The Greensheet: January 2026

Key Themes

• DDR5 RDIMM Enters a Hyperinflation Phase: DDR5 RDIMM pricing has surged to unprecedented levels. 64GB modules now trade above $1,500, and 96GB units exceed $3,200, with weekly price resets becoming standard. Samsung, SK Hynix, and Micron are allocating over 75% of output directly to hyperscalers, leaving traditional OEMs and distributors scrambling in a seller’s market with NCNR terms and no visibility beyond Q1.
• Automotive DRAM Chip Supply Tightens as AI Allocation Accelerates: Micron, SK Hynix, and Samsung are devoting most of their resources to AI and server-class HBM and DRAM. AI-driven demand far exceeds current capacity, pushing automotive-critical DRAM further down allocation priorities through 2026. Automotive OEMs are already entering the market early to secure buffer supply.
• Nexperia Fragmentation Deepens: While limited domestic shipments from Nexperia China have resumed under yuan-based transactions, global OEMs, including automotive Tier 1s, are permanently qualifying On Semi, Diodes Inc., and Infineon as replacements. Authenticity concerns persist for parts produced after October 2025, and Malaysia-based production remains unreliable, with many high-runners still facing lead times into 2027. A hearing began on 1/14 in the Netherlands, and so far there appears to be no clear compromise or resolution.
• Intel’s 10nm Portfolio Becoming Less Viable: Intel is enforcing double-digit price hikes across its entire 10nm CPU stack (12th–14th Gen desktop, mobile, and server) effective Q2 2026, citing unsustainable economics. Small-core CPUs (N97, N305, N355) remain in short supply, with allocations being decommitted as far out as March 2026. Customers report Intel is now demanding 6-month forecasts just to secure minimal volumes.
• SSD & HDD Allocation Collapses for Non-Hyperscalers: Enterprise SSDs from Solidigm, Samsung, and Kioxia are now effectively reserved for top customers. Channel partners receive <30% allocation and face 20–70% Q1 price hikes. Similarly, 20TB+ HDDs from Seagate and WD are backlogged into mid-2026, while low-capacity drives (1–2TB) are being phased out entirely to force upgrades.

Product Updates

Integrated Circuits

• ADI implements 15% portfolio-wide price increase effective February 1, 2026, with military-grade MPNs seeing ~30% hikes. Lead times for key op-amps (e.g., OPA series) now exceed 40 weeks due to backend capacity limits and substrate shortages.
• TI signals significant price increases for new lead-time orders in Q1 2026, per multiple specialist vendors. Customers are rushing to lock in 2025 pricing before formal notices drop.
• Marvell and Realtek extend lead times to 32–52 weeks as TSMC redirects wafer capacity to AI memory. Mid-and low-end Ethernet switch ICs (e.g., 88E6320) are especially impacted.
• Sony industrial sensor ICs face decommits through Q1 due to end-of-fiscal-year demand surges in Japan, disrupting dashcam and machine vision builds.

CPUs 

Pricing Trends and Supply Constraints

• Intel confirms 10–15% price hikes on Server CPUs in Q2 2026, driven by AI business prioritization. SRV5F and SRWPD SKUs are in extreme shortage, with lead times stretching into April.
• AMD sells out all server CPU capacity through 2026, with Genoa/Turin lead times at 24+ weeks. Cloud providers are absorbing nearly all output; channel customers report zero allocation guarantees.
• Intel Alder Lake-N series (N97, N305, etc.) remains critically constrained, with March 2026 allocations already decommitted. Intel is deprioritizing low-margin SKUs to focus on Panther Lake/18A AI PC ramp.

GPUs

• NVIDIA RTX 5090 production remains minimal until April 2026 due to GDDR7 shortages. Spot prices have surged, and Nvidia is not accepting new POs.
• NVIDIA RTX PRO Ada workstation GPUs face lead times of 8–12 weeks as Nvidia prioritizes Blackwell B-series and RTX 6000 for AI deployments.

Networking

• Nvidia bundles CX8 NICs exclusively with GB-series GPUs, making standalone purchases nearly impossible. MCCX-7 availability remains erratic despite prior improvement signals.

Memory

• Samsung hikes DDR5 contract prices by up to 60%, forcing customers into weekly, non-negotiable NCNR agreements. 32GB / 64GB / 96GB and 128GB modules are now quoted daily, with prices rising 5–10% week-over-week.
• LPDDR4/5 and eMMC supply collapses for automotive. Wafer capacity is being reallocated to HBM/DDR5 for AI. Automotive-grade memory prices have surged, with brokers holding the only viable spot stock. eMMC allocation covers only ~15% of orders, with Q4 2025 seeing 20–30% price hikes on 4–8GB modules.

RDIMM

• Module suppliers require acceptance within 30 days of PO or risk losing allocation, highlighting extreme volatility. Daily pricing is now standard practice.
• AI server demand drives an acute shortage of 96GB/128GB modules. Samsung’s 96GB (Q-die) is slightly available, but 128GB remains extremely tight due to shared die constraints.
• DDR5 fulfillment rates plummet with Q1 pricing expected to rise 50–70%. Lead Times for new orders are on average 54+ weeks.

SSD

• Solidigm, Samsung, and Micron hike Q1 SSD prices by 20–70%, with enterprise SKUs like 7.68TB NVMe trading above $1,500. Allocation is restricted to hyperscalers.
• Kioxia BG6 NVMe SSDs are nearly unobtainable, signaling emerging NAND flash shortages. Samsung halts distributor sales of SATA SSDs entirely in January. As Memory manufacturers focus on keeping DRAM and HBM production high, NAND production has fallen.

HDD

• Toshiba is unable to support 20TB+ models (e.g., MG10ACA20TE); lead times for 10–24TB drives stretch into Q2 2026. Backorder pricing is not being honored.
• Seagate raises prices “almost daily” on low-capacity SKUs, releasing allocations in shrinking batches. 1TB drives are effectively EOL.
• Even major CSPs report HDD shortages for 16–24TB capacities, with Toshiba excluded from most qualified vendor lists.

Passives

• Panasonic announces significant Q1 2026 price hikes on tantalum capacitors, with full capacity booked through the quarter. Lead times now exceed 40 weeks.
• MLCCs from Samsung, Murata, and Taiyo Yuden remain at 24–28 weeks, exacerbated by upstream shortages of low-Dk fiberglass and HVLP copper foil, with both prioritized for AI server PCBs.
• Vishay MOSFETs (e.g., SQ2389CES) see 25-week lead times, with distributors reporting allocation cuts across resistors, capacitors, and discretes.

Supply Chain Trends 

• Trump invokes Section 232: A new 25% tariff on certain advanced computing chips that includes Nvidia’s H200 chip and AMD’s MI325X. However, chips imported to support building up the US technological supply chain would be exempt.
• China takes cautious stance on US H200 chip export: China is likely to accept Nvidia’s H200 artificial intelligence (AI) chips following US clearance, driven by immediate demand in the country’s vast tech sector.
• SK hynix has responded to rumors that the company is exiting the consumer business: SK hynix will still satisfy consumer demand through OEM channels, but balancing it with AI supply is becoming difficult.
• Micron addresses Crucial exit backlash: Micron warned that, despite breaking ground on new memory fabs, customers shouldn't expect to see meaningful output impacting memory supply until at least 2028.

Manufacturer News and Updates 

• Samsung Forecasts Record Q4 Profit: Driven by “tight supply and soaring prices” in memory, Samsung projects its highest quarterly earnings since 2022. New DRAM fabs in Korea remain on track for mid-2027 completion. [Source: Reuters]
• TSMC forecasts nearly 30% revenue growth: TSMC held its earnings conference call on January 15, 2026, and guided for the first quarter. The company expects revenue to range between US$34.6–35.8 billion. Gross margin is projected at 63–65%, while operating margin is expected to rise significantly to 54–56%. [Source: SeekingAlpha]

Explore our latest Market Intelligence on Agentic AI

A focused look at how Agentic AI workloads are reshaping demand for memory, GPUs, advanced packaging, and infrastructure—and what sourcing teams should be watching next. Read now.