transceiver-db/RESEARCH-revenue-lifecycle-prediction.md

Revenue Lifecycle Prediction Models for Optical Networking Equipment

Research Date: 2026-03-28 Scope: Optical transceivers, switches, routers — product lifecycle revenue prediction

---

Table of Contents

1. Revenue Lifecycle Prediction Models
2. Historical Data Points for Optical Transceivers
3. Regional/Country-Level Adoption Differences
4. Conference-to-Market Timeline Analysis
5. Switch/Router Refresh Cycles
6. Predictive Models for Future Products
7. Recommended Implementation for TIP

---

1. Revenue Lifecycle Prediction Models

1.1 Bass Diffusion Model (Foundation)

The Bass model (1969) is the foundational framework for technology adoption forecasting.

Core Equation:

f(t) = (p + q * F(t)) * (1 - F(t))

Where:

• f(t) = instantaneous rate of adoption at time t (fraction of market potential)
• F(t) = cumulative fraction of adopters at time t
• p = coefficient of innovation (external influence / "advertising effect")
• q = coefficient of imitation (internal influence / "word-of-mouth effect")

Closed-form cumulative adoption:

F(t) = (1 - exp(-(p+q)*t)) / (1 + (q/p)*exp(-(p+q)*t))

Revenue form (units * price):

R(t) = m * f(t) * P(t)

Where m = total market potential, P(t) = price at time t.

Typical parameter ranges (telecom/technology):

• p: 0.01 - 0.03 (innovation coefficient)
• q: 0.2 - 0.4 (imitation coefficient)
• Peak adoption occurs at: t_peak = (1/(p+q)) * ln(q/p)

Source: Bass, F.M. (1969). "A New Product Growth for Model Consumer Durables." Management Science, 15(5), 215-227.

• Bass diffusion model - Wikipedia
• GeeksforGeeks explanation

1.2 Norton-Bass Multi-Generation Diffusion Model (CRITICAL for TIP)

The Norton-Bass (NB) model (1987) extends Bass to handle successive technology generations — exactly the pattern seen in optical transceivers (1G → 10G → 40G → 100G → 400G → 800G → 1.6T).

Two-Generation Formulation:

Generation 1 introduced at t=0, Generation 2 at t=τ₂.

Units-in-use for G1:
  N₁(t) = m₁ * F₁(t)                                for t < τ₂
  N₁(t) = m₁ * F₁(t) * (1 - F₂(t - τ₂))            for t ≥ τ₂

Units-in-use for G2:
  N₂(t) = 0                                           for t < τ₂
  N₂(t) = (m₂ + m₁ * F₁(t)) * F₂(t - τ₂)           for t ≥ τ₂

Where:

• Fᵢ(t) = Bass cumulative adoption for generation i
• mᵢ = incremental market potential for generation i
• τ₂ = introduction time of generation 2

Key finding: p and q parameters are generally the same between successive generations — only market potential (m) changes.

Three-Generation Extension:

N₁(t) = m₁*F₁(t)*(1-F₂(t-τ₂))                                    for τ₂ ≤ t < τ₃
N₁(t) = m₁*F₁(t)*(1-F₂(t-τ₂))*(1-F₃(t-τ₃))                      for t ≥ τ₃

N₂(t) = (m₂+m₁*F₁(t))*F₂(t-τ₂)*(1-F₃(t-τ₃))                     for t ≥ τ₃

N₃(t) = (m₃ + (m₂+m₁*F₁(t))*F₂(t-τ₂) + m₁*F₁(t)*(1-F₂(t-τ₂)))*F₃(t-τ₃)

Source: Norton, J.A. & Bass, F.M. (1987). "A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products." Management Science, 33(9), 1069-1086.

• INSEAD working paper
• Semantic Scholar

1.3 Generalized Norton-Bass (GNB) Model

Jiang & Jain (2012) extended Norton-Bass to differentiate leapfrogging from switching — critical for optical transceivers where some data centers skip generations (e.g., skip 40G, go from 10G to 100G).

Leapfrogging: Potential adopters skip older generation and directly adopt newer generation. Switching: Existing adopters of older generation migrate to newer generation.

Two-Generation GNB Formulation:

Leapfrog adoptions of G2:
  L₂(t) = m₂ * F₂(t - τ₂)

Switching adoptions from G1 to G2:
  S₂(t) = m₁ * F₁(t) * F₂(t - τ₂)

Total G2 units-in-use:
  N₂(t) = L₂(t) + S₂(t) = (m₂ + m₁*F₁(t)) * F₂(t - τ₂)

G1 remaining units:
  N₁(t) = m₁ * F₁(t) * (1 - F₂(t - τ₂))

Empirical validation (DRAM generations):

• 4K, 16K, 64K DRAM quarterly shipments 1974-1984
• Adjusted R² values: 0.9853, 0.9707, 0.999
• Of 64K DRAM adoptions: 60% new adopters, 33% switching from 16K, rest leapfrogging

Software: Available in R via the diffusion package (Nortonbass function).

Source: Jiang, Z. & Jain, D.C. (2012). "A Generalized Norton-Bass Model for Multigeneration Diffusion." Management Science, 58(10), 1887-1897.

• Full PDF - Iowa State
• SSRN
• INFORMS
• R package docs

1.4 Gompertz Curve for Revenue Lifecycle

The Gompertz curve is particularly effective for modeling the asymmetric S-curve of technology market growth, where early adoption accelerates fast but saturation is gradual.

Formula:

y(t) = K * exp(log(y₀/K) * exp(-α*t))

Where:

• K = carrying capacity (maximum market size / saturation level)
• y₀ = initial value
• α = growth rate coefficient
• Inflection point occurs at 36.8% of upper asymptote (vs. 50% for logistic)

Alternative parametrization:

y(t) = a * b^(c^t)

Where a = upper asymptote, 0 < b < 1, 0 < c < 1.

Application to semiconductors: Wally Rhines (Mentor Graphics) demonstrated that the Gompertz curve can determine where particular semiconductor market segments are in their lifecycle by plotting cumulative unit production against the Gompertz S-curve. By determining the three coefficients early in the cycle, the remainder of the cycle can be predicted.

Gompertz vs. Logistic: When Y is low, Gompertz grows faster; when Y is high, Gompertz grows slower. This asymmetry better matches technology markets where early adoption is driven by innovators (fast) but late-stage saturation is drawn out by laggards.

Source:

• EE Times - Gompertz for semiconductor prediction
• Semiengineering - Gompertz model
• FasterCapital - Business growth

1.5 Weibull Distribution for Lifecycle Curves

The Weibull distribution provides a flexible framework for modeling both growth and decline phases with varying shapes.

Lifecycle formulation:

f(t) = (β/η) * (t/η)^(β-1) * exp(-(t/η)^β)

Where:

• β = shape parameter (β < 1: decreasing failure/decline rate, β > 1: increasing)
• η = scale parameter (characteristic life)

A 2019 paper proposes a two-step Weibull distribution with four parameters for modeling bimodal product lifecycle diffusion curves — fitting both the rise and fall of product sales.

Source: "Using Weibull Distribution for Modeling Bimodal Diffusion Curves: A Naive Framework to Study Product Life Cycle." International Journal of Innovation and Technology Management, 2019.

• World Scientific
• Technological Forecasting & Social Change - Weibull for tech change

1.6 Revenue Duration Model (Composite)

For TIP, the recommended composite model for a single transceiver generation:

Revenue(t) = Units(t) * ASP(t)

Where:
  Units(t) = Norton-Bass adoption model (accounts for cannibalization by next gen)
  ASP(t) = ASP₀ * exp(-λ*t)  (exponential price erosion)

Duration above 50% peak revenue:
  Solve for t₁, t₂ where R(t) = 0.5 * R_peak
  Duration = t₂ - t₁

---

2. Historical Data Points for Optical Transceivers

2.1 Total Optical Transceiver Market Revenue by Year

Year	Total Market Revenue	Growth	Source
2019	~$7.5-8.0B	Declined	LightCounting (derived)
2020	~$8.8-9.3B	+17%	LightCounting
2021	~$10.0B+	+10%	LightCounting milestone
2022	~$11.0-11.5B	+14%	LightCounting
2023	~$10.7-10.9B	-6%	LightCounting; telecom downturn
2024	~$13.6B	Strong rebound	MarketsandMarkets; AI-driven
2025	~$23B (projected)	+60%+	LightCounting Dec 2025

Datacom optical segment specifically:

• 2024: ~$9B (Cignal AI)
• 2025: >$16B (Cignal AI, +60%)
• 2026: ~$12B high-speed datacom segment projected (Cignal AI, as 800G peaks)

Sources:

• LightCounting newsletter
• Cignal AI Jan 2025
• Cignal AI May 2025
• MarketsandMarkets

2.2 Generation Lifecycle Timelines

Generation	Datacom Launch	Peak Revenue Window	Years to Peak	Cycle → Next Gen
1G SFP	~2002	~2008-2012	~6-8 yrs	~5 yrs
10G SFP+	~2007-2010	~2013-2016	~4-6 yrs	~4 yrs
40G QSFP+	~2011-2013	~2015-2017	~3-4 yrs	~3 yrs (largely skipped)
100G QSFP28	~2014	~2018-2020	~4 yrs	~3-4 yrs
400G QSFP-DD	~2018-2019	~2022-2024	~3-4 yrs	~3 yrs
800G OSFP	~2023-2024	~2025-2026 (proj)	~2-3 yrs	~2 yrs
1.6T OSFP-XD	~2025-2026	~2027-2028 (proj)	~2 yrs	~2 yrs

KEY FINDING: Innovation cycles are compressing from 3-4 years historically to ~2 years currently.

Sources:

• Introl blog
• Cignal AI
• Medium - Module Evolution

2.3 Price Erosion Curves

100G QSFP28 SR4 Price History

Period	Approx. ASP	Notes
2015-2016	>$2,000	Early production, few suppliers
2017	~$800-$1,200	Volume ramp begins
2018	~$400-$700	Chinese suppliers enter
2019	~$200-$400	Commoditization
2020	~$100-$250	COVID demand + continued pressure
2021-2022	~$80-$150	Mature market
2024-2026	~$29-$99	Third-party vendors (FS.com, Optcore)

Overall decline: ~60% in 5 years, ~95%+ from launch to commodity phase.

Price erosion model:

ASP(t) = ASP₀ * exp(-λ*t)

For 100G QSFP28:
  ASP₀ ≈ $2,000 (launch year 2015)
  λ ≈ 0.35-0.40 per year (aggressive phase)
  Half-life: ~2 years

800G Module Pricing (2024)

Module Type	ASP (2024)
800G Multimode (SR8, VCSEL)	~$500
800G LPO	~$600
800G Single-mode (EML)	>$700
NVIDIA LinkX 800G (bulk)	~$1,000
800G FR4/DR8 (reseller)	$1,000-$3,800

1.6T Module Pricing

Period	ASP
Q4 2024 (initial)	~$2,000
2025 (maturity)	~$1,500 (projected)

Sources:

• Deep Fundamental - Optical Module Market
• Approved Networks
• FS.com

2.4 Shipment Volumes

Year	400G+800G Units	800G Alone	1.6T
2022	~5M (est.)	Early	—
2023	~8M (est.)	Ramp	—
2024	>20M	~10M	~300K (Q4)
2025	—	12-15M (proj)	2-6M (proj)

GPU-to-module ratio: 1 H100 = 2.5x 800G modules (training); 1 B200 = 2.5x 1.6T modules.

Sources:

• Cignal AI
• Deep Fundamental

2.5 400G ZR Coherent Timeline (Case Study)

Milestone	Date	Volume
OIF 400ZR spec finalized	~2019-2020	—
First commercial shipments	Late 2021	>60,000 units
OFC 2022 demos / volume ramp	2022	~190,000 units
Mass deployment (hyperscale + telco)	2023-2024	Bulk of WDM bandwidth
800G ZR GA announced	March 2025	Next gen arriving

Timeline: Spec → first shipment: ~18-24 months. First shipment → volume: ~12 months. Total spec → volume: ~30-36 months.

Sources:

• FiberMall
• Coherent 800G ZR announcement
• PrecisionOT

---

3. Regional/Country-Level Adoption Differences

3.1 Adoption Tier Framework

Based on research findings, optical transceiver adoption follows a tiered geographic pattern:

Tier	Region	Adoption Lag	Primary Drivers
Tier 1	US Hyperscalers (Google, Meta, Amazon, MS)	Reference (0 months)	AI training, scale-out DC
Tier 1B	Chinese Hyperscalers (Alibaba, Tencent, ByteDance)	6-12 months	Domestic manufacturing, export controls
Tier 2	Japan/Korea (NTT, SK Telecom)	12-18 months	Early coherent, methodical qualification
Tier 3	European Telcos (DT, Orange, Telefonica)	24-36 months	Regulatory, longer procurement cycles
Tier 4	India/SEA/LATAM	36-60 months	Infrastructure buildout, cost sensitivity

3.2 US Hyperscalers (Tier 1)

• Lead adopters for every generation — first to deploy at scale.
• Google's hyperscale DCs have deployed optical circuit switching at massive scale.
• NVIDIA/Meta/Google driving LPO adoption: >40% of short-reach 800G links by late 2025.
• NVIDIA's bulk 800G LinkX price: ~$1,000/transceiver at 100K+ volumes.
• 92% of 2025 hyperscale DC contracts specify OSFP-XD for 1.6T.

Source: Hector Weyl blog

3.3 Chinese Market (Tier 1B)

• Manufacturing dominance: Chinese manufacturers (Innolight, Eoptolink, Accelink) hold ~60% of merchant 800G market share.
• Innolight: ~40% global 800G share; >50% of NVIDIA procurement.
• Eoptolink: ~20% of NVIDIA's 800G LPO orders.
• Critical vulnerability: Chinese vendors remain dependent on US silicon — 5nm/3nm DSPs sourced almost exclusively from Broadcom and Marvell.
• Current export restrictions target compute chips, NOT networking signal processors — but this could change.
• Tencent was first deployer of Broadcom Humboldt CPO (2021).
• Accelink upgraded 1.6T OSFP224 at OFC 2025; Eoptolink launched Gen2 1.6T at OFC 2025.
• Asia-Pacific holds 30% of optical interconnect market share (fastest-growing region).

Source: Substack - Pluggables, Power, and Geopolitics

3.4 Europe (Tier 3)

• European presence focuses on equipment vendors (Ciena, Nokia) rather than hyperscale deployments.
• Ciena active in hyper-rail photonics, 1600ZR/ZR+ pluggables (acquired Nubis Communications).
• European telcos typically 2-3 years behind hyperscalers in adopting new transceiver generations.
• Regulatory and procurement cycle overhead extends adoption timelines.

3.5 Bass Model with Geographic Heterogeneity

Academic research confirms that Bass model parameters vary significantly across countries:

Key findings:

• Multi-country diffusion modeling helps overcome the "data hunger" problem — use earlier-adopting countries' data to predict later-adopting ones.
• BRIC mobile adoption study: India's q value was much higher than other BRIC countries.
• European broadband study: Bass model parameters for OECD countries showed peak adoption has already passed.
• 3G mobile across 35 countries: NLMIXED approach with pooled multi-country data.

Recommended approach for TIP:

For each region r:
  F_r(t) = Bass(p_r, q_r, m_r, t - lag_r)

Where lag_r = geographic adoption lag (months):
  US Hyperscaler:  lag = 0
  China Hyperscaler: lag = 6-12
  Japan/Korea:     lag = 12-18
  Europe Telco:    lag = 24-36
  India/SEA/LATAM: lag = 36-60

And p_r, q_r may be adjusted per region:
  Hyperscalers: higher p (innovation-driven), lower q
  Telcos: lower p, higher q (imitation-driven)
  Emerging: lower p, lower q, much higher m (larger potential)

Sources:

• ScienceDirect - Heterogeneity in diffusion
• ScienceDirect - Broadband diffusion Europe
• Academia.edu - Bass model BRIC
• Tandfonline - Agent-based Bass

---

4. Conference-to-Market Timeline Analysis

4.1 Standards Pipeline

The typical pipeline from concept to product:

OIF electrical interface → IEEE formal standard → MSA form factor spec → Product GA

Typical timing:
  OIF spec → IEEE ratification: 12-18 months
  MSA spec → first product samples: 6-12 months
  First samples → GA shipping: 6-12 months
  GA → volume production: 6-12 months

TOTAL: OIF spec → volume production: 30-48 months

4.2 Historical Conference-to-Market Timelines

400G ZR

Event	Date
OIF 400ZR spec finalized	~2020
First commercial shipments	Q4 2021
OFC 2022 demos / ramp	2022
Volume deployment	2022-2023
Spec → volume: ~24-30 months

800G

Event	Date
800G Pluggable MSA founded	Sept 2019
MSA PSM8 spec (first 800G pluggable)	2020
OSFP 800G spec released	June 2021
First shipments	2023
Volume production	2024
MSA founding → volume: ~5 years; Spec → volume: ~3-4 years

1.6T

Event	Date
OFC 2025 demos (multiple vendors)	April 2025
OFC 2026 demos (400G/lambda DR4)	March 2026
IEEE 802.3dj 200G/lane expected	Mid 2026
Sampling	Late 2025
Production ramp (projected)	Late 2026
Volume deployment	2027
Demo → volume: ~24 months

3.2T

Event	Date
Coherent demos at OFC 2026	March 2026
Expected arrival	~2026-2027 (samples)
LightCounting added 3.2T to forecast	July 2024

4.3 Conference-to-Market Formula for TIP

T_volume = T_demo + Pipeline_Lag

Where Pipeline_Lag depends on technology maturity:

  Incremental (same platform, higher speed):
    Pipeline_Lag = 18-24 months

  New platform (new form factor, new SerDes):
    Pipeline_Lag = 30-36 months

  Paradigm shift (CPO, new physics):
    Pipeline_Lag = 48-60 months

Key signals to monitor:

1. OIF electrical interface spec release → 30-48 months to volume
2. MSA spec release → 24-36 months to volume
3. IEEE standard ratification → 12-24 months to volume (spec often trails products)
4. Multiple vendors demoing at OFC/ECOC → 18-24 months to volume
5. LightCounting adding category to forecast → 24-30 months to volume

Sources:

• LPO MSA
• IEEE 802.3
• FS.com MSA intro
• Eoptolink OFC 2026
• EDN - OFC 2025 1.6T innovations
• Coherent 1.6T at OFC 2025

---

5. Switch/Router Refresh Cycles

5.1 Broadcom Tomahawk ASIC Timeline (Sets Industry Cadence)

Gen	Year	Bandwidth	Process	Key Optics
TH1	2014	3.2 Tb/s	28nm	10G/25G
TH2	2016	6.4 Tb/s	16nm	25G/50G
TH3	2017-18	12.8 Tb/s	16nm	50G/100G
TH4	2019-20	25.6 Tb/s	7nm	100G/400G
TH5	2022	51.2 Tb/s	5nm	400G/800G
TH6	2025	102.4 Tb/s	3nm	800G/1.6T
TH7	~2027	204.8 Tb/s	(planned)	1.6T/3.2T
TH8	~2029	409.6 Tb/s	(planned)	3.2T+

Cadence: Bandwidth doubles every ~2 years. A single TH5 replaces 48 TH1 switches (95% power reduction).

CRITICAL: Pluggable optics consume ~50% of system power and >50% of system cost.

Sources:

• Broadcom TH5
• Broadcom TH6 launch
• TechInsights - TH5
• NextPlatform - TH6 102.4T
• ServeTheHome - TH6
• NADDOD - TH6

5.2 Cisco Nexus Refresh Cycle

Platform	Generation	Release	Optics Support
Nexus 9364C	Cloud Scale	~2018-2019	100G/400G
Nexus 9364D-GX2A	Current gen	May 2022	400G
Nexus 9364C-H1	Updated	April 2024	400G
Nexus 9364E variants	Next gen	Feb 2025	800G
Nexus 9364C (EOL)	—	EOS Aug 2023	Support ends Jan 2029

Refresh cycle: ~2-3 years per platform generation.

Source: Cisco Nexus 9000 series

5.3 Arista Refresh Cycle

Platform	ASIC	Timeline
7800R3	Jericho 2	Prior gen
7800R4	Jericho 3-AI/3+	Shipping 2024-2025

The 7800R4 supports 1,152x 400G or 576x 800G ports. Existing 7800R3 systems can be upgraded with R4 fabric modules.

Source: Arista 7800R4

5.4 NVIDIA Networking

• Spectrum-X switches with ConnectX-7 NICs: current generation for AI clusters.
• ConnectX-8 / Spectrum-4 expected to follow standard ~2-year NVIDIA cadence.
• Quantum-X800: 144 ports of 800G CPO (unveiled 2025).
• Each GPU requires 6 pluggable transceivers consuming 30W each.
• 100K GPU cluster = ~200K transceivers (100K scale-up + 100K scale-out).
• Scaling to 1M GPUs would consume ~180MW in optics alone.

Source: NVIDIA LinkX

5.5 ASIC-to-Transceiver Demand Formula

Transceiver_Demand_Surge = f(ASIC_GA + Switch_GA_Lag + Qualification_Lag)

Where:
  ASIC_GA: Broadcom ships to OEMs
  Switch_GA_Lag: OEM builds switch (+6-12 months)
  Qualification_Lag: Customer qualifies transceiver (+3-6 months)

Total: ASIC ship → transceiver demand surge: 9-18 months

Demand magnitude:
  Per TH5 switch: 64x 800G transceivers = 64 modules
  Per TH6 switch: 64x 1.6T or 128x 800G transceivers

---

6. Predictive Models for Future Products

6.1 3.2T Transceivers

Signals to watch:

• Coherent demoed 3.2T pluggable technologies at OFC 2026
• LightCounting added 3.2T to forecasts in July 2024
• IEEE 802.3 expected to start 400G/lane standardization work post-802.3dj
• Broadcom TH7 (204.8T) roadmapped for ~2027

Predicted timeline:

• Samples: 2027
• GA: 2028
• Volume: 2029

6.2 CPO (Co-Packaged Optics)

Market forecasts:

Source	2025	2026	2030+
Precedence Research	$95M	$124M	$1,055M (2034)
Mordor Intelligence	$121M	$165M	$764M (2031)
IDTechEx	—	—	$20B+ (2036)
LightCounting	—	—	LPO+CPO >$10B (2026)

Key milestones:

• Broadcom Humboldt (1st gen CPO): Jan 2021 (Tencent deployed)
• Broadcom Bailly (TH5 CPO, 51.2T): 2024 — 50K+ shipped in 2025
• Broadcom Davisson (TH6 CPO, 102.4T): 2025 announced
• NVIDIA Quantum-X800: 144x 800G CPO, shipping H2 2025
• IEEE 802.3 CPO at 800G/1.6T ratification: expected late 2027
• Large-scale CPO deployments: 2028-2030 (Yole Group)

Impact on pluggable revenue:

• Pluggables remain majority of DC optical links through the decade (LightCounting).
• CPO captures scale-up (GPU-to-GPU) first; pluggables retain scale-out (DC-to-DC).
• CPO for scale-up is the "killer application."

Sources:

• Precedence Research
• IDTechEx
• EDN - CPO in 2026
• Lightwaveonline
• Broadcom CPO

6.3 LPO (Linear Pluggable Optics)

Adoption timeline:

• 2024: ~few hundred 800G LPO units (NVIDIA primary customer)
• 2025: 1-2M units; >40% of short-reach 800G links in AI DCs by late 2025
• 2027: >8M 1.6T LPO ports expected
• LPO MSA 100G/lane spec finalized: March 2025
• CAGR >35% through 2033

Power advantage: 1.6T LPO = ~10W vs. conventional 1.6T = 30W+

Source:

• LPO MSA
• Gigalight - LPO & CPO

6.4 Silicon Photonics vs. InP Market Share Evolution

Year	SiPh Share	InP/GaAs Share
2022	24%	76%
2025	30%	70%
2028	44% (projected)	56%
2030	60% (projected)	40%

Driver: LPO and CPO designs overwhelmingly use SiPh platforms. All LPO/CPO devices (except VCSELs) will be SiPh-based.

InP retains strategic importance for: coherent transceivers, high-performance lasers, and vertical integration (Coherent, Lumentum).

Source:

• LightCounting SiPh report
• EE Times

---

7. Recommended Implementation for TIP

7.1 Core Model: Multi-Generation Norton-Bass with Price Erosion

interface TransceiverGeneration {
  name: string;           // e.g., "100G QSFP28"
  speed_gbps: number;     // 100, 400, 800, 1600
  launch_year: number;    // datacom first commercial ship
  market_potential_m: number; // total addressable units (millions)
  p: number;              // innovation coefficient (0.01-0.03)
  q: number;              // imitation coefficient (0.2-0.4)
  asp_launch: number;     // ASP at launch ($)
  price_decay_lambda: number; // exponential decay rate
  form_factor: string;    // SFP+, QSFP28, QSFP-DD, OSFP, OSFP-XD
}

// Revenue model for generation i at time t
function generationRevenue(gen: TransceiverGeneration, t: number, nextGen?: TransceiverGeneration): number {
  const F_t = bassCumulativeAdoption(gen.p, gen.q, t - gen.launch_year);

  // Cannibalization by next generation
  let cannibalization = 0;
  if (nextGen && t >= nextGen.launch_year) {
    const F_next = bassCumulativeAdoption(nextGen.p, nextGen.q, t - nextGen.launch_year);
    cannibalization = F_next;
  }

  const units_in_use = gen.market_potential_m * F_t * (1 - cannibalization);
  const asp = gen.asp_launch * Math.exp(-gen.price_decay_lambda * (t - gen.launch_year));

  return units_in_use * asp;
}

// Bass cumulative adoption
function bassCumulativeAdoption(p: number, q: number, t: number): number {
  if (t < 0) return 0;
  return (1 - Math.exp(-(p + q) * t)) / (1 + (q / p) * Math.exp(-(p + q) * t));
}

7.2 Calibrated Parameters for Known Generations

Generation	m (M units)	p	q	ASP₀ ($)	λ (decay/yr)	Launch
10G SFP+	500	0.015	0.30	500	0.25	2008
40G QSFP+	100	0.010	0.25	800	0.30	2012
100G QSFP28	400	0.020	0.35	2000	0.38	2015
400G QSFP-DD	300	0.025	0.35	1500	0.35	2019
800G OSFP	250	0.030	0.40	700	0.30	2024
1.6T OSFP-XD	200	0.035	0.40	2000	0.35	2026

Note: These are initial estimates to be calibrated against LightCounting/Cignal AI data. Parameters should be fitted using nonlinear least squares on observed shipment data.

7.3 Geographic Revenue Multiplier

interface RegionConfig {
  name: string;
  adoption_lag_months: number;
  market_share_pct: number;
  p_multiplier: number;  // adjust innovation coefficient
  q_multiplier: number;  // adjust imitation coefficient
}

const REGIONS: RegionConfig[] = [
  { name: "US Hyperscaler",    adoption_lag_months: 0,  market_share_pct: 35, p_multiplier: 1.5, q_multiplier: 0.8 },
  { name: "China Hyperscaler", adoption_lag_months: 9,  market_share_pct: 25, p_multiplier: 1.2, q_multiplier: 1.0 },
  { name: "Japan/Korea",       adoption_lag_months: 15, market_share_pct: 10, p_multiplier: 1.0, q_multiplier: 1.1 },
  { name: "Europe Telco",      adoption_lag_months: 30, market_share_pct: 15, p_multiplier: 0.7, q_multiplier: 1.2 },
  { name: "India/SEA/LATAM",   adoption_lag_months: 48, market_share_pct: 15, p_multiplier: 0.5, q_multiplier: 0.6 },
];

7.4 Conference Signal Pipeline Tracker

interface TechnologySignal {
  technology: string;
  signal_type: "OIF_SPEC" | "IEEE_STANDARD" | "MSA_SPEC" | "OFC_DEMO" | "ECOC_DEMO" | "LC_FORECAST_ADD" | "FIRST_SHIP" | "VOLUME";
  date: Date;
  predicted_volume_date: Date; // computed
  confidence: number;         // 0-1
}

// Pipeline lag by signal type (months to volume production)
const SIGNAL_TO_VOLUME_LAG: Record<string, number> = {
  "OIF_SPEC":        36,  // 30-42 months
  "IEEE_STANDARD":   18,  // 12-24 months
  "MSA_SPEC":        30,  // 24-36 months
  "OFC_DEMO":        21,  // 18-24 months (multiple vendor demos)
  "ECOC_DEMO":       24,  // 18-30 months
  "LC_FORECAST_ADD": 27,  // 24-30 months
  "FIRST_SHIP":      12,  // 9-15 months
};

7.5 ASIC Demand Correlation Model

Transceiver_Revenue(t) = Σ [Switch_Shipments(ASIC_gen, t - lag) * Ports_Per_Switch * ASP(speed, t)]

Where:
  ASIC generations: TH4→TH5→TH6→TH7
  lag = 9-18 months (ASIC ship → transceiver surge)
  Ports_Per_Switch: 64 (TH5), 64-128 (TH6)

Monitor: Broadcom ASIC announcements as leading indicator
         → OEM switch GA as confirming signal
         → Transceiver qualification as demand signal

7.6 Key Metrics Dashboard for TIP

For each transceiver generation, TIP should compute and display:

1. Lifecycle Stage: {Pre-launch | Ramp | Growth | Peak | Decline | EOL}
2. Time to Peak Revenue: Derived from Norton-Bass fit
3. Current ASP vs. Launch ASP: Price erosion percentage
4. Revenue Duration >50% Peak: How many quarters remaining above half-peak
5. Cannibalization Index: What % of market potential is being captured by next gen
6. Geographic Heatmap: Adoption stage by region
7. Leading Indicators: Conference demos, spec milestones, ASIC launches

7.7 Data Sources for Calibration

Source	Data Type	Access	Cost
LightCounting	Revenue, shipments, ASP by speed	Subscription	$$$
Cignal AI	Datacom revenue, component market	Subscription	$$$
Dell'Oro	Ethernet switch/router market	Subscription	$$$
Yole Group	SiPh, CPO market forecasts	Reports	$$
IDTechEx	CPO market forecasts	Reports	$$
Broadcom press releases	ASIC launch dates	Free	$0
OFC/ECOC proceedings	Demo tracking	Conference fee	$
IEEE 802.3 minutes	Standards timeline	Free	$0
Company earnings calls	Revenue by segment, guidance	Free (SEC filings)	$0
Innolight/Coherent 10-K	Supplier revenue, growth rates	Free (SEC/CSRC)	$0

---

Appendix A: Key Reference Papers

1. Bass, F.M. (1969). "A New Product Growth for Model Consumer Durables." Management Science.
2. Norton, J.A. & Bass, F.M. (1987). "A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products." Management Science, 33(9).
3. Jiang, Z. & Jain, D.C. (2012). "A Generalized Norton-Bass Model for Multigeneration Diffusion." Management Science, 58(10), 1887-1897.
4. Meade, N. & Islam, T. (2006). "Modelling and forecasting the diffusion of innovation - A 25-year review." International Journal of Forecasting.
5. Tsai, B.H. (2013). "Predicting semiconductor industry growth." Technological Forecasting and Social Change. (Gompertz curve application)
6. Jaafari, A. (2019). "Using Weibull Distribution for Modeling Bimodal Diffusion Curves." Int. J. Innovation and Technology Management.

Appendix B: All Sources Used

• Bass diffusion model - Wikipedia
• IEEE Xplore - Technology forecasting using Bass model
• GNB Model - INSEAD
• GNB Model - INFORMS
• GNB Model - SSRN
• GNB Model - Iowa State
• R diffusion package
• Heterogeneity in diffusion - ScienceDirect
• Bass model broadband Europe - ScienceDirect
• Bass model BRIC - Academia.edu
• Agent-based Bass - Tandfonline
• Gompertz for semiconductors - EE Times
• Gompertz for semiconductors - Semiengineering
• Weibull for bimodal PLC - World Scientific
• Weibull for tech change - ScienceDirect
• MarketsandMarkets - Optical Transceiver
• Cignal AI - 800G shipments 2025
• Cignal AI - 20M 400G/800G 2024
• LightCounting - Sales of 800G
• LightCounting - $23B in 2025
• LightCounting - Ethernet optics 2024
• LightCounting - Market forecast
• Coherent - 800G ZR GA
• Coherent - 1.6T VCSELs
• Coherent - 3.2T at OFC 2026
• Eoptolink - 1.6T DR4 OFC 2026
• PrecisionOT - 400G ZR
• Deep Fundamental - Module Market
• Pluggables Power Geopolitics - Substack
• Broadcom TH5
• Broadcom TH6
• Broadcom TH4
• Broadcom CPO
• TechInsights - TH5
• NextPlatform - TH6
• NextPlatform - CPO
• ServeTheHome - TH6
• Arista 7800R4
• Cisco Nexus 9000
• NVIDIA LinkX
• Precedence Research - CPO
• IDTechEx - CPO
• EDN - CPO 2026
• Lightwaveonline - LPO CPO
• LPO MSA
• LightCounting - SiPh
• EE Times - AI reshapes photonics
• Nature Communications - SiPh roadmap
• AIM Photonics - Commercialization
• IEEE 802.3 - Wikipedia
• FS.com - MSA standards
• Hector Weyl - AI optical networking