TrendForce rates six of the seven components it tracks for AI servers as short of supply, and on 2 November it moved the seventh, GPUs, from “Balanced” to “Tightening”, according to its Weekly Radar 008. ABF substrates, used to package AI chips, have the longest wait: 48 to 56 weeks against 12 weeks in a balanced market, up to about 4.3 times normal by our arithmetic at the midpoint.
The research firm’s summary of the update: “GPU shifts from Balanced to Tightening, MLCC lead times extend to 38 weeks, and DRAM eases to 19 weeks but remains Very Tight, especially for 32GB/64GB RDIMMs.”
Business Pill · GPUs Explained
An evergreen Business Engineer explainer on GPUs, the chips at the centre of AI servers. It teaches the general idea only and does not discuss this research.
The key insight: TrendForce’s 2 November radar rates six of seven AI server components as short (three Very Tight, three Tight) and moves the seventh, GPUs, to Tightening. ABF substrate lead times of 48 to 56 weeks are up to about 4.3 times the 12 weeks of a balanced market (our arithmetic).
Source: TrendForce table dated 2 November 2026 (Weekly Radar 008). *Week-on-week changes as summarised by Meritz Securities from the same TrendForce update.
GPU. TrendForce writes that from the fourth quarter of 2026 Nvidia and other AI chip suppliers are expected to shift to new platforms, that “availability of new platforms such as Rubin could be constrained due to upstream component shortages”, and that supply is likely to go first to hyperscale cloud providers.
CPU. “AI inference continues to drive server CPU demand, while x86 server CPU supply remains constrained”, with vendors expected to allocate supply, which TrendForce says opens opportunities for Arm-based CPUs.
HDD. New recording technologies will not add capacity “until late 2027 or 2028”, so lead times stay at 50 weeks.
ABF. Capacity is “being allocated primarily to major customers, leaving other buyers facing lead times of nearly a year.”
MLCC. Equipment makers are “rushing to secure 2027 supply”. In a separate release on 29 October, TrendForce said it expects the high-end MLCC shortage to extend through the first half of 2027.
NAND. TrendForce sees possible relief: lead times “could potentially improve by 1Q27” because of softer demand in China, but for now they stay at 16 weeks.
Lead time in weeks for seven AI server components, in a balanced market and now, with ranges plotted at their midpoint. Data: TrendForce, Weekly Radar 008 (2 November 2026). Midpoints are our arithmetic.
Status count: three components are Very Tight (DRAM, HDD, ABF), three are Tight (CPU, NAND, MLCC) and one is Tightening (GPU). In TrendForce’s key, Very Tight means severe shortage, Tight means constrained supply and Tightening means conditions worsening; none of the seven is Balanced or Easing (our count from TrendForce’s table).
Current lead time against balanced, at the midpoint of each range: ABF about 4.3 times, MLCC about 3.2 times, HDD about 3.1 times, DRAM about 2.4 times, NAND 2.0 times, CPU about 1.5 times and GPU about 1.1 times (our arithmetic).
The GPU line is the closest to normal on lead time; what changed this week is its direction.
As we read the table, the longest waits are not for the chips themselves: ABF substrates, HDDs and MLCCs have the highest ratios to a balanced market, while GPUs are the closest to it.
As we read TrendForce’s allocation notes for GPUs and ABF, the largest buyers are served first, which leaves smaller buyers with the longest waits.
TrendForce, table dated 2 November 2026
“Available capacity is being allocated primarily to major customers, leaving other buyers facing lead times of nearly a year.”
Three Implications
FOR AI SERVER BUYERS ABF substrates, HDDs and MLCCs carry lead times of 38 to 56 weeks, per TrendForce.
FOR CHIPMAKERS TrendForce says new GPU platforms such as Rubin could be constrained by upstream component shortages.
FOR MEMORY WATCHERS TrendForce lists DRAM at 19 weeks, down one week, and says NAND lead times could improve by 1Q27.
The framework says: “The AI race has a new bottleneck, and it’s not what most observers expect.”
As we read it, TrendForce’s table extends that bottleneck from memory to packaging and storage. The framework does not discuss this research.
What Is Not Established
We read TrendForce’s table and the free summary of its radar; the full radar is behind a paywall. The week-on-week changes come from Meritz Securities’ summary of the same update, not from TrendForce directly. Lead times describe how long buyers wait for new orders; they are not prices, and TrendForce does not forecast company earnings in the table.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
TrendForce rates six of the seven AI server components it tracks as short of supply and marks the seventh, GPUs, as tightening; ABF substrates wait up to about 4.3 times their normal lead time.
A note on sourcing. Component statuses, lead times and quoted notes are from TrendForce’s table dated 2 November 2026 and the free summary of its Weekly Radar 008; the MLCC outlook is from TrendForce’s 29 October release. Week-on-week changes are from Meritz Securities’ summary of the same TrendForce update (our translation). The piece was prompted by a post on X from Milk Road AI that quoted a summary of the table. Ratios and counts are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
