MaxLinear Doubled Its Optical Guidance and Lost Half Its Market Cap

The Numbers Say One Thing

MaxLinear's third quarter was not an incremental beat. Revenue came in at $218.1 million, up 29% sequentially from $168.8 million and 72% year over year from $126.5 million. Non-GAAP gross margin hit 60% and non-GAAP operating margin reached 29.5% — a company that was guiding to break even eighteen months ago is now printing the operating leverage its long-term model always promised. The segment mix is the real headline: infrastructure revenue of roughly $125 million grew 210% year over year and, in CEO Kishore Seendripu's words, is “now larger than the rest of our revenue categories combined” — Kishore Seendripu · 2026-10-22 — versus broadband at ~$53 million, connectivity at ~$25 million and industrial/multimarket at just ~$15 million.

The company's own fundamentals file still shows the prior quarter (filed 2026-07-23), and it is worth holding both pictures side by side. Then, total revenue was $169 million with a −3.9% operating margin and a negative $25 million in free cash flow. Q3 flipped revenue and margin; the cash line did not flip nearly as cleanly, and that is the thread the bears are pulling.

What Actually Changed This Quarter

The most consequential new fact was not in the prepared remarks — it surfaced under questioning. Tore Svanberg of Stifel caught the shift directly: “I think you said you expect 1.3 revenue already first half of next year. I think in the last quarter, I think you talked about maybe middle of the year. So what changed there?” — Tore Svanberg · 2026-10-22 The answer confirmed a genuine pull-forward: “there is an initial design win that will start shipping in the first half” — Kishore Seendripu · 2026-10-22. Go back one call and the language was softer — in July, management expected “the ramps at 1 or 2 opportunities to start somewhere in the second half of the year” — Kishore Seendripu, Chief Executive Officer · 2026-07-23. Go back three calls, to October 2025, and Rushmore was a hedging exercise: “Rushmore, yes, best case will be the end of '26, but I'm not -- I'm being realistic. And -- but Keystone, Keystone, Keystone.” — Kishore Seendripu, CEO · 2025-10-23 The 1.6-terabit gigabit per lane generation has migrated from aspirational to scheduled in nine months. That is new, and it is company-specific.

Second, the content math changed. The 800G-to-1.6T step was always going to be a roughly 2x ASP event; the addition of the Washington TIA on top makes it more than that, because MaxLinear now sells two chips per module instead of one. The attach rate story is the sleeper: management puts current TIA attach near 80%, heading to 100% within about eighteen months. Third, and easier to miss, is Rack Commander — a USB/PCIe/UART bridge family aimed at telemetry, power management and protection inside rack-scale systems. This is MaxLinear pushing from data movement into the control plane of the AI data center, a market it has never addressed. Add Anapurna's electrical interconnect ambitions (management is targeting 20–30% of the 1.6T AEC cycle, deliberately confined to one or two US data centers) and the portfolio breadth claim stops sounding like a slide.

On a snapshot basis in 2030, you should expect 60% to 70% pluggable, 20% NPO, and 10% all various forms of it. That is my personal view. Not a market view.

Note what that implies: the optical pluggable franchise — which the debate has spent two years writing off in favor of co-packaged optics — is the CEO's own base case for two-thirds of the market four years out. The co package optics basket in the global tape is up 72% over 360 days but down 5% over 90, which tells you the market has already begun fading that narrative before the technology ships.

What Quietly Left the Script

For four straight quarters, DOCSIS 4.0 was a load-bearing keyword for this company — a promised 50% content increase per gateway, an upgrade cycle management repeatedly dated to late 2026 and into 2027. On this call it barely exists. Broadband is now described as seasonally down in Q4 with growth resuming in 2027, full stop. The franchise that drove the entire 2020–2021 peak is now a rounding error next to infrastructure, and management has effectively stopped defending it. Industrial and multimarket was reduced to a single sentence and $15 million.

The other omission is structural: once again, no 10% customer. Svanberg flagged it, and it is now a multi-quarter pattern rather than a quirk — a $218 million quarterly revenue base spread across module makers and hyperscalers with no single dominant relationship. That is either diversification or an absence of anchor design wins, and management's own answer to the market-share question invites the second reading: “It should be in the teens for sure... in a 12- to 24-month window, we could be neck and neck with 1 of our 2 competitors.” — Kishore Seendripu · 2026-10-22 That is a goal, not a forecast.

What did not change is the supply constraint, which has hardened into a permanent feature of the growth cycle framing:

The optic supply is easing. We are really constrained by foundries on the DSP supply... there is really, really a fab capacity shortage.

That constraint is the reason the balance sheet moved. Cash fell to $109.9 million, operating cash flow was negative $13.6 million, and the revolver was expanded by $70 million to $200 million — all in service of wafer prepayments for backlog that now extends into 2028. CFO Steven Litchfield was explicit that “I expect prepaids to continue to go up from here” — Steven G. Litchfield · 2026-10-22. This is a company borrowing capacity to buy its own growth, with effective net cash of negative $31 million replacing the $121 million net cash position of 2023.

The Disconnect

Here is the tension worth owning. MaxLinear's stock peaked at $128.03 on June 30 and now sits roughly 49% below that, down 22.6% over the last 90 trading days — even as the optical data server outlook went from $110–130 million in January, to $150–170 million in April, to a $50 million-plus raise in July, to $250–270 million today. Roughly a doubling of guidance in nine months, against a halving of the share price.

The explanation is not company-specific. The AI data center theme in the global tape is up 52% over 360 days with 41 advancers against 13 decliners — and down 5.6% over 90 days with 6 advancers against 48 decliners. The whole complex de-rated while MaxLinear's own AI data center exposure inflected. The company is riding a wave that the tape has already decided to stop paying for at the sector level.

Which makes the valuation argument the live one. The fundamentals file prints price to revenue of 20.4x, which looks absurd until you note it is calculated on trailing revenue of roughly $560 million. Simple arithmetic on the Q3 run rate — $218 million annualized — implies something closer to 7x on a $5.9 billion market cap. The multiple debate is entirely a bet on whether $250 million of optical revenue becomes $1 billion, and management is funding that bet with R&D at 33% of revenue, a level that leaves no room for a demand air pocket.

The bull case is that DSP supply is scarce, module makers are queuing, and this remains early innings — a phrase management has now used across multiple quarters. The bear case is that the easy re-rating is done, the cash is being pre-spent, and the sector's multiple has already voted. This quarter added real evidence to the first case. It did not add any to the second.