T. Rowe Price's Rizzo: The entire data memory industry is in an 'extreme shortage' right now

Key Concepts

• Memory Shortage: Current extreme shortage in both DRAM and NAND memory impacting the tech industry.
• Inelastic Demand: The current demand for memory is largely price-insensitive due to AI needs.
• Memory-Logic Coupling: The cyclical nature of memory and logic chip markets – they tend to rise and fall together.
• Semi Capital Equipment: Investing in companies that manufacture equipment for semiconductor production (e.g., ASML) as a way to benefit from the upcycle.
• Systems Architecture Approach: NVIDIA’s strategy of tackling problems by designing complete systems, potentially disrupting traditional storage vendors.
• High Bandwidth Memory (HBM): A type of memory offering significantly increased bandwidth, favored by the speaker’s firm.

---

Memory Market Dynamics & Investment Strategy

The discussion centers on the current state of the memory market, specifically the severe shortage of both DRAM and NAND, driven largely by the demands of the Artificial Intelligence (AI) revolution. Dom Rizzo of T. Rowe Price highlights that the industry is experiencing an “extreme shortage” across the entire memory supply chain. He notes that their firm has primarily allocated investments within the memory sector to SK Hynix due to their “leadership in high bandwidth memory.” This preference for HBM reflects its importance in supporting AI workloads.

The speaker emphasizes the increasing convergence of memory and logic chips, citing NVIDIA’s BlueField GPU announcement at CES as an example of this trend. This integration suggests a future where memory and processing are more tightly coupled.

Navigating the Boom-Bust Cycle

Rizzo addresses the historically cyclical nature of the memory market, acknowledging its “boom-bust cycles.” However, he argues that the current AI-driven demand presents a potentially prolonged upcycle. He describes the current demand curve as “completely inelastic,” meaning that customers are willing to pay whatever price is asked, indicating strong underlying demand.

Despite this, he cautions against immediate exuberance, explaining that the impact of recent price increases (50-80% quarter-over-quarter) on earnings will take “a few quarters to play out” due to pre-existing contracts negotiated six months prior.

Investment Recommendations: Semi Cap Equipment

Rizzo suggests that the most prudent investment strategy in the current environment is to focus on “semi capital equipment” companies, specifically mentioning ASML as an example. He believes these companies offer a more stable way to benefit from both the memory and logic upcycles, as they are “coupled” and move in tandem. He notes that ASML “have just started their move higher [and] still trade at relatively reasonable valuations.” This approach aims to avoid the volatility often associated with direct investments in memory manufacturers like SK Hynix and Samsung.

Concerns Regarding Traditional Storage Vendors

The conversation touches upon the implications for companies like Pure Storage and NetApp. Rizzo expresses skepticism about their ability to differentiate themselves, stating he has “always struggled with those names.” He believes NVIDIA’s “systems architecture perspective” – designing complete solutions rather than simply providing components – poses a significant challenge to traditional storage vendors attempting to “add value on top of that memory storage.”

Avoiding Core Memory Players & Prioritizing Responsible Navigation

Rizzo clarifies that his preference for semi-cap equipment stems from a desire to avoid the inherent risks associated with investing directly in core memory manufacturers (“they break your ankles so often”). He emphasizes the importance of “navigating all the different markets responsibly” and views semi-cap equipment as a way to participate in both the memory and logic markets simultaneously.

Notable Quote

“The issue right now in memory is we basically have a completely inelastic demand curve, right? It's whatever the price is, people are willing to pay it.” – Dom Rizzo, Portfolio Manager at T. Rowe Price.

Technical Terms

• DRAM (Dynamic Random-Access Memory): A type of volatile memory commonly used in computers.
• NAND (Not AND): A type of non-volatile flash memory used for storage.
• HBM (High Bandwidth Memory): A high-performance RAM interface for 3D-stacked synchronous dynamic random-access memory (SDRAM).
• Semi Cap Equipment: Short for Semiconductor Capital Equipment – the machinery and tools used to manufacture semiconductors.
• Logic Chips: Integrated circuits that perform computational functions.
• Systems Architecture: The conceptual model that defines the structure, behavior, and more views of a system.

Logical Connections

The discussion flows logically from identifying the current memory shortage and its drivers (AI) to analyzing the risks and opportunities within the sector. Rizzo’s argument for investing in semi-cap equipment is presented as a direct response to the cyclical nature of the memory market and the challenges faced by traditional storage vendors. The connection between memory and logic upcycles is consistently emphasized as a key factor in his investment thesis.

Data & Statistics

• Memory Price Increase: Memory prices have increased by 50-80% quarter-over-quarter.
• Contract Timing: Current demand is largely based on contracts written six months ago.

Synthesis/Conclusion

The primary takeaway is that the current memory shortage, fueled by AI demand, presents a unique opportunity, but requires a cautious investment approach. Dom Rizzo advocates for investing in semi-cap equipment companies like ASML as a way to benefit from the upcycle while mitigating the risks associated with the volatile memory market and the disruptive influence of companies like NVIDIA. He expresses skepticism towards traditional storage vendors and emphasizes the importance of responsible market navigation. The key to success lies in recognizing the coupling of memory and logic cycles and positioning investments accordingly.