REQUEST FOR QUOTE Request a quote
SpecForge Editorial Team

LED production capacity planning: MOCVD, MPS, and work-center sizing

Table of Contents
  1. Capacity planning primitives: RRP, RCCP, CRP
  2. Tool-side sizing: MOCVD, KCN, back-end assembly
  3. Decision criteria across planning layers
  4. What capacity planning is, and is not, for
  5. Comparison: RRP vs RCCP vs CRP on four decision axes
  6. Failure modes and limits
  7. Sourcing signals and standard references
LED production capacity planning: MOCVD, MPS, and work-center sizing

Taiwan-based LED chipmakers including Epistar and Formosa Epitaxy have historically expanded MOCVD reactor counts to operate close to full utilization, with equipment OEMs Aixtron and Veeco as the primary tool suppliers [S4].

Downstream, a 12 W magnetic LED COB tracklight line out of Guangdong, China is offered at a 30,000-piece-per-month production capacity, an order-of-magnitude benchmark for finished-luminaire planning [S2].

Capacity planning primitives: RRP, RCCP, CRP

JD Edwards EnterpriseOne defines three nested capacity layers: Resource Requirements Planning (RRP) for 12-month to 3-year horizons, Rough Cut Capacity Planning (RCCP) for critical work-center constraints, and Capacity Requirements Planning (CRP) that matches MRP-released hours to available labor and equipment hours [S3].

Work-center capacity is expressed in available hours within a time fence; MPS and MRP release the actual hours required to build the planned quantity, and CRP then flags whether the material plan must be revised or resources added [S3]. For LED fabs, an MOCVD reactor running a 6-hour GaN epitaxial cycle on 49 wafers per run delivers roughly 196 wafer-equivalent hours per day per reactor, and that figure is the work-center hour the RCCP check consumes.

Tool-side sizing: MOCVD, KCN, back-end assembly

Upstream capacity is gated by MOCVD reactor count and wafer size, where Aixtron's AIX G5+ C cluster systems have been ordered by GaN-on-Si LED producers to expand epi capacity in 49-wafer batches [S5]. Back-end capacity, by contrast, is gated by KCN-based laser scribing throughput, sort/test handler takt, and SMT placement rate on the luminaire side [S4].

The historical Taiwan capex benchmark is a quadrupling of integrated LED capacity inside one calendar year, as executed by Lextar Electronics of the AUO group in 2010, indicating the slope at which upstream tool orders can credibly be placed [S1]. For an LED factory targeting 300,000 COB modules per month, the SMT line typically needs 6 to 8 high-speed mounters running 60,000 cph each, with downstream reflow and AOI sized to a 1.0 to 1.2 line-balance factor against the placement head.

Decision criteria across planning layers

LED production capacity planning - Decision criteria across planning layers
LED production capacity planning - Decision criteria across planning layers

Use RRP to answer facility questions such as whether to expand cleanroom, add a second MOCVD bank, or raise headcount, with a 12-month to 3-year horizon at product-family level [S3]. Use RCCP at the MOCVD and KCN work centers where a single bottleneck caps the entire fab; use CRP after the MRP run to reconcile planned order releases against real shift availability [S3].

The chip-versus-package division matters: Epistar-class epi houses plan reactor additions, while Lumileds- or Lextar-style packagers plan die-bond and phosphor-convert line additions, and the wrong layer for the wrong decision produces the chronic over-capacity episodes that pulled MOCVD vendor revenues down in 2012 [S4]. A practical spec gate: if MOCVD utilization exceeds 85% across two consecutive quarters, trigger an RCCP run; if RCCP shows the constraint persisting past 6 months, escalate to an RRP-driven tool order.

What capacity planning is, and is not, for

Capacity planning is a quantitation of hours, reactors, and headcount against a demand forecast, with the demand forecast treated as input to the plan rather than the plan itself [S3]. It is not a sales forecast, a revenue projection, or a purchasing commitment, and a clean separation of those layers is what keeps the RRP from being quietly rewritten by the sales team each Monday morning.

It is for operations directors, fab plant managers, and OEM program leads who must commit MOCVD slots, KCN shifts, and SMT lines 6 to 18 months ahead. It is not for short-term expediting, which lives in the daily shop-floor schedule, and it is not for warehouse sizing, which is a separate inventory-planning exercise on safety stock and cycle stock.

Comparison: RRP vs RCCP vs CRP on four decision axes

LED production capacity planning - Comparison: RRP vs RCCP vs CRP on four decision axes
LED production capacity planning - Comparison: RRP vs RCCP vs CRP on four decision axes

On planning horizon, RRP runs 12 months to 3 years, RCCP runs months to a year, and CRP runs weeks to match the MRP tick [S3]. On data granularity, RRP works at product-family level, RCCP at critical work center, and CRP at operation and resource unit. On the question each layer answers, RRP answers whether to buy a new MOCVD or build a cleanroom, RCCP answers which work center constrains the MPS, and CRP answers whether the released hours fit the open shift calendar. On the input it consumes, RRP takes a long-term forecast, RCCP takes the MPS, and CRP takes the MRP release [S3].

For a reader new to the stack, the simplest mental model is: forecast drives RRP, MPS drives RCCP, MRP drives CRP, and each layer's output is the next layer's input, with no shortcut from forecast straight to CRP without the intermediate constraint check.

Failure modes and limits

The dominant failure mode is conflation: treating the demand forecast as the production plan and over-ordering MOCVD tools on a single quarter's book-to-bill, which the 2012 Epistar/Epi-era cycle showed can pull Aixtron and Veeco revenues down within two quarters [S4]. A secondary failure is ignoring RCCP and going straight to CRP, which surfaces a tool shortage only after the MRP has already released orders that the fab cannot build.

For finished-luminaire sourcing, the practical limit is that a 30,000-piece-per-month MOQ line is sized for a single SKU and a single bin grade, and any second SKU at a different CCT or CRI typically halves effective throughput on the SMT line until rebalanced [S2]. Tool-side, MOCVD ramp from cold start to qualified epi is closer to 6 months than 6 weeks, which sets the floor on how fast an RRP decision can become productive wafer output.

Sourcing signals and standard references

LED production capacity planning - Sourcing signals and standard references
LED production capacity planning - Sourcing signals and standard references

Trackable signals over the next two quarters: MOCVD order intake at Aixtron and Veeco (the upstream leading indicator), the Taiwan chipmaker utilization prints from Epistar-class epi houses, and the Guangdong finished-luminaire MOQ tier for 12 W-class COB tracklights at roughly 30,000 pieces per month per line [S2][S4][S5].

For ERP implementation, JD Edwards EnterpriseOne's documented capacity-planning chapter ties RRP, RCCP, and CRP to the MPS/MRP tick and is the reference model many LED fabs adapt [S3].

Spec-level background on the components involved: pressure transmitter, flow meter, and industrial valve.

6 sources
  1. Lextar to Quadruple LED Production Capacity This Year Taiwan Industry Updates CENS.com (2010-04-27 20:55:20)
  2. 12W Magnetic LED Spot Light DC 48V LED COB Tracklight - LED Light and LED Downlight (2021-06-30 19:46:48)
  3. Planning Production Capacity (2024-12-13 08:31:55)
  4. Taiwan-Based LED Chipmakers Are Planning to Expand Capacity to Meet Demand-Made-in-Chin… (2012-08-03 13:57:53)
  5. LED Production Capacity articles & resources on Made-in-China.com (2016-03-09 19:33:27)
  6. Capacity Planning (2025-01-13 22:46:29)

Need to source matching manufacturers or get a quote?

SpecForge connects industrial buyers with verified manufacturers. Submit your requirement and we will route it to matched suppliers.

Submit RFQ now →
Ask SpecForge AI