Open in interactive viewer → charts, metric popovers & call review

ams-OSRAM Wants You to Buy 'Wide and Slow' — Just as the Tape Dumps Optical Interconnect

A 16.3% margin quarter with a fixed balance sheet, and a growth narrative that quietly leapfrogged smart glasses on the way to the data hall
AMS.SW · Earnings Call · 2026-11-13

The quarter beneath the headline

ams-OSRAM delivered EUR 829 million of Q3 revenue and a 16.3% adjusted EBITDA margin, above the midpoint of guidance, on a EUR ~1.8 billion market cap. The semiconductor core portfolio grew roughly 3% like-for-like and Opto Semiconductors grew 8%, while the reported top line fell 3% because the portfolio is being dismantled on purpose — roughly EUR 60 million of year-on-year drag from divestments. Lamps & Systems put up a 23.1% margin on an 11% sequential revenue lift, and management again credited the collapse of rival First Brands for pulling customers its way, a refrain that has now run for several quarters. The freshest thing in the quarter was not in the slides but in the Q&A. Consumer revenue fell 13% year-on-year, and the cause was not demand: “That's why the Android phone business is down so much because less phones are being built because of availability and price of memory there.” — Aldo Kamper, CFO or Senior Executive (likely CFO) · 2026-11-13 Aldo Kamper went further, flagging PCB shortages as "much more visible" in recent weeks, with OEM customers rumbling about slowing over Christmas. That is a genuinely company-specific channel through which the AI buildout is hurting handsets rather than helping — and it shows up in the keyword set, where memory appears as one of the quarter's sharpest new risers at ams-OSRAM after barely registering previously. The other half of the portfolio behaved as advertised. “you could say that for LSP, the 9% is kind of the new reality.” — Rainer Irle, CFO or Senior Executive (likely CFO) · 2026-11-13 Management is admirably unromantic about it: strip out the near-zero-margin manufacturing services supplied to Infineon post-close and that segment would have printed around 14%.

The narrative migrated, quietly

Three months ago the sequencing was unambiguous. On the August call Kamper framed the roadmap as a relay: “The opportunity for tomorrow is in smart glass. The opportunity for the day after tomorrow is in the data center space.” — Aldo Kamper, CEO · 2026-08-04 Asked this quarter whether that order still held, he said “No, I would still see it in that sequence.” — Aldo Kamper, CFO or Senior Executive (likely CFO) · 2026-11-13 But the disclosed emphasis has inverted. AI photonics now owns the middle of the prepared remarks and the majority of the Q&A, complete with hard technical proof points — error-free 32 Gb/s NRZ at roughly 0.25 picojoules per bit on a sub-50-micron pitch — while smart glasses gets relegated to an answer about content per pair. The keyword record confirms the migration. microVCSEL is among the strongest gainers of the quarter; smart glasses is among the sharpest decliners. The stated sequence is intact; the investor-facing one has already reordered itself. That gap is the story.

Instead of pushing a small number of optical lanes to ever-higher speeds, we distribute bandwidth across many parallel optical channels operating at lower rates per channel. The target is to remember: optical reach, copper-like latency and a path towards copper-like economics.

Aldo Kamper, CFO or Senior Executive (likely CFO) · 2026-11-13
The credibility scaffolding is real but early: a Credo-led OCP workstream for interoperable short-reach micro-emitter links, the company's micro-emitters inside Credo's active light cable demonstrator, and a fiber/connector partnership with BizLink. Kamper's own framing keeps it honest: “The key elements of the technology stack have been validated, but it is, of course, not yet commercial revenue.” — Aldo Kamper, CFO or Senior Executive (likely CFO) · 2026-11-13 MicroVCSEL first, into active optical cables over 1–75 metres; microLED later, below a metre, for memory disaggregation, with first commercialization pegged around 2029–2030 — roughly the same window Marvell has sketched for scale-in.

The tape is voting the other way

Here is the contrast worth holding on to. Over the last 90 days the market has been actively de-rating the exact complex ams-OSRAM is now leaning into. Co-packaged optics names are down roughly 30% with thirteen of fifteen falling; data center interconnect is down more than 30%; high-performance compute and high-bandwidth memory sit in the same bucket. The 30-day window is not better. A company with a niche, unproven, pre-revenue architecture is asking investors to underwrite it precisely as the tape marks down the incumbents that are already shipping. That is not to say the bet is wrong — wide-and-slow targets a different optimization point than fast-and-narrow, and a parallel architecture that tolerates cheap multimode fiber is a legitimate wedge. It is to say the burden of proof has shifted from narrative to silicon, and 32 Gb/s at pixel level is a component proof point, not a system one.

What went quiet

The sensor business disposal to Infineon dominates the keyword set, and it genuinely repaired the financial profile: EUR 1.39 billion of cash, roughly EUR 1.9 billion of available liquidity, net debt down to EUR 883 million excluding the Kulim 2 sale-and-leaseback (EUR 1.34 billion including it). Q3 free cash flow of EUR 462 million looks spectacular until you notice operating cash flow was negative EUR 61 million — the quarter was financed by EUR 569 million of divestment proceeds. Two threads have gone conspicuously soft. Kulim 2, a top-ten keyword as recently as early 2026, now surfaces only because Robert Sanders asked, and the answer is unchanged: nothing really changed. And free cash flow has been quietly pushed out a year — in August the guidance was a burn “a bit more than EUR 300 million” — Rainer Irle, CFO · 2026-08-04 for 2026 ex-divestments; today the commitment is merely "positive free cash flow" for fiscal 2027. The Simplify program, which spiked as a headline theme earlier in the year, now merits one line: more than EUR 20 million of run-rate savings against a EUR 200 million target by 2028. None of that makes this a bad quarter. The core is growing through mixed end markets, the balance sheet is fixed, and design wins of EUR 1.3 billion in the quarter put the company on track for roughly EUR 5 billion for the year. But strip out the divestment gain and the story is a small-cap semiconductor supplier with a solid 3–4% core, a heavily diluted sensor segment, and a photonics option whose commercial clock has not started. The market is being asked to pay for the option today while pricing the architecture it belongs to as if the cycle has already peaked. That is the tension — and it will not be resolved by another demonstrator.