The Greensheet is Fusion Worldwide’s monthly report on trends that impact the open market supply chain. Our latest report details the changes in supply and demand influencing the integrated circuit, CPU, memory, GPU, and networking product industries.
- Pricing volatility impacts automotive components as constrained supply causes lead times to stretch for MOSFETs, IGBTs, and SiC components. MCUs see lead times expand beyond 99 weeks.
- PC market activity improves, but cost savings remains the major motivator for customers. However, finding prices lower than what manufacturers offer has proven difficult as OEMs hold the best cost-efficient offerings.
- AI has strengthened GPU demand and stabilized memory module pricing, but it has yet to impact the storage market positively. After a weak quarter, manufacturers plan to cut NAND production until demand recovers.
While some manufacturers are still managing excess, other companies continue to see shortages for components that are critical for the automotive, new energy, and AI. As companies shift manufacturing capacity to better support demand, it will still take time for these strategies to improve availability and bring lead times down.
Learn more about the market happenings in the full report below.
MOSFET Lead Times Extend, Impacting the Automotive, New Energy, and Storage Industries
Overall, the supply of MOSFETs has gradually improved for series with lower demand. However, there are still areas where supply cannot match demand, and lead times for manufacturers like Infineon and onsemi are on the rise. Items like SiC, IGBTs, and high-voltage MOSFETs are experiencing lead times of 60 weeks or higher. Prices for all components are on the rise because of the limited availability.
Due to the demand for these products, onsemi is shifting production focus to high and low-voltage MOSFETs, with a particular emphasis on components with new energy, automotive, and energy storage applications.
Despite attempting to bolster support for constrained products, the manufacturer has reportedly told customers that allocation will be challenging to secure for some time. The limited availability has pushed some customers to find alternative products or brands to ensure that production timelines are maintained.
Q2 Earnings Causes Pricing Volatility for STM Automotive Components as Shortages Persist
According to the STMicroelectronics Q2 earnings report, the company's automotive and power discrete segments saw double-digit growth and a 34.4% increase in revenue. Due to the strength in the automotive market, STM is expanding its focus on this industry, which has caused a shift in pricing volatility.
While heightened demand for automotive ICs has been ongoing, there has been a particular spike in activity for the VN series since early April. This series of high-side drivers, especially those with the prefix VNLxxx, VNNxxx, VNQxxx, and VNDxxx, has experienced numerous reports of shortages over the last year, and allocation remains tight. The constraints may extend to smaller tier-1 suppliers if availability does not improve. Outside of automotive, these components enable power management and protection against fault conditions for the following industries:
- Industrial: Supports efficient and reliable control of high-power loads in industrial machinery.
- Consumer: Help control power supply to various components, with use cases for products like home appliances, power tools, and audio systems.
- Battery Management Systems: Protect against overcharging and over-discharging.
MCU Shortages Set to Continue for NXP as Production Pivots Elsewhere
NXP’s supply of MCUs has undergone shortages throughout the year, with the MCIMX6 series starting in 2023 with lead times above 99 weeks. Supply has not improved, and prices have increased by 10% to 15% for the MCIMX6 series. This shortage will continue as NXP is shifting its production lineup and capacity to support the MCIMX8 series. Lead times for the MCIMX6 have come down but are still hovering around 87 weeks.
Additionally, supply constraints have lingered for automotive MCUs like the S912ZVL series due to increased demand for EVs. Products like the MC9S, MCF, and MK series have limited manufacturing capacity and allocation, causing supply constraints. NXP does not intend to increase production for older models like the S912ZVL, MC9S, and MCIMX6 as limitations on the number of chips that can fit on a wafer influence utilization rates.
This situation has caused unit prices and lead times to expand, with lead times sitting between 38 to 52 weeks. Certain pin packaging, particularly the TS18UHV and CLN40LP, may face supply constraints in the future because of the increasing costs. The S912ZVL12ACLF series, specifically with TS18UHV pin packaging, is in high demand and used in industrial control, AI, and other applications.
Central Processing Units
Intel Outpaces AMD in Desktop CPU Demand
While market demand for desktop CPUs is still low, there have been signs of improvement as activity within the PC market has increased since the beginning of August. Supply of Intel’s 10th Gen Comet Lake series is being controlled carefully due to the minimal spot demand, which has focused on the i3 and i5 processors because of their balance between performance and affordability. The i3 processors are suitable for more basic computing tasks, while i5 processors are more powerful and preferred for more demanding applications.
Customers are primarily searching for prices lower than what manufacturers offer, which is becoming increasingly difficult to find. These kinds of cost savings are limited because of price reductions and rebates available to OEMs, so they usually have the most cost-efficient offerings possible. Unless there is a sudden upside in demand, pricing is already at rock bottom, and supply is stable.
Alternatively, AMD’s activity has been even lower than Intel’s demand. There was some optimism in AMD’s estimates for order size, but forecasts were proven incorrect and were 20% lower than expectations. Due to the low demand and transactions, there has been an increase in rumors surrounding workforce reductions for AMD’s authorized distributors. Any changes would likely impact PC CPU partners.
Upward Trend for Intel Mobile CPU Prices
Transactions for Intel’s 11th Gen Tiger Lake mobile CPU are slowing down as costs continue to spike. Supply availability has dwindled after the EOL announcement in July, but despite the constraints, customers buying behavior depends on urgency due to the elevated pricing.
However, interest in the 12th Gen Alder Lake and 13th Gen Raptor Lake mobile CPUs is expanding. Customers looking to secure supply are still only placing orders based on cost savings, but overall supply appears healthy. The demand landscape will be more apparent towards the end of Q3 heading into Q4 once the Raptor Lake Refresh and Meteor Lake become available.
AMD and Intel Continue to Battle Over Server Market Share
AMD’s server market share has continued to expand, and the company’s CEO recently disclosed that AMD’s market share now exceeds 25%. This announcement came after confirmation that TSMC would produce the newest Instinct MI300 series generative AI accelerator, which will be released in Q4. Open market trends are currently favoring AMD’s Milan and Genoa series server CPUs. Demand mainly focuses on the following series:
- Genoa family: EPYC 9334, EPYC 9554, and EPYC 9654.
- Milan family: EPYC 7713 and EPYC 7763.
The adoption rate has increased for the Genoa series, likely due to the significant performance improvements it offers compared to its predecessors.
To regain its edge in the server segment, Intel is suspending special pricing for specific distribution channels to increase profitability. In particular, this strategy is impacting the older series of Intel Cascade Lake Refresh processors. While providing better discounts should help capture more customers, especially amongst Tier 1 clients, demand in the open market must first improve. However, this is likely the result of vendors maintaining lean inventory levels to avoid oversupply issues. Once stock has depleted, customers will have to stock up on parts once more, which should bolster demand.
With demand still not quite matching forecast expectations for both Intel and AMD, server CPU prices remain highly negotiable.
Storage Pricing Downtrend Proves Persistent, Leading to Production Cuts
Reductions made to NAND flash production earlier in the year have proved prudent as demand remains low. Prices have continued to erode, with companies like Samsung and SK Hynix reporting in the Q2 earnings calls costs for NAND have come down by as much as 10%. Both companies have stated that there will be adjustments to NAND production moving forward, at least until demand picks up once more. With the demand landscape remaining slow for consumer electronics, plus the sluggish destocking amongst Chinese smartphone brands, oversupply and further price drops around zero to 5% will likely materialize in Q3.
Additionally, this situation has had a ripple effect on SSDs. SSD pricing has declined as manufacturers and vendors try offloading inventory to cut losses. The number of orders for client SSD will likely dip below last year’s peak season volume, which may contribute to already increasing supply as production capacity has gradually resumed normal levels of output. Client SSD prices may fall by 3% to 8% within Q3 as end customers have cut down their forecasts, which has prompted distributors to make reductions to avoid excess inventory buildup.
Meanwhile, enterprise SSD prices should remain flat compared to Q2 due to the strength in this market. Forecasts indicate that the global procurement capacity of enterprise SSDs will increase by 10% sequentially if demand holds steady.
Memory Module Costs Start to Stabilize as DDR5 Supply Constraints Worsen
After a period of steady decline, memory module prices have begun to stabilize now that the supply and demand imbalance has shifted. It will take some time for costs to expand significantly. This process may be sped up by manufacturers cutting DRAM production. Prices will increase once supply in the open market depletes, and demand outweighs availability.
Meanwhile, DDR5 modules, especially higher-density products, still have supply issues. The trigger for the constraints has been mounting demand merging with the issue of there being no backward compatibility with certain server applications. Currently, the adoption of DDR5 for servers is predicated on the adoption of Intel Sapphire Rapids and AMD’s Genoa series. At present time, cloud service providers and AI server manufacturers have barely surpassed the double-digit line of use. With Sapphire Rapids, there is about a 7% to 11% adoption rate. By June of next year, adoption will likely grow closer to 50%.
Production is already struggling to keep up with order volume, even with the low adoption rate. As adoption rates increase and demand rises, manufacturers will have to ramp up capacity to satisfy demand. At the moment, allocation is being prioritized for Tier 1 customers, which could continue if supply remains limited.
Furthermore, manufacturers are reducing special pricing approval to maintain current price points. However, the ample open market availability is an obstacle to this strategy since customers can leverage alternative sources to find cost-saving opportunities.
GPU Demand Gets a Boost from AI
The AI market boom has brought significant demand to the enterprise GPU market, increasing competition for supply and pricing. Market prices have expanded rapidly, and existing order shipments have already seen delays. Due to this spike in the market, vendors receive less special project price approval, and allocation from manufacturers is currently limited.
This trend may have further implications for the consumer GPU market, particularly NVIDIA’s RTX4090, as there has been an uptick in smaller companies using these products for their own AI builds. Consumer GPU supply is currently healthy, as demand from the gaming industry has been minimal this year. Plus, there are cost savings opportunities for the RTX4090, even with the boost in order.
Regarding popularity, NVIDIA’s Tesla series GPUs, the A100, A800, and H100, remain the most in-demand and consequently have the tightest supply. The limited availability and intense competition for components have caused lead times to extend. As the A100 becomes increasingly challenging to find thanks to its EOL status, customers are being pushed towards the H100, which has better availability but still maintains lead times between 45 to 52 weeks.
With the launch date of NVIDIA’s new L40s fast approaching, the company may continue to encourage customers to move away from the Tesla A100 and A800s for AI-related applications. The current landscape of limited supply may mean that the manufacturer will take this opportunity to phase out the older series in time for the Q4 launch of the L40.
Networking Product Lead Times Expand Beyond 20 Weeks
The AI effect has spread to the networking product market, increasing demand for NIC cards, fiber optic cables, optical transceivers, and high-speed switches. Currently, customers are focused on higher-end parts, causing lead times to extend as production struggles to match mounting demand. Some of the quoted lead times for in-demand products are as follows:
- Certain manufacturers have extended Optocoupler series lead times to about 40 weeks.
- Fiber optic cables have seen lead times hover around 30 weeks.
Most booking lead times for high-speed products have extended to at least 16 to 20 weeks, at a minimum, and manufacturers are struggling to ramp up production capacity.
Stay ahead of shortages and changes to the electronic component supply chain by signing up for Fusion’s biweekly report, the Industry Insider, or reach out to a representative today.
