Capex cycle

Coherent’s capex cycle through FY2026–FY2028 is the operational expression of its AI-photonics thesis. Sherman, TX is the load-bearing capacity site — the world’s first 6-inch InP wafer fab and the source of Coherent’s structural cost advantage in EML production. Capex has accelerated from ~7% of revenue in FY2024 to ~9% in FY2026 H1, with management guidance of continued sequential increases through FY2026 to support Datacenter & Communications capacity expansion. The DENSO + Mitsubishi $1B SiC JV (Dec 2023) restructured the SiC capex burden — the Silicon Carbide LLC subsidiary funds its own capex from internal cash flow, lifting the SiC build out of Coherent’s parent-level capex envelope.

Capex trajectory

Period	Capex ($M)	% of revenue	Cumulative ($M)
FY2023	~$330	~6.7%	$330
FY2024	~$340	~7.2%	$670
FY2025	$440.8	7.6%	$1,111
H1 FY2026	$257.5	7.9%	$1,368
Q2 FY2026 alone	$154	9.1%	n/a
FY2026 (forecast)	$550–650	~9–10%	$1,660–1,760
FY2027 (forecast)	$700–900	~10–12%	$2,360–2,660
FY2028 (forecast)	$700–900	~10–11%	$3,060–3,560

Source: FY2025 results ✓; Q1 + Q2 FY2026 releases ✓; FY2027–28 projections are derived from management capex commentary on capacity-build acceleration.

The capex acceleration is consistent with management framing on Q2 FY2026: “support rapid capacity expansion in datacenter and communications” with sequential capex increases planned (Q2 FY2026 release) ✓.

Sherman TX 6-inch InP fab capacity expansion — load-bearing

Coherent announced in March 2024 the world’s first 6-inch InP scalable wafer fabs, located at Sherman TX and Järfälla Sweden (Coherent press release) ✓. Key parameters:

Metric	Value
Wafer-size advantage	4× more chips per wafer vs 3-inch; ~2.5× more vs 4-inch
Die-cost reduction	~60% vs 3-inch (Coherent disclosure)
Sherman TX capex	Multi-hundred-million-dollar multi-year build
Texas Semiconductor Innovation Fund grant	$14,076,031 awarded Feb 2026; project total >$154M (Semiconductor Today Feb 9, 2026) ◐
6-inch ramp progress (Q1 FY2026)	~80% of target wafer-start rate
Internal InP capacity doubling	Targeted Q4 CY2026
Products on 6-inch	200G EML, DFB-MZ, 100G EML, photodetectors, high-power CW lasers

The Sherman fab build is the single largest operational capex commitment in Coherent’s near-term horizon. The 6-inch ramp is ahead of plan — at ~80% target wafer-start rate as of Q1 FY2026 commentary, with the doubling target anchored at Q4 CY2026.

SiC substrate capacity — JV-funded structure

Per the December 2023 transaction with DENSO and Mitsubishi Electric (Coherent press release) ✓:

DENSO + Mitsubishi each invested $500M for 12.5% non-controlling interest = $1.0B total at JV-level
Coherent retained 75% ownership in Silicon Carbide LLC
All operating and capital expenses of the Business will be funded by the Business

Translation: SiC capex no longer rolls up to Coherent’s parent capex line. The SiC subsidiary funds its own 200 mm wafer capacity expansion from JV-level cash, with the JV partners contributing growth capital as needed. This reduces Coherent’s parent-level capex burden by approximately $100–200M annually vs the pre-2023 structure.

200G/lane EML ramp + 1.6T module capacity

Capacity-build for 200G/lane EML production is the most strategically-leveraged piece of the capex envelope:

Capacity action	Timing	Capex bucket	Strategic role
6-inch InP wafer-fab ramp Sherman TX	CY2025–2026	Sherman TX fab capex ($hundreds of millions)	Source-laser supply for 1.6T pluggables + CPO
200G EML production ramp on 6-inch	CY2026	Embedded in fab capex	Capacity for AI-cycle demand
1.6T module assembly capacity	Allen TX + Sherman TX assembly	$50–100M annually	Module-form-factor flexibility
VCSEL 200G ramp	2H CY2026	Sherman TX existing line	Short-reach 1.6T option
Silicon-photonics scaling	OFC 2026 demos; production timeline confidential	$50–100M annually	CPO-engine optionality
CPO module assembly	CY2027–2028	TBD	Vertical-integration into CPO

The aggregate incremental capex through FY2028 is on the order of $2–3B vs FY2024 baseline (cumulative). The NVDA $2.0B equity injection plus the multi-billion-dollar purchase commitment underwrites the demand-side of this capex envelope — the ROI math depends on capacity utilization at NVDA-anchor demand levels through FY2028.

CHIPS Act / federal-incentive uplift

Coherent benefits from US-domestic-fab incentives:

Program	Estimated benefit
CHIPS Act 25% Investment Tax Credit (ITC)	~25% of qualifying US fab capex; on Sherman TX 6-inch InP fab build, this translates to potentially $50–100M+ of credits
Texas Semiconductor Innovation Fund grant	$14,076,031 cash (Feb 2026) ◐
CHIPS Act direct manufacturing grants	Coherent applied; specific awards confidential ⚠
DOE / DOD photonics R&D grants	Various smaller awards ⚠

The 25% federal ITC effectively reduces net Sherman TX capex by ~25% on qualifying spend — a material economic uplift for the fab investment math. This is one factor making the 6-inch InP capacity build economically attractive even at the high upfront capital intensity.

Capacity utilization and ROI math

The capex-ROI argument depends on:

Capacity utilization at NVDA-anchor demand — NVDA’s multi-billion-dollar purchase commitment underwrites a meaningful share of capacity dedicated to NVDA. The remainder must be filled by other hyperscaler / module-vendor demand.
ASP discipline holding — supply-tightness pricing supports above-historical gross margins on EML chips and 1.6T modules.
6-inch cost advantage realized at scale — the 60% die-cost reduction from 3-inch to 6-inch is a learning-curve outcome, not a starting-yield outcome. Cumulative wafer starts drive the realization of cost savings.
CPO transition timing — the same fab capacity supports CPO source lasers in 2027–2028, meaning the capex is dual-purpose (pluggable + CPO) rather than single-architecture.

Under the base case (capacity utilization 80%+ through FY2028), the capex envelope supports gross margins in the 38–42% range and operating margins of 19–22%. Under the bear case (NVDA capex digestion + capacity overhang), gross margin compression to 33–35% would compress returns materially.

Capex sensitivity scenarios

Scenario	FY2026–28 cumulative capex ($B)	Capacity utilization	Margin profile	NPV impact
Bull	$2.5–3.0	90%+	GM 40–43%	Positive — 6-inch advantage compounds
Base	$2.0–2.5	80–85%	GM 38–40%	Neutral — capex returns ~hurdle rate
Bear	$1.5–2.0 (under-build)	<80% (over-build)	GM 32–36%	Negative — capacity stranded if AI-capex digests

The bear case requires NVDA demand to digest below committed-volume levels — which would require NVDA to renegotiate the multi-billion-dollar purchase commitment. As of April 2026, no such signal has emerged. The bull case requires hyperscaler-non-NVDA demand to fill the residual capacity at attractive ASPs.