checks indicate that in Q4, server DRAM contract prices have surged by nearly 70%, NAND contract prices are up 20–30%, and suppliers hold unprecedented pricing power.
The firm maintains Overweight ratings on SK hynix and Samsung Electronics, expecting rising memory prices to push share prices to new highs and for memory makers’ profits to significantly beat expectations.
Morgan Stanley noted that the core drivers of this cycle have changed in quality. The buyers are no longer traditional customers sensitive to price, but AI data centers and cloud-service giants engaged in an arms race to build compute infrastructure.
For them, securing memory is a strategic “must-have,” and price sensitivity has fallen to the lowest level. At the same time, HBM (high-bandwidth memory) production continues to structurally cannibalize traditional DRAM capacity. As Morgan Stanley emphasized in the report:
“The peculiar point is as follows: today’s memory demand has evolved into a competition led by AI data centers (compute-intensive platforms) and cloud service providers, and they are not as price-sensitive as traditional customers… The exponential upward trend in inference demand provides a solid foundation for this, and this is what makes this cycle fundamentally different from any previous one.”
According to Morgan Stanley’s latest channel checks, the DRAM price outlook turned sharply bullish in just two weeks. Q4 server RDIMM contract prices jumped about 70%, far surpassing the previous 30% expectation. The DDR5 (16Gb) spot price soared 336%, from $7.50 in September to $20.90 now. DDR4 prices also saw a 50% increase in offer levels. Most contract transactions will not be finalized until the end of this month, but customer acceptance seems inevitable—because they fear further price hikes and supply constraints.
Furthermore, NAND has fallen into a severe shortage:
NAND has become a key component of AI compute infrastructure and video storage. Prices of 3D NAND wafers (TLC and QLC) are expected to rise 65–70% quarter-over-quarter to respond to capacity constraints. Nearline storage specifications are shifting from 128TB to 256TB QLC SSDs. TrendForce forecasts that enterprise SSD bit-based server demand in 2026 will increase by about 50% year-over-year. Samsung’s bit production in the first half of 2025 was constrained by the transition from V6 176-layer to 321-layer V8 NAND, and in the second half it will ramp only gradually, resulting in this year’s bit shipment growth rate being limited to 10%.
There is still ample room for price increases, and the cycle is not at its peak yet
The market is often constrained by “peak fear,” assuming that once stock prices hit new highs a reversal will soon follow. However, Morgan Stanley emphasizes that in this AI-led market, the ultimate determinant is earnings growth, not historical valuation:
Currently, server DRAM prices are $1/Gb, whereas the peak of the cloud super cycle in Q1 2018 was $1.25/Gb. Given the scale of AI infrastructure investment and the dynamic nature of hyperscale customers, the peak price of this cycle is highly likely to surpass the previous high. Memory cycles typically last 4–6 quarters, and profits are in the process of being realized. The key, however, is comparison with market expectations—the market is showing noticeably greater enthusiasm for general-purpose memory prices. Valuation is not a predictor of future returns; it reflects supply and demand, not historical precedents.
Based on the strong long-term drivers of AI, memory price increases have already entered “no-man’s-land,” and earnings prospects far exceed the market’s general expectations. This implies there is still enormous upside potential in stock prices.
“As AI-related capital expenditures continue to expand, the share of memory in total spending is likely to keep rising—this will support P/B to far exceed historical peaks. In our view, this is a story where multiple expansion is layered on top of cyclical earnings recovery…
We believe analysts’ data revisions are always lagging—for SK hynix and Samsung, our 2026 and 2027 earnings forecasts are 31~48% and 38~51% higher than market consensus, respectively.”
In sum, the drivers of this memory “super cycle” are more durable, the magnitude of price increases has surpassed historical records, and earnings prospects are higher still. Coupled with a strong cycle, this creates a rare investment opportunity for memory manufacturers that hold pricing power.
Google AI comes up with excellent detail on just what the Neon tester does and why it is a integral test for HBM:
share.google
google.com
The Cohu Neon is an inspection and metrology platform used for physical quality control of High Bandwidth Memory (HBM) devices, not an electrical tester that directly measures bandwidth. It handles the physical high-volume demands of manufacturing for these high-bandwidth semiconductors.
High bandwidth demand in semiconductors, such as HBM for AI and high-performance computing (HPC), primarily refers to the electrical data transfer rate of the final packaged chip or wafer. The Neon system addresses the manufacturing challenges associated with these complex devices by ensuring physical integrity and quality at high speeds.
Cohu Neon and High Bandwidth Demand
- Inspection, not electrical testing: The Neon platform is an automated handling and vision inspection system, used to perform full 6-sided optical inspection and metrology (measurement) on tiny devices (dies or WLCSPs) at high speeds. It uses advanced vision systems, infrared inspection, and AI software to detect physical defects, not to test electrical signal bandwidth.
- Support for HBM manufacturing: Cohu has received orders for its Neon platform from leading memory manufacturers specifically for use with HBM devices. In this context, the "demand" is for a reliable inspection solution that can handle the specific physical form factor, fragility, and high throughput requirements of HBM components during manufacturing.
- Physical Metrology: The Neon system accurately and reliably measures thousands of micro-pillars across each HBM device, which are critical for the vertical interconnects in stacked memory. This precise physical metrology is crucial for the final assembly process to work correctly and achieve the high bandwidth required in the end application.
Electrical Bandwidth Testing
For actual electrical testing of high bandwidth, Cohu offers different Automatic Test Equipment (ATE) platforms and instrumentation, such as the Diamondx or Pax, which utilize specific instruments:
- High-speed digital and RF instruments: These testers use high-speed waveform digitizers and source instruments with wide analog bandwidths (e.g., up to 800 MHz or 8.5 GHz for RF measurements) to generate and analyze high-frequency signals.
- Physical layer testing: They perform physical layer testing using built-in PRBS (Pseudo-Random Binary Sequence) BERT (Bit Error Rate Tester) Transmitters/Receivers and high-bandwidth drive/compare memory for protocols like SerDes (Serializer/Deserializer).
The Neon ensures the physical components are correct so that the subsequent assembly and electrical testing of HBM chips can be successful.
Would you like to know more about the specific instrumentation Cohu uses for electrical bandwidth measurement?
What vision and AI capabilities does the Neon system use for HBM inspection?
What types of physical defects does the Cohu Neon system detect on HBM devices?
Tell me more about the instruments for electrical testing
13 sites
Neon - Next Generation Inspection Platform
Market Leading Inspection Yield with Uncompromised Throughput. Neon inspection & metrology platform can handle fragile wafer-level...
Cohu·
0:15
Cohu Enters HBM Memory Market with Neon Inspection ...
Nov 14, 2024 — ... Neon platform was selected by a leading U.S. memory and data storage technologies customer for inspection metrolog...
www.storagenewsletter.com
Cohu Secures Additional Neon Orders, Raises 2025 ...
Sep 15, 2025 — POWAY, Calif. --(BUSINESS WIRE)--Sep. 16, 2025-- Cohu, Inc. (NASDAQ: COHU ), a global supplier of equipment and servic...
Cohu
I think some serious short covering has moved Cohu UP!
|
