Ray Wang: Memory chip prices are driving production cuts at Apple and Nintendo, AI demand is creating a supply-demand imbalance, and HBM production complexities are impacting profitability | Odd Lots

Ray Wang: Memory chip prices are driving production cuts at Apple and Nintendo, AI demand is creating a supply-demand imbalance, and HBM production complexities are impacting profitability | Odd Lots

Rising DRAM prices driven by AI demand could lead to higher costs for major tech companies.

by Editorial Team | Powered by Gloria

Key takeaways

  • Memory chip prices are affecting major companies like Apple and Nintendo, potentially leading to price hikes or production cuts.
  • The surge in DRAM prices is driven by AI demand, creating supply-demand imbalances.
  • Demand for memory is booming due to rapid demand acceleration and constrained supply.
  • The production of High Bandwidth Memory (HBM) is more complex than commodity DRAM, allowing for differentiation and better margins.
  • AI’s demand for DRAM is pervasive across training and inference, necessitating substantial memory resources.
  • The importance of memory in AI will grow as context windows expand, increasing hardware demands.
  • The smartphone market is expected to see a 10-15% reduction in outlook by 2026.
  • The market impact of price hikes will become evident in the second half of 2026.
  • Clean room constraints are a significant challenge for semiconductor supply this year.
  • There is a significant technological gap between Chinese and Korean memory producers, impacting competitive dynamics.
  • The current cycle driven by AI demand is expected to last until the second half of 2027.
  • The DRAM shortage is expected to worsen in 2027 due to increasing AI-driven demand.
  • Hyperscalers face challenges in negotiating long-term agreements due to the dynamic supply-demand environment.

Guest intro

Ray Wang is a Seoul-based analyst at SemiAnalysis covering AI accelerators and memory technologies including DRAM, HBM, and NAND. He recently co-authored the report “Memory Mania: How a Once-in-Four-Decades Shortage Is Fueling a Memory Boom,” which examines the surge in DRAM demand driven by AI workloads. Wang brings deep expertise in semiconductor market dynamics and has contributed analysis to major financial and technology publications.

The impact of memory chip prices on tech giants

  • Memory chip shortages are causing companies like Apple to consider raising prices or reducing production. – Ray Wang
  • Nintendo shares are getting hammered due to the memory chip shortage.

    — Ray Wang

  • The supply chain issues are significantly impacting the tech industry, particularly in relation to memory chips.
  • The surge in spot DRAM prices correlates with growing demand from the AI industry. – Ray Wang
  • The big surge in DRAM prices wasn’t until late last year.

    — Ray Wang

  • Companies are struggling to keep up with the rapidly accelerating demand for memory. – Ray Wang
  • On one side your demand is accelerating so fast on the other hand on the supply side your supply just couldn’t catch up.

    — Ray Wang

  • The memory market is experiencing a significant supply-demand imbalance, affecting pricing and availability.

The role of AI in memory demand

  • AI advancements are driving a surge in demand for memory, impacting both training and inference processes. – Ray Wang
  • You need tons of HBM and a lot of CPU DRAM for AI applications.

    — Ray Wang

  • AI requires substantial memory for processing and storing large data sets during training and inference. – Ray Wang
  • For training, you need tons of data to address, and for inference, you need one data after another.

    — Ray Wang

  • The importance of memory in AI will continue to grow as context windows expand. – Ray Wang
  • The memory importance will only increase, especially with expanding context windows.

    — Ray Wang

  • AI’s demand for memory is pervasive across various processes, indicating strong market demand.
  • The increase in AI server demand is significantly impacting DRAM market dynamics. – Ray Wang

The complexities of HBM production

  • HBM manufacturing is more complex than commodity DRAM, allowing for differentiation and better margins. – Ray Wang
  • HBM manufacturing involves more complex packaging, both front end and back end.

    — Ray Wang

  • The production dynamics of HBM and commodity DRAM are fundamentally different, impacting supply and profitability. – Ray Wang
  • On the same wafer basis, you can produce three more bits of commodity DRAM than HBM.

    — Ray Wang

  • By Q2 2025, there will likely be a significant shortage of both commodity DRAM and HBM due to supply constraints. – Ray Wang
  • Probably Q2 2025, people will realize we can’t find enough DRAM or HBM.

    — Ray Wang

  • The memory chip industry resembles a commodities market due to difficulty in product differentiation. – Ray Wang
  • It’s hard for a memory supplier to significantly differentiate their products.

    — Ray Wang

Strategic responses to supply constraints

  • Chip purchasing strategies involve securing contracts with suppliers based on demand forecasts. – Ray Wang
  • You negotiate a new contract pricing for the next quarter based on demand forecasts.

    — Ray Wang

  • The limited wafer capacity in chip manufacturing necessitates a focus on non-migration processes to increase DRAM bit production. – Ray Wang
  • Producers are migrating as much as they can to increase DRAM bit production despite limited wafer capacity.

    — Ray Wang

  • There will still be a significant shortage in chip supply this year despite efforts to increase production. – Ray Wang
  • Even with all factors considered, we’ll see a significant shortage this year.

    — Ray Wang

  • Producers are likely reluctant to build new fabs due to costs and potential lower prices. – Ray Wang
  • They’re reluctant to build new fabs because it means money out and lower prices.

    — Ray Wang

The future of memory market dynamics

  • The current cycle driven by AI demand is expected to last until the second half of 2027. – Ray Wang
  • We’re looking at this cycle lasting until the second half of 2027, which is rare.

    — Ray Wang

  • The key difference in the current super cycle is the emergence of AI as a new demand driver. – Ray Wang
  • There’s a new demand driver in this super cycle: AI.

    — Ray Wang

  • The DRAM shortage is expected to worsen in 2027 due to increasing AI-driven demand. – Ray Wang
  • 2027 will still see a DRAM shortage as demand increases.

    — Ray Wang

  • The market dynamics may stabilize by 2027. – Ray Wang
  • In 2027, we might see the market balance out.

    — Ray Wang

The influence of corporate spending on macroeconomic trends

  • The current pace of spending by major companies is creating a fiscal boom impacting macroeconomic indicators like CPI. – Ray Wang
  • The pace of spending is like a fiscal boom, impacting CPI.

    — Ray Wang

  • The crowding out effect from AI investments is leading to resource allocation shifts affecting consumers. – Ray Wang
  • AI investments are reallocating resources, affecting consumer goods.

    — Ray Wang

  • The increase in capital expenditures by hyperscalers is not solely driven by rising DRAM prices. – Ray Wang
  • Capex increase is not just because of DRAM price hikes.

    — Ray Wang

  • Hyperscalers face challenges in negotiating long-term agreements for memory pricing. – Ray Wang
  • Long-term agreements for memory pricing are difficult due to market dynamics.

    — Ray Wang

Technological gaps and competitive dynamics

  • There is a significant technological gap between Chinese and Korean memory producers. – Ray Wang
  • There’s a three to four-year technological gap between Chinese and Korean producers.

    — Ray Wang

  • The margin of commodity DRAM is currently higher than HBM, influencing production priorities. – Ray Wang
  • Commodity DRAM margins are higher than HBM, affecting production strategies.

    — Ray Wang

  • HBM will be a new growth driver for memory suppliers over the next few years. – Ray Wang
  • HBM is a new growth driver for the company, expected to last for years.

    — Ray Wang

  • The server DRAM and HPN markets will be top priorities for memory companies in the coming years. – Ray Wang
  • Server DRAM and HPN will be top priorities due to their market significance.

    — Ray Wang

The future of personal computing and AI integration

  • The future of personal computing may shift towards relying on external AI resources. – Ray Wang
  • It might make more sense to rely on external AI resources than powerful personal devices.

    — Ray Wang

  • The necessity for powerful personal devices depends on the specific tasks being performed. – Ray Wang
  • The need for powerful devices depends on task requirements, like video editing or document work.

    — Ray Wang

  • The current market resembles a commodity super cycle. – Ray Wang
  • The market is like a commodity super cycle.

    — Ray Wang

Ray Wang: Memory chip prices are driving production cuts at Apple and Nintendo, AI demand is creating a supply-demand imbalance, and HBM production complexities are impacting profitability | Odd Lots

Ray Wang: Memory chip prices are driving production cuts at Apple and Nintendo, AI demand is creating a supply-demand imbalance, and HBM production complexities are impacting profitability | Odd Lots

Rising DRAM prices driven by AI demand could lead to higher costs for major tech companies.

by Editorial Team | Powered by Gloria

Share

Add us on Google

Key takeaways

  • Memory chip prices are affecting major companies like Apple and Nintendo, potentially leading to price hikes or production cuts.
  • The surge in DRAM prices is driven by AI demand, creating supply-demand imbalances.
  • Demand for memory is booming due to rapid demand acceleration and constrained supply.
  • The production of High Bandwidth Memory (HBM) is more complex than commodity DRAM, allowing for differentiation and better margins.
  • AI’s demand for DRAM is pervasive across training and inference, necessitating substantial memory resources.
  • The importance of memory in AI will grow as context windows expand, increasing hardware demands.
  • The smartphone market is expected to see a 10-15% reduction in outlook by 2026.
  • The market impact of price hikes will become evident in the second half of 2026.
  • Clean room constraints are a significant challenge for semiconductor supply this year.
  • There is a significant technological gap between Chinese and Korean memory producers, impacting competitive dynamics.
  • The current cycle driven by AI demand is expected to last until the second half of 2027.
  • The DRAM shortage is expected to worsen in 2027 due to increasing AI-driven demand.
  • Hyperscalers face challenges in negotiating long-term agreements due to the dynamic supply-demand environment.

Guest intro

Ray Wang is a Seoul-based analyst at SemiAnalysis covering AI accelerators and memory technologies including DRAM, HBM, and NAND. He recently co-authored the report “Memory Mania: How a Once-in-Four-Decades Shortage Is Fueling a Memory Boom,” which examines the surge in DRAM demand driven by AI workloads. Wang brings deep expertise in semiconductor market dynamics and has contributed analysis to major financial and technology publications.

The impact of memory chip prices on tech giants

  • Memory chip shortages are causing companies like Apple to consider raising prices or reducing production. – Ray Wang
  • Nintendo shares are getting hammered due to the memory chip shortage.

    — Ray Wang

  • The supply chain issues are significantly impacting the tech industry, particularly in relation to memory chips.
  • The surge in spot DRAM prices correlates with growing demand from the AI industry. – Ray Wang
  • The big surge in DRAM prices wasn’t until late last year.

    — Ray Wang

  • Companies are struggling to keep up with the rapidly accelerating demand for memory. – Ray Wang
  • On one side your demand is accelerating so fast on the other hand on the supply side your supply just couldn’t catch up.

    — Ray Wang

  • The memory market is experiencing a significant supply-demand imbalance, affecting pricing and availability.

The role of AI in memory demand

  • AI advancements are driving a surge in demand for memory, impacting both training and inference processes. – Ray Wang
  • You need tons of HBM and a lot of CPU DRAM for AI applications.

    — Ray Wang

  • AI requires substantial memory for processing and storing large data sets during training and inference. – Ray Wang
  • For training, you need tons of data to address, and for inference, you need one data after another.

    — Ray Wang

  • The importance of memory in AI will continue to grow as context windows expand. – Ray Wang
  • The memory importance will only increase, especially with expanding context windows.

    — Ray Wang

  • AI’s demand for memory is pervasive across various processes, indicating strong market demand.
  • The increase in AI server demand is significantly impacting DRAM market dynamics. – Ray Wang

The complexities of HBM production

  • HBM manufacturing is more complex than commodity DRAM, allowing for differentiation and better margins. – Ray Wang
  • HBM manufacturing involves more complex packaging, both front end and back end.

    — Ray Wang

  • The production dynamics of HBM and commodity DRAM are fundamentally different, impacting supply and profitability. – Ray Wang
  • On the same wafer basis, you can produce three more bits of commodity DRAM than HBM.

    — Ray Wang

  • By Q2 2025, there will likely be a significant shortage of both commodity DRAM and HBM due to supply constraints. – Ray Wang
  • Probably Q2 2025, people will realize we can’t find enough DRAM or HBM.

    — Ray Wang

  • The memory chip industry resembles a commodities market due to difficulty in product differentiation. – Ray Wang
  • It’s hard for a memory supplier to significantly differentiate their products.

    — Ray Wang

Strategic responses to supply constraints

  • Chip purchasing strategies involve securing contracts with suppliers based on demand forecasts. – Ray Wang
  • You negotiate a new contract pricing for the next quarter based on demand forecasts.

    — Ray Wang

  • The limited wafer capacity in chip manufacturing necessitates a focus on non-migration processes to increase DRAM bit production. – Ray Wang
  • Producers are migrating as much as they can to increase DRAM bit production despite limited wafer capacity.

    — Ray Wang

  • There will still be a significant shortage in chip supply this year despite efforts to increase production. – Ray Wang
  • Even with all factors considered, we’ll see a significant shortage this year.

    — Ray Wang

  • Producers are likely reluctant to build new fabs due to costs and potential lower prices. – Ray Wang
  • They’re reluctant to build new fabs because it means money out and lower prices.

    — Ray Wang

The future of memory market dynamics

  • The current cycle driven by AI demand is expected to last until the second half of 2027. – Ray Wang
  • We’re looking at this cycle lasting until the second half of 2027, which is rare.

    — Ray Wang

  • The key difference in the current super cycle is the emergence of AI as a new demand driver. – Ray Wang
  • There’s a new demand driver in this super cycle: AI.

    — Ray Wang

  • The DRAM shortage is expected to worsen in 2027 due to increasing AI-driven demand. – Ray Wang
  • 2027 will still see a DRAM shortage as demand increases.

    — Ray Wang

  • The market dynamics may stabilize by 2027. – Ray Wang
  • In 2027, we might see the market balance out.

    — Ray Wang

The influence of corporate spending on macroeconomic trends

  • The current pace of spending by major companies is creating a fiscal boom impacting macroeconomic indicators like CPI. – Ray Wang
  • The pace of spending is like a fiscal boom, impacting CPI.

    — Ray Wang

  • The crowding out effect from AI investments is leading to resource allocation shifts affecting consumers. – Ray Wang
  • AI investments are reallocating resources, affecting consumer goods.

    — Ray Wang

  • The increase in capital expenditures by hyperscalers is not solely driven by rising DRAM prices. – Ray Wang
  • Capex increase is not just because of DRAM price hikes.

    — Ray Wang

  • Hyperscalers face challenges in negotiating long-term agreements for memory pricing. – Ray Wang
  • Long-term agreements for memory pricing are difficult due to market dynamics.

    — Ray Wang

Technological gaps and competitive dynamics

  • There is a significant technological gap between Chinese and Korean memory producers. – Ray Wang
  • There’s a three to four-year technological gap between Chinese and Korean producers.

    — Ray Wang

  • The margin of commodity DRAM is currently higher than HBM, influencing production priorities. – Ray Wang
  • Commodity DRAM margins are higher than HBM, affecting production strategies.

    — Ray Wang

  • HBM will be a new growth driver for memory suppliers over the next few years. – Ray Wang
  • HBM is a new growth driver for the company, expected to last for years.

    — Ray Wang

  • The server DRAM and HPN markets will be top priorities for memory companies in the coming years. – Ray Wang
  • Server DRAM and HPN will be top priorities due to their market significance.

    — Ray Wang

The future of personal computing and AI integration

  • The future of personal computing may shift towards relying on external AI resources. – Ray Wang
  • It might make more sense to rely on external AI resources than powerful personal devices.

    — Ray Wang

  • The necessity for powerful personal devices depends on the specific tasks being performed. – Ray Wang
  • The need for powerful devices depends on task requirements, like video editing or document work.

    — Ray Wang

  • The current market resembles a commodity super cycle. – Ray Wang
  • The market is like a commodity super cycle.

    — Ray Wang