InP 晶圆与 EML 短缺：800G 和 1.6T 光模块瓶颈对 BOM 意味着什么

随着 InP 晶圆和 200G/通道 EML 产能跟不上超大规模 AI 建设，800G 和 1.6T 光模块需求约超出合格供应 30%。

已发布 已更新 4 分钟阅读 ✓ AI 就绪 分享

TL;DR——三句话

  • Transceiver demand runs approximately 30% ahead of qualified supply, with 800G+ shipments past 60 million units in 2026 (S1).
  • The tightest bottlenecks are InP crystal boule growth, wafer epitaxy yield, and qualified 200G/lane EMLs, not module assembly (S1).
  • Three suppliers - Sumitomo Electric, AXT (Beijing Tongmei), and JX Advanced Metals - dominate polished merchant InP wafers; China refines roughly 70% of raw indium (S1).

摘自 key_takeaways （真实字段，最多 3 条）。事实与分析分别标注。

InP 晶圆与 EML 短缺：800G 和 1.6T 光模块瓶颈对 BOM 意味着什么Editorial cover for a component sourcing brief. Chip motif; headline "InP 晶圆与 EML 短缺：800G 和 1.6T 光模块瓶颈对 BOM 意味着什么". No reported key numbers on this brief, so no number strip is drawn. 1 cited source. Static cover (no animation).InP 晶圆与 EML 短缺：800G 和 1.6T 光模块瓶颈对 BOM意味着什么XAIPART
Cover image generated from data already shown on this page: the category motif, the headline, and reported key numbers. No decorative illustration; no third-party image.

您从 AI 回答跳转而来——这是可验证的原文

AI 摘要可能遗漏了其结论的依据。原始预测、受影响料号及所有引用均在本页，均可追溯至来源。

XAIPART 预测

本篇简报包含可证伪的预测。

跳转至完整预测及其证伪规则 ›
受影响的料号

此报道在目录中无匹配——我们选择不显示，而不是推断匹配。

了解判定方式 ›
需要采购料号？发送请求 → 询价

Reported by Exponential Industry on 2026-10-12. Summarized and analyzed by XAIPART Editorial.

What happened

Optical transceivers at 800G and 1.6T speeds depend on indium phosphide (InP), the only commercial material that efficiently emits light at the 1310 nm and 1550 nm wavelengths used in AI optical fabrics; even silicon-photonics designs must pair with an external InP continuous-wave laser.S1 The source identifies the tightest constraints not in module assembly but upstream: growth of high-purity InP crystal boules, wafer epitaxy yield, and qualification of 200G-per-lane electro-absorption modulated lasers (EMLs).S1

Transceiver demand reportedly runs approximately 30% ahead of qualified supply, with 800G-and-above shipments scaling past 60 million units in 2026, driven primarily by hyperscale AI clusters at Nvidia, Google, Microsoft, Amazon, and Meta.S1 Total shipments of approximately 150 million 800G-and-above modules are expected across 2026 and 2027.S1 Suppliers including InnoLight, Eoptolink, Coherent, and Lumentum are already shipping 1.6T modules in volume, with production backlogs extending into 2027.S1

On the substrate side, most raw indium is a byproduct of zinc smelting, with China refining roughly 70% of global output; polished merchant InP wafers are concentrated among Sumitomo Electric, AXT (via Beijing Tongmei), and JX Advanced Metals.S1 After China introduced export restrictions and licensing rules, 6-inch wafer prices rose, prompting AXT, Lumentum, and Coherent to secure capacity with substantial prepayments.S1 Mitigations under way include a transition from 2-inch to 6-inch InP wafers, multi-year supply contracts, a shift toward silicon photonics with external CW lasers, and linear pluggable optics (LPO) to cut DSP power ahead of co-packaged optics (CPO) scaling.S1 Post-consumer recycling of indium from retired modules is described as economically impractical because modules contain only sub-milligram amounts.S1

Key numbers
MetricReported value
Demand vs. qualified supply~30% ahead of supplyS1
800G+ shipments, 2026past 60 million unitsS1
800G+ shipments, 2026–2027~150 million unitsS1
China share of global indium refiningroughly 70%S1
Exact EML or wafer capacity figuresNot disclosed

Background timeline

  1. - 800G-and-above shipments scale past 60 million units, led by hyperscale AI clusters.S1
  2. - The source analysis of the InP and EML bottleneck is published with a 2026-09-21 dateline.S1
  3. - Production backlogs at InnoLight, Eoptolink, Coherent, and Lumentum extend into 2027, with roughly 150 million 800G+ units expected across 2026 and 2027.S1

XAIPART take

Technical impact

Designers of 800G and 1.6T links should treat laser choice as a supply decision, not just a link-budget decision: qualified 200G/lane EMLs from Lumentum and peers are the binding constraint. Specifying silicon photonics with an external CW laser, or an LPO variant that removes the DSP, widens the pool of qualifying designs and lowers module power. Sub-micron fiber attach still concentrates precision assembly at specialists such as Fabrinet, so second-source qualification takes longer than typical board-level components.

Market impact

With demand running about 30% ahead of qualified supply and backlogs reaching into 2027, non-hyperscale buyers face the longest queues. Procurement teams should expect allocation-based ordering, multi-year agreements, and prepayment terms similar to those AXT, Lumentum, and Coherent have already accepted. Wafer-price increases after China's export rules suggest substrate cost pass-through into 2027 module quotes.

Action

Place 800G/1.6T module orders with named allocation windows rather than open-ended lead times, and qualify at least one LPO or silicon-photonics alternative per link. For networking line cards and switch rebuilds, XAIPART recommends confirming DSP (Broadcom/Marvell) and module vendor lead times before locking board-spin schedules. Ask suppliers explicitly about wafer-diameter transition status, since 6-inch capacity changes cost and availability curves.

XAIPART Forecast

XAIPART analysis: if qualified 200G/lane EML capacity does not close the roughly 30% supply-demand gap by Q4 2027, pluggable 800G/1.6T module lead times for enterprise and carrier buyers will remain above the allocation queues hyperscale buyers face, and LPO adoption will exceed current design-in share as a DSP-power workaround before Coherent-led CPO reaches volume. If published hyperscale shipment figures for 2027 fall short of the source's trajectory toward 150 million cumulative units, this forecast is wrong.

XAIPART analysis

Facts are as reported by the cited source. The design-in, sourcing and forward-looking statements above are XAIPART's analysis, not the source's claims. The ~30% supply gap, exact EML capacity volumes, and specific contract values are single-source figures the source does not fully quantify; the article cites third-party data (TrendForce, Yole, Nikkei) that XAIPART has not independently verified.

Corrections: none at publication. This brief will be updated with a timestamp if the source corrects its report.

Dated developments, extracted from this page's timeline (curated milestones = hollow markers).
Timeline of dated developmentsDated entries: 2026-01-01 - 800G-and-above shipments scale p…; 2026-09-21 21, 2026 - The source analysis of …; 2027-01-01 - Production backlogs at InnoLight….2026-01-01- 800G-and-above shipments scale p…2026-09-2121, 2026 - The source analysis of …2027-01-01- Production backlogs at InnoLight…

替代料与生命周期

已收录的替代料——仅为目录范围，不代表市场排名
料号关系生命周期依据
— 暂无收录替代料 待补充数据 — 需 match_type 后才可显示任何前代标签

“前代”和“替代”标签仅来自已记录的匹配类型、生命周期状态或来源文本。没有记录匹配时，不显示任何标签。我们自己的商品列表不构成市场参考。

关注要点

可证伪预测——阈值、时间窗口和当前状态。被推翻的预测仍保留可见。
论断阈值发布方状态
: if qualified 200G/lane EML capacity does not close the roughly 30% supply-demand gap by Q4 2027, pluggable 800G/1.6T module lead times for enterprise and carrier buyers will remain above the allocation queues hypers... published hyperscale shipment figures for 2027 fall short of the source's trajectory toward 150 million cumulative units Q4 2027 Pending

被证明错误的预测仍保留可见并加以标注。我们不会删除它们。

受影响的料号

目录中无匹配。 此事件未映射到 XAIPART 当前上架的任何料号。我们不推断匹配，也不显示占位列表。

各维度成本压力

Source: FRED · Generated · reference output z and composite are algorithm outputs, not a fetch snapshot

Cost pressure by dimension — composite stateDiverging bars for three dimensions around a zero midpoint. Input Cost has no reading; Supply Demand z -1.4711, notable; Fx Energy z 0.2546, neutral. Composite index -9.1, band neutral.0投入成本no reading供需z −1.4711 · notable46 d汇率与能源z +0.2546 · neutral5 dcomposite −9.1 · neutral−100 · −40 · −15 · 0 · +15 · +40 · +100
Underlying values — copyable for verification.
DimensionFRED series IDUnitLagAs of
投入成本PCOPPUSDM, PALUMUSDM, PNICKUSDMUSD/mt52 d
供需PCU334413334413, IPG3344Sindex(1982=100) / index(2017=100)46 d
汇率与能源DEXCHUS, DCOILBRENTEU, DHHNGSPrate / USD/bbl / USD/MMBtu5 d

No free authoritative interface for: gold · silver · rare earth · polysilicon · lithium carbonate · market share — listed rather than approximated.

本文如何生成

过程证据——抓取、来源可靠性检查、模型与校验
本简报的审计记录
Source fetched Exponential Industry · exponentialindustry.com · 6,848 chars
Source sanity (S1) ok — passed before any model call (no s1_ block recorded)
Model / prompt glm-5.3-flash · prompt Blog Agent - News Summary (v2.8-final) · 7,661 tokens · 39.7s · run completed
Gate verdict hold · 6/6 applicable checks passed · spec v2.9 · C10 ratio 0.015849 · C10 rounds 0
Time contract 3 <time datetime> values in this brief — 0 violations (+08:00 or date-only)

上述每一行均由流水线生成，而非人工撰写。

编辑评分

确定性校验结果——纯 PHP 实现，判定过程不涉及模型
校验评分明细 6 项适用检查， 6项通过： 100 %. Hold items: C13 (falsifiable forecast) — sent to human review. Gate spec: v2.9 (S0 + S1 + C1–C20) · Source affinity (C20): 0.8784. 适用检查已通过 6 / 6 = 100% · 判定： hold Hold items: C13 (falsifiable forecast) — sent to human review Gate spec: v2.9 (S0 + S1 + C1–C20) · Source affinity (C20): 0.8784 是否发布由校验决定。独立审计评分仅供参考。

相关报道

发现错误？ 告诉我们哪里有问题 ——更正内容会与原文一同发布并记录在案。我们不会悄悄修改文章。

评论

成为第一个评论的人。

来源

  1. Inside the 800G and 1.6T optical transceiver bottleneck InP wafers and AI fabrics — https://exponentialindustry.com/blog/2026-09-21-800g-1-6t-optical-transceivers-inp-ai-supply-chain · rss · 访问于

事实标注为 S1/S2，并可追溯至所链接的来源。标注为 分析 的内容是 XAIPART 自身的推理，与报道的事实分开呈现。来源链接会被访问核实，而非 nofollow ——署名是有意为之。