Optics Replacing Copper, Now Backed by Earnings: Broadcom Extends Visibility to 2027, but the Hardest Bottleneck Is a Single Laser (W24)

2 days ago

7 min read

This week the bull case for optical communications got a new spokesperson — not analyst presentations, but Broadcom's income statement. Yet what's truly scarce can't be bought with earnings.

Introduction

On June 3, Broadcom reported Q2 FY2026: consolidated revenue of $22.2 billion, up 48% year over year, including $10.8 billion from AI semiconductors, up 143%. Guidance for next quarter's total revenue went straight to $29.4 billion, up 84% (sources: Motley Fool earnings call transcript, Broadcom IR). This isn't a company "expecting AI to be good" — it's a company writing the fact that AI is already ramping into the line items of its financial statements.

Over the past two years, lines like "optics replacing copper" and "AI infrastructure eats all the bandwidth" have been repeated countless times, but mostly in analyst price targets and trade show demos. This week's watershed: earnings from Broadcom and Credo, plus Marvell's completed product lineup, let the demand side of this narrative "speak entirely in numbers" for the first time. Only one question remains — with demand locked in by earnings, can the supply chain keep up? The answer lies in a laser at the very top of the chain that few people watch.

1. Optics Replacing Copper: From PowerPoint to Income Statement

Starting Point: Broadcom Puts Its "AI Twin Engines" in the Past Tense

Broadcom's bull thesis has always been the twin engines of custom ASICs plus AI networking. This quarter's call moved that story from future tense to realized numbers: AI semiconductors at $10.8 billion, up 143%, with next-quarter guidance accelerating further to 84% year-over-year growth (sources: Motley Fool, Broadcom IR). When a company's AI business, already growing triple digits off a high base, keeps accelerating rather than slowing next quarter, it means demand exceeds supply — not that orders have peaked.

The technology track is moving too. This quarter Broadcom taped out a 200T switch chip, the next-generation Tomahawk 7 (2027) doubles bandwidth again, and third-generation CPO advances to 200G per lane (sources: TechTimes, Broadcom IR). Stack these together and Broadcom has effectively extended visibility for the entire AI networking supply chain out to 2027 — for downstream players like TSMC advanced packaging, ASE, and switch maker Accton, it's an order map with a clear schedule.

Evidence: Broadcom Isn't the Only One Reporting Good News

If it were only Broadcom, this could be read as the story of a single strong player. But this week, good news came across the board. Credo's Q4 FY2026 revenue topped $1.3 billion, more than tripling year over year, with non-GAAP net income above $662 million; FY2027 guidance calls for growth above 80%, with optical revenue targeted above $600 million (sources: Seeking Alpha, The Globe and Mail). Marvell rounded out its full 1.6T optical DSP family (Ara T / Ara X / Petra / Aquila M) in one go, then acquired plasmonics silicon photonics company Polariton, pointing its optical performance roadmap beyond 3.2T (sources: Marvell IR, HPCwire).

Real shipments on the demand side line up as well. Innolight's 2026 Q1 revenue was RMB 19.496 billion, up 192% year over year, with net income of RMB 5.735 billion, up 262%; it holds more than 40% of the global optical module market and over half of 1.6T (sources: Stockstar, East Money). Looking further up at orders: Goldman Sachs raised its 2025/2026 800G module volume estimates to 19.9 million / 33.5 million units, and total demand for 800G and above is estimated to jump from about 24 million units in 2025 to nearly 63 million in 2026, roughly 2.6x (sources: Anue, Yuanta Investment Consulting).

Interpretation: Valuation Led the Run — Now Earnings Take the Baton

Put the three threads side by side — earnings (Broadcom, Credo), product lineups (Marvell's 1.6T family), and shipment volumes (Innolight Q1 plus Goldman's numbers) — and this week's significance is clear: the 2026 optical communications rally has evolved from "the market expects 1.6T to ramp" to "earnings prove 1.6T is ramping."

The difference matters more than it sounds. The former is valuation running on analyst optimism; the latter is earnings running on shipment orders. When earnings take over from valuation, the supply chain's visibility no longer depends on sentiment but on capacity. That's exactly why the deciding factor is no longer "who tells the best story" but "who can actually ship." Follow this logic upstream and you hit a component that everyone needs but only a handful can make.

2. The Hardest Bottleneck Isn't the DSP or CPO — It's the Laser

Starting Point: Shortages Through 2027, at the Very Top of the Chain

While downstream DSP, switch and optical module makers race to expand capacity, the tightest link in the chain has actually moved to the very top: the light source. This week's signals are blunt: shortages of EML (electro-absorption modulated laser) and CW (continuous-wave) lasers now extend into 2027, and NVIDIA, to lock in capacity, has pushed EML laser chip lead times beyond 2027 (sources: Commercial Times, MoneyDJ, Business Today).

What makes this so damaging is where it sits. Whether optical modules can be built, and whether silicon photonics CPO can light up, ultimately comes back to whether there's a laser to serve as the light source. DSPs can add another foundry, PCBs can add another line, but the learning curves for InP substrate growth, epitaxy and EML processes are measured in years. When the top of the chain is stuck, no amount of downstream capacity expansion does anything but spin its wheels.

Evidence: LandMark's Long-Term Contract and NVIDIA's Equity Stake Are Really a Grab for Strategic Supplies

To see how scarce this laser is, just watch where capital is moving. LandMark Optoelectronics (3081) pushed gross margin to about 48% this quarter — an almost luxurious number for an upstream optical component maker. On April 22 it signed a five-year InP substrate contract with Sumitomo Electric covering 2026 to 2030 and launched an expansion of 4 InP plus 1 GaN MOCVD tools, while NVIDIA invested roughly NT$40 billion for an equity stake (sources: SinoPac RichClub, MoneyDJ).

When a customer like NVIDIA is willing to take equity stakes to lock in capacity at upstream laser makers (Lumentum and Coherent), it's no longer an ordinary procurement relationship — it's stockpiling laser chips as strategic supplies. This shortage also ripples through a whole row of Taiwanese companies — Luxnet (4979), Truelight (3234), VPEC (2455) and GCS Holdings (4991) — with InP epitaxy outsourced to IntelliEPI (4971) and VPEC (sources: Commercial Times, Anue). The more downstream demand is proven real by earnings, the more pricing power concentrates upstream in this laser.

Interpretation: In This Cycle, the Bottleneck Decides Who Makes the Cleanest Money

Optical communications investors tend to focus on switches, DSPs and CPO timelines, because that's where the buzz is. But put this week's two threads together and you reach a counterintuitive conclusion: once demand is locked in by earnings, the players most certain to benefit from the supply-demand imbalance are actually the upstream light-source makers.

Downstream competes by adding capacity; upstream competes on scarcity. When everyone is short of the same laser, pricing power isn't in the buyer's hands.

The practical takeaway for STT readers: to gauge the health of this optical rally, don't just look at how fast module makers' revenue is growing — watch upstream light-source lead times and margins. Longer lead times and higher margins are hard proof that demand is real and supply can't keep up. The day EML lead times start to loosen and upstream margins come down is when this supply chain needs a fresh checkup. Right now, not a single loosening signal has appeared.

3. Other Signals Worth Tracking This Week

The five largest cloud providers' combined 2026 capex is projected at $660–690 billion, nearly double 2025, with about 75% ($450 billion) tied to AI infrastructure; Microsoft also disclosed $80 billion in Azure orders it can't fulfill due to power constraints (sources: Tom's Hardware/Futurum). The compute bottleneck is spilling over from chips to power.
NVIDIA's Spectrum-X Photonics CPO switch integrates CPO into the ASIC, with bandwidth up to 409.6 Tb/s and availability in 2H 2026, touting 5x power savings and 10x reliability; initial customers named include CoreWeave, Lambda, Meta, Microsoft and Oracle (source: NVIDIA Newsroom). The timeline for CPO's first year of commercial deployment is pinned down one notch tighter.
TSMC's silicon photonics COUPE is set for volume production in 2026, with the accompanying 200 Gbps MRM also entering volume production in 2026 (sources: SemiWiki, SinoPac RichClub). The foundation for Taiwan's CPO ecosystem is being laid.
Accton's May revenue rose 184.76% year over year, riding the switch upgrade cycle from 800G to 1.6T (source: SinoPac RichClub).
Fabrinet's optical communications revenue was $833 million, up 29%, and telecom revenue was $554 million, up 59% (sources: TradingView/Zacks); Coherent disclosed that scale-out CPO revenue will start in 2H 2026 and scale-up in 2H 2027 (sources: Futurum, Lumentum IR).

Conclusion

This week's takeaway is clean: the bull case for optical communications is now backed by earnings — Broadcom up 143%, Credo tripling, Innolight nearly tripling — and demand is no longer an expectation but a reality. Yet on the same supply chain map, the hardest bottleneck has moved to the laser at the very top: EML/CW lasers are short through 2027, and even NVIDIA has to take equity stakes to lock in capacity. The more real the demand, the more valuable the light source.

Two things are worth watching next. First, whether Q2 revenue at downstream module and switch makers can keep up with the visibility set by these earnings — Broadcom has raised the bar to 2027, and Taiwanese companies' numbers this quarter will be the first report card. Second, and more critically, upstream light-source lead times and margins: whether LandMark's 48% gross margin holds and whether EML lead times loosen will be the best thermometer for how much longer this supply-demand imbalance can run. Until those two signals turn, earnings will keep speaking for the optics-replacing-copper narrative.

This article is for technology and industry trend analysis only and does not constitute investment advice.

STT Market Insights | June 8, 2026 | Compiled by the STT Editorial Team