The Greensheet | October 2026
TI and ADI tighten further: TI’s quoted 12–16 week lead times are stretching closer to six months, with some customers seeing 52+ weeks and significant PO decommits. Another 10–15% TI price increase is expected in October. ADI lead times are now commonly 60–70+ weeks, with customers being asked to order at least six months ahead and reporting that previous price-lock and rebate terms are no longer being honored.
AI demand squeezes mature-node capacity: TSMC’s wafer plan is reportedly flat year over year, meaning new 2nm starts are displacing older-node production, while 28nm capacity is being redirected toward CoWoS packaging. This is contributing to severe shortages of Altera 10M08, 10M02 and 10M04. Broadcom PCIe Gen5 SS24 and SS26 are reportedly sold out for 2027 due to substrate constraints, with one customer seeing lead times of around 90 weeks and being advised to order up to two years of demand.
Memory allocation tightens as higher-capacity RDIMM pricing flattens: Samsung reportedly suspended October DDR5 quotes, with supply not expected to recover until mid-November, while October RDIMM increases are running around 12–25%. Micron SSD pricing is up 20–25% on average for October EMEA shipments, Samsung SSDs are up 14–15%, and Kioxia Gen5 is up 5–10% in Q4. Open-market pricing for 64GB, 96GB and 128GB RDIMMs has largely flattened, while 32GB continues to rise. Micron expects supply in 2028 to be tighter than in 2026.
Professional GPU demand remains exceptionally strong: NVIDIA RTX PRO 6000-class products, including NVIDIA 900-2G153-0000-000 and PNY kits and cards, led the three-week SO ranking. One large server OEM reports NVIDIA is accepting orders it cannot commit to supplying, quoting 26 weeks despite required ICs carrying 52-week lead times. PNY also reports no RTX PRO 6000 inventory and a US backlog exceeding 20,000 units.
Market Volatility Index
October tightness remains concentrated: GPU/accelerators, advanced packaging, HBM/RDIMM and high-density NOR remain at peak tightness, with conditions stable but showing no signs of easing. Enterprise SSD has moved from 9 to 10 and high-reliability MLCC from 8 to 9, while 800G/1.6T optics is expected to reach 10 within the next 30–60 days. The biggest shift is in memory, where HBM, high-density server DRAM and enterprise SSD supply is increasingly being allocated through strategic agreements and programme commitments. Contract pricing is rising into Q4 2026, making generic channel pricing a less reliable indicator of qualified infrastructure availability.
|
Indicator |
September Baseline |
October Reading |
Next 30–60 Days |
MoM Change |
Risk Status |
Intelligence Context |
|
AI GPU / accelerator platforms |
10 / 10 |
10 / 10 |
10 / 10 |
→ |
Critical |
Programme demand continues to pull through accelerators, HBM, fabrics, power and storage. |
|
Advanced packaging / leading-edge foundry |
9 / 10 |
9 / 10 |
9 / 10 |
→ |
Critical |
TSMC’s Q3 revenue rose 41.2% YoY; leading-edge demand and the 2nm ramp continue to command capacity. |
|
HBM / high-density DDR5 RDIMM |
10 / 10 |
10 / 10 |
10 / 10 |
→ |
Critical |
Micron has contracted the vast majority of 2027 HBM supply; DRAM remains supply-constrained into 2027. |
|
Enterprise SSD / high-density QLC NAND |
9 / 10 |
10 / 10 |
10 / 10 |
↑ |
Critical |
Enterprise SSD demand is forecast to exceed 80% YoY in 2026; 4Q price growth is expected to accelerate. |
|
800G / 1.6T optics, AECs & switch silicon |
9 / 10 |
9 / 10 |
10 / 10 |
→ |
Critical |
102.4T switching and 224G/lane connectivity increase design-specific release and qualification dependencies. |
|
High-density NOR (256Mb / ≥ Octal SPI) |
9 / 10 |
9 / 10 |
9 / 10 |
→ |
Elevated |
Limited effective supply and qualification time continue to constrain high-density grades through 2H26. |
|
High-reliability MLCC / AI power chain |
8 / 10 |
9 / 10 |
9 / 10 |
↑ |
Elevated |
AI-power transition and 4Q MLCC actions sustain targeted component exposure. |
Scores are a relative supply-pressure assessment (1 = low, 10 = extreme) based on Fusion industry research. They are not published manufacturer indices or price data.
October 2026: components most likely to delay an AI deployment if sourced independently
|
AI Infrastructure Component |
Supply Status |
Price Direction |
Primary Constraint |
Procurement Priority |
|
HBM / stacked memory |
Allocation Only |
↑↑↑ |
2027 supply is increasingly reserved under strategic customer agreements and platform commitments. |
Lock 2027 platform demand, stack preference and qualified vendor path; do not rely on spot substitution. |
|
Advanced GPUs / accelerator platforms |
Allocation Only |
↑↑↑ |
Platform scheduling combines foundry, package, memory, fabric and rack dependencies. |
Request allocation at complete-platform level and manage every critical BOM dependency together. |
|
Advanced packaging ecosystem |
Capacity Constrained |
↑↑ |
Leading-edge process demand and the 2nm ramp sustain priority-driven capacity allocation. |
Validate package route, node, substrate and test dependencies at design and forecast release. |
|
800G / 1.6T optics and switching |
Constrained |
↑↑ |
102.4T switches and 224G/lane scale-up fabrics raise optical, AEC, DSP and retimer design requirements. |
Release approved fabric BOMs early, including optics/AECs, switch silicon, NICs/DPUs and cabling. |
|
DDR5 RDIMMs / server memory |
Allocation Only |
↑↑ |
Server DRAM remains undersupplied as suppliers prioritise advanced-process capacity for higher-value products. |
Reserve configuration-level volume and pre-qualify density/module alternatives before build freezes. |
|
High-capacity enterprise SSDs |
Allocation Only |
↑↑↑ |
CSP inference, cache and vector-database demand has committed most incremental supply. |
Commit capacity, controller and NAND alternatives early; link storage buys to actual deployment milestones. |
What changed since September: DDR5 RDIMM moved from Critical to Allocation Only, and enterprise SSD price direction rose from ↑↑ to ↑↑↑. Operational risk now centers on the interaction between allocated memory and architecture-specific fabric components, not GPU availability alone.
-
TI lead times extend as decommits accelerate: Official Texas Instruments lead times of 12–16 weeks are running closer to six months in practice, with customers reporting 52+ weeks, MCUs at 40+ weeks, and some SN-series parts at 99+ weeks.
- TI pricing rises as expedite options disappear: Texas Instruments is expected to raise prices another 10–15% in October, although long-term agreement customers are expected to be exempt. Customers are being told expedites are unavailable and to source through the TI Store, while SN74L, TPS62 and TPS54 remain among the most requested and highest-risk families. TI has also reportedly discontinued production of some low-margin diodes.
- ADI lead times and pricing remain under pressure: Analog Devices lead times of 60–70+ weeks are now typical, with customers being asked to order at least six months ahead and some deliveries pushed into mid-2027 or 2028. Expedite and pull-in support is effectively unavailable.
- Broadcom PCIe switches face severe substrate constraints: Broadcom PCIe Gen5 SS24 and SS26 families are reportedly sold out for 2027 due to substrate constraints, with price increases above 10% indicated across the product line. Lead times are around 90 weeks, and customers have been advised to order up to two years of demand.
- FPGA supply tightens as prices rise: Altera 10M08, 10M02 and 10M04 are in severe shortage as TSMC reallocates 28nm capacity toward CoWoS packaging. AMD/Xilinx adaptive computing pricing increases on December 27, 2026, for both backlog and new orders, with distributors indicating around 20%, while expedites remain unavailable through Q1 2027. Lattice is also raising all device prices from November 1, with increases starting around 30% and some legacy products rising as much as 100%.
-
Discrete semiconductor lead times remain elevated: Diodes Inc. lead times are largely running at 48–52+ weeks, with some products at 60–74+ weeks and mid-size customers receiving only around 10% of requested Q4 allocation. Onsemi lead times have risen from 26+ weeks in early 2026 to 50+ weeks, with top customers receiving around 70% of requested MOSFET and discrete supply following October price increases.
- Power semiconductor shortages deepen: Vishay AI-related MOSFETs are at 104+ weeks and automotive MOSFETs at 52+ weeks, with customers expecting only around 10% of requested allocation over the next 4–6 months. Infineon TDA power stages are running at 80–90+ weeks, with prices up 10–50% over the past quarter. Nexperia lead times are at 36–40+ weeks following August price increases of 5–70%, with another increase expected in January.
- Marvell and STMicro lead times continue to extend: Marvell DSP lead times exceed 40 weeks, with 88E-series parts reportedly moving from 135 to 270 days and pricing up around 20% since mid-September. STMicroelectronics is quoting 52+ weeks on MCUs, with automotive customers specifically raising concerns over availability.
-
Intel server CPU constraints extend into 2027: Demand remains heavily concentrated around the latest high-performance server platforms, while supply continues to be limited by key ecosystem components rather than processor production alone. Distributors report ongoing allocation challenges, with availability favoring premium AI-focused configurations. Industry participants expect supply conditions to remain tight through 2027, while pricing pressure persists amid strong demand and constrained capacity across the supply chain.
-
AMD server demand outpaces available capacity: AMD is guiding to more than 80% server growth in H2 2026 and more than 70% in 2027, but supply remains constrained by TSMC capacity. EPYC pricing is expected to rise 10%+ in Q4, likely around mid-October, while special pricing is increasingly limited to top customers. Lead times range from three weeks to six months, with Turin F-series around 3–4 months and Turin the highest-risk generation. Distributors also expect further Q1 2027 increases of around 10% for Genoa and 15% for Turin.
- RTX PRO 6000 shortages persist despite strong demand: Supply conditions remain tight across server, workstation, and mobile variants, with channel inventory reported as largely depleted. Lead times continue to stretch as demand outpaces production, and buyers are facing mounting order backlogs. Market participants indicate that availability through authorized channels is unlikely to improve materially before year-end.
- NVIDIA H200 allocations shift into 2027: Supplier allocation for H200 has reportedly moved from November 2026 to March 2027 as production shifts away from individual cards toward HGX and GB300/VR systems. Lead times are being quoted at 40+ weeks, while open-market pricing is running at around twice OEM-direct levels.
-
RTX 5090 channel pricing reaches extreme levels: RTX 5090 e-tail pricing has climbed above $8,000, with some available inventory in the $6,000–$7,000 range. One major US retailer is also listing the card as no longer stocked.
- Samsung DDR5 allocation tightens as pricing moves higher: Supply conditions for DDR5 continue to firm, with buyers reporting reduced allocation visibility and increasingly selective availability. RDIMM pricing remains on an upward trajectory, supported by strong AI server demand and a tightening supply environment.
- Micron locks future capacity as legacy DRAM supply contracts: Micron expects 2028 supply to be tighter than 2026, with long-term agreements absorbing almost all 2027 and 2028 capacity and DRAM bit growth limited to the low-20% range. Micron is proceeding with DDR3 EOL and associated supply decommits, while 16Gb industrial DDR4 pricing has increased another 20–30%.
- SK hynix shifts toward direct allocation: SK hynix is expected to stop supplying RDIMM modules through distributors in 2027 and move toward direct customer engagement. The company reportedly plans to meet only 30–50% of projected 2027 DRAM chip demand.
- RDIMM pressure concentrates on lower capacities: Open-market DDR5 RDIMM pricing continues to rise for 32GB, currently the highest-demand module, while 64GB, 96GB and 128GB pricing has largely flattened. Supply is expected to tighten again in Q1–Q2 2027 as manufacturers increase HBM production, which could consume around 50% of DDR5 wafer capacity in 2027.
- Legacy automotive memory remains constrained: Supply of mature automotive memory technologies, particularly LPDDR4 and DDR3, remains tight, with pricing continuing to trend higher. While some segments are showing signs of stabilization, long-term supply constraints and limited industry investment in legacy nodes are expected to support a firm market environment through the coming years.
-
SSD pricing continues to rise: Major suppliers are implementing broad price increases as demand remains strong and supply discipline tightens. Enterprise SSDs face the greatest pressure, while procurement is becoming more closely tied to broader memory purchasing. Product transitions and phase-outs are also reducing availability of select enterprise SSD families, supporting expectations for a firmer market into 2027.
-
Enterprise SSD allocation favors hyperscalers: Micron 30TB and 61.44TB enterprise SSD lead times have extended to 26–36+ weeks, with supply reportedly fully allocated to leading hyperscalers and limited 2027 support available for smaller customers. Smaller-capacity SSD availability is also tightening as NAND is prioritized toward higher-capacity drives, creating shortages of server boot drives.
-
High-capacity HDD shortages continue to worsen: Supply of high-capacity enterprise HDDs remains extremely tight, particularly above 20TB, as hyperscale demand absorbs much of the available capacity. Availability is limited across major suppliers, lead times remain extended, and pricing continues to move higher. With supply-demand imbalances expected to persist into 2027, further pricing pressure is likely in the coming quarters.
-
NAND and NOR emerge as major 2027 supply risks: NAND is becoming a significant 2027 supply concern, with Micron NAND pricing up 30% last quarter compared with 17–18% for DRAM. 2027 eMMC allocation remains uncertain as telecom eMMC moves toward EOL, while small-capacity NOR is not receiving full H1 2027 allocation and industrial 1Gb and 2Gb products are seeing significant decommits. Macronix is experiencing last-minute delays, while Micron parallel NOR families QLHP, QLJP, QLKP, and QLLP have a last-order date of September 29, 2027, and a final shipment date of September 29, 2028.
-
Optical transceivers emerge as a key AI networking bottleneck: Lead times for high-speed optical interconnects continue to extend as AI infrastructure deployments accelerate. Supply constraints across transceivers and supporting networking components are driving allocation challenges, prompting some buyers to explore alternative sourcing channels to secure capacity.
- DSP and networking component shortages persist: Extended lead times for critical networking components continue to constrain optical transceiver availability, prompting some buyers to explore alternative sourcing strategies. Despite these supply challenges, regulatory conditions remain supportive of broader participation in the optical transceiver market.
-
NIC supply tightens as pricing increases: Demand is shifting from DPUs toward CX6 and CX7 NICs, increasing the risk of CX7 200G/400G shortages in early 2027. MCX6 and MCX7 lead times are running at 3–6 months, while NVIDIA is removing switch discounts, resulting in increases of around 10% for many affected customers. Broadcom NIC pricing rises November 1, with increases of 7.5% on lower-port-count products and as much as 22% on higher-port-count models.
- Broadcom Trident 4 allocation remains constrained: Trident 4 availability is severely restricted, with allocation unavailable to some networking customers through the remainder of 2026 and Q1 2027.
-
MLCC shortages broaden across major suppliers: More than 300 open shortages are affecting a broad range of component categories, with MLCC gaps concentrated around Murata, Yageo, Taiyo Yuden, TDK and Samsung. Murata and Yageo remain among the most constrained suppliers through December, pushing server manufacturers into the spot market for MLCCs. Samsung Electro-Mechanics Q4 increases are running 25–30% on consumer X5R and 10–20% on AI-server X6S, while nearly all Taiyo Yuden products are on allocation.
-
MLCC utilization reaches critical levels: Yageo is running at 98–100% utilization with a 1.7 book-to-bill ratio and is not expanding capacity, while Murata is reportedly prioritizing AI-related MLCC demand. The 47µF X6S 0603 and 0805 segment is currently the tightest class, with conditions expected to worsen in 2027.
- PCB and glass-cloth constraints intensify: PCB availability is emerging as a major gating factor for hyperscale infrastructure, driven by E-glass shortages and strong demand, with lead times reportedly doubling over the past year. Glass cloth used in FR-4 laminates remains particularly exposed due to constraints at a shared upstream supplier.
Hottest MPNs
|
Part Number |
Manufacturer |
Product Type |
|
NVRTXPRO6000TCGPU-KIT |
PNY Technologies |
Professional GPU kit |
|
900-2G153-0000-000 |
NVIDIA |
Professional GPU card |
|
M321RAJA0MB2-CCP |
Samsung |
Server memory module |
|
MT25QU512ABB8ESF-0SIT |
Micron |
NOR flash |
|
900-5G153-2200-000 |
NVIDIA |
Professional GPU card |
|
MT41K256M16TW-107:P |
Micron |
DDR3 SDRAM IC |
|
VCNRTXPRO6000MQ-BLK |
PNY Technologies |
Professional GPU card |
Priority Manufacturer Watchlist
|
Manufacturer |
October Development |
Supply / Sourcing Implication |
Action |
|
NVIDIA |
Q2 FY2027 data-centre revenue was $89.0bn; Q3 revenue guide is $108.0bn ±2% (as reported Aug 26). |
Accelerator demand continues to pull through HBM, networking, power and enterprise storage, not GPUs in isolation. |
Secure a complete rack and fabric BOM against deployment dates. |
|
TSMC |
Q3 2026 revenue was US$46.72bn, +41.2% YoY; advanced technologies were 79% of wafer revenue. Q4 guide is US$52.3–54.3bn. |
Leading-edge process and advanced-package exposure remains customer- and programme-priority dependent. |
Confirm node, package, substrate and test path during forecast and design reviews. |
|
Micron |
FQ4 revenue was $54.23bn. Micron has contracted the vast majority of its 2027 HBM bit supply and sees DRAM/NAND tightness into 2027–28. |
Memory access is increasingly secured by strategic agreements rather than transactional sourcing. |
Refresh 2027 HBM, RDIMM and enterprise-SSD demand with approved configurations. |
|
Broadcom |
At OCP, Broadcom featured 102.4T switching, 800G NICs, PCIe components and co-packaged optics for AI racks. |
Network architecture transitions increase switch, NIC, optics and CPO ecosystem dependencies. |
Release architecture-specific fabric requirements early and qualify viable alternatives. |
|
Marvell |
A new 3.2T retimer and second-generation 1.6T AEC DSP are sampling for 224G/lane AI scale-up fabrics. |
Retimers and active electrical cables become earlier design and qualification gates as lane rates rise. |
Bring AEC/DSP/retimer requirements into design-release and sourcing plans. |
|
Infineon |
Infineon completed the C2i acquisition and introduced a 27kW three-phase PSU solution for 800V/±400V AI-server architectures. |
AI power delivery is moving toward higher-density, digitally controlled designs with narrower approved component sets. |
Identify power-stage, controller, MOSFET/SiC and magnetic dependencies before design freeze. |
|
Nexperia |
Q2 revenue improved 18.4% QoQ as supply-chain stabilisation progressed; China entities remain outside the group’s consolidated disclosure (as reported Aug 28). |
Exposure remains legal-entity, COO and customer-qualification specific, not a uniform stockout. |
Verify COO, date code, entity and acceptance before committing substitutes. |
Lead Times & Pricing
→ Stable / selective movement ↑ Rising ↑↑ Strongly rising ↑↑↑ Critical
|
Product (Sub-Category) |
Brand / Ecosystem |
Lead-Time Condition* |
Allocation |
Pricing Trend |
|
AI Compute & Networking |
||||
|
AI Accelerator Platforms |
NVIDIA / AMD |
Programme allocation; quote-specific |
Allocation Only |
↑↑↑ |
|
Advanced Packaging / Custom AI ASICs |
TSMC ecosystem |
Capacity-constrained; customer-specific |
Capacity Constrained |
↑↑ |
|
800G / 1.6T Optics, AECs & DSPs |
Optics / cable ecosystem |
Extended; design and qualification-specific |
Constrained |
↑↑ |
|
High-Radix Switch / DPU / NIC Silicon |
Broadcom / NVIDIA / Marvell |
Extended; architecture-specific |
Constrained |
↑↑ |
|
Memory & Storage |
||||
|
HBM |
Samsung / SK hynix / Micron |
LTA / allocation-driven |
Allocation Only |
↑↑↑ |
|
DDR5 RDIMM, High Density |
Samsung / SK hynix / Micron |
Extended; allocation-specific |
Allocation Only |
↑↑ |
|
Enterprise PCIe 5.0 / 6.0 SSD |
Samsung / Solidigm / Micron / Kioxia |
Extended; capacity mix dependent |
Allocation Only |
↑↑↑ |
|
Mainstream DDR5 / NAND Contract |
Major suppliers |
Variable; contract supply tighter than channel indications |
Selective / Tight |
↑↑ |
|
Embedded, Power & Industrial |
||||
|
NOR Flash 256Mb / Octal ≥ SPI |
Winbond / GigaDevice / MXIC |
Extended in high-density grades |
Constrained |
↑↑ |
|
AI Server Power / Multiphase Controls |
Infineon / TI / onsemi |
Design and qualification-specific |
Elevated |
↑ |
|
High-Reliability MLCC |
Murata / TDK / SEMCO / KEMET |
Application and rating-specific |
Elevated |
↑↑ |
|
Small-Signal Discretes |
Nexperia / onsemi / Vishay |
COO, manufacturer and qualification-specific |
Constrained / Segmented |
↑ |
|
Mature-Node Analog & Power ICs † |
Texas Instruments |
Quoted 12–16 weeks versus about six months in practice; many customers hearing 52+ weeks |
Constrained; decommits widespread |
↑↑ |
|
Precision Analog & Linear Power † |
Analog Devices |
60–70+ weeks typical; some deliveries pushed to mid-2027 or 2028 |
Constrained; expedite support minimal |
↑ |
|
Power MOSFETs & Discrete Diodes † |
Diodes Inc / onsemi / Vishay / Infineon |
48–104+ weeks by brand and product; AI-related MOSFETs longest |
Partial Allocation |
↑↑ |
*Lead-time conditions are October 2026 are Fusion’s market observations. Public disclosures do not provide universal lead times. Confirm final lead time and allocation against exact MPN, quantity, product revision, COO and end-customer qualification.
Commodity Risk Brief
|
Commodity / Supply Chain |
Condition |
Price Signal |
What Has Changed |
Procurement Response |
|
HBM & high-density server DRAM |
Critical / Allocation Only |
↑↑↑ |
Forward HBM is largely contracted; 4Q conventional DRAM contract prices are forecast +10–15% QoQ. |
Protect approved configurations and vendor alternatives before platform release. |
|
Enterprise SSD / high-density NAND |
Critical / Allocation Only |
↑↑↑ |
Enterprise SSD demand is forecast +80% YoY in 2026; most incremental supply is committed and 4Q price growth accelerates. |
Commit high-capacity configurations and controller/NAND alternatives early. |
|
High-density NOR flash |
Elevated / Supply Compression |
↑↑ |
256Mb+ grades remain constrained by limited effective capacity and qualification time. |
Qualify density, package and supplier options for firmware-heavy designs. |
|
AI fabric: switching, optics, AEC & DSP |
Critical / Capacity Constrained |
↑↑ |
102.4T switching and 224G/lane designs add retimer, cable, DSP and optical dependencies. |
Source the complete fabric BOM at initial release, by deployment milestone. |
|
MLCC & high-density AI power chain |
Elevated / Targeted |
↑↑ |
4Q MLCC actions and 800V/±400V server designs add architecture-specific exposure. |
Map approved MLCC, controller, power-stage, MOSFET/SiC and magnetics. |
|
Mature-node analog ICs (TI, ADI) † |
Critical / Constrained |
↑↑ |
TI’s 12–16 week quotes are running closer to six months in practice, with decommits widely reported and another 10–15% increase expected in October. ADI lead times of 60–70+ weeks are typical, and customers report ADI is no longer honoring price-lock and rebate terms. |
Order at least six months ahead where programmes allow, test delivery commitments against actual rather than quoted lead times, and review price-protection assumptions in contracts. |
|
Power MOSFETs & discrete diodes † |
Critical / Partial Allocation |
↑↑ |
Diodes Inc lines mostly run 48–52+ weeks, with mid-size customers reporting about 10% of requested Q4 allocation. onsemi is at 50+ weeks versus 26+ in early 2026, Vishay AI-related MOSFETs at 104+ weeks, and Infineon prices are up 10–50% in the past quarter. |
Qualify alternates by package and customer approval, recognising that brand switching adds load to Diodes Inc and onsemi, and weigh multi-month PO or LTA requests against demand certainty. |
October marks a sharper separation between allocated infrastructure-grade memory and broader channel signals. Enterprise SSD is the only memory category where 4Q price growth is expected to accelerate, so allocation access and configuration approval matter more than headline spot movement.
Regional Sourcing Lens
|
Region |
Supply Signal |
Implication for Procurement |
|
US / North America |
CSP deployment schedules continue to govern HBM, server-memory, enterprise-SSD and fabric allocation. |
Submit platform and deployment detail; generic spot enquiries carry less weight. |
|
Europe |
Nexperia exposure remains COO-, legal-entity- and qualification-specific; AI power adds approved-BOM dependencies. |
Pre-clear COO, traceability and alternatives with quality and customer teams. |
|
China |
High-density memory and AI infrastructure are allocation-led; origin and compliance determine usable supply. |
Validate origin, export-control, end-use and customer requirements before award. |
|
SEA |
EMS builds can be held up by memory or fabric allocations despite normal finished-goods supply. |
Track component commits against build milestones and protect critical configuration buffers. |
30-Day Procurement Actions
|
Priority |
Action Required in the Next 30 Days |
Why It Matters |
Owner |
|
1 |
Align 4Q26 and 2027 AI demand into a single allocation strategy covering HBM, RDIMM, SSDs, networking and accelerators. |
AI infrastructure is increasingly allocated at the platform level; missing one component can delay complete system deployment. |
Procurement + Engineering |
|
2 |
Convert validated DRAM, HBM and enterprise SSD forecasts into supplier commitments. |
Pricing is rising and available capacity is increasingly locked into long-term agreements. |
Commodity Management |
|
3 |
Finalize and release 224G/1.6T networking BOMs, including optics, DSPs, NICs, switches and cables. |
Advanced networking components are emerging as critical-path items with extended lead times. |
Network Engineering + Sourcing |
|
4 |
Map power-critical components across current and next-generation AI platforms. |
Higher-density power architectures are creating new supply and qualification constraints. |
Engineering + Supply Chain |
|
5 |
Maintain COO, legal-entity, traceability and customer-acceptance controls for Nexperia discretes, high-density NOR and alternatives. |
Qualification and compliance requirements may limit substitution options despite technical compatibility. |
Quality + Supply Chain |
|
6 |
Re-baseline analog and discrete lead times against actual delivery history, and place orders at least six months ahead for critical TI, ADI and power-discrete lines. |
Quoted TI lead times of 12–16 weeks are running closer to six months in practice, ADI asks for orders six months ahead, and discrete lead times run from 48 to 104+ weeks. |
Procurement + Supply Chain |
Based on Fusion market observations for October 2026 rather than public disclosures.