They grew revenue 72% and generated negative $13.6M in operating cash flow. The CFO expanded the revolver $70M specifically to fund wafer prepayments for backlog he says runs through 2028.
Thesis: MXL sits on the constrained side of the industry's real bottleneck: foundry DSP wafer capacity. Seendripu stated plainly that fab capacity is not coming online, lead times are 26-28 weeks with orders beyond that window, and DSP ASPs are doubling from 800G to 1.6T with TIA stacking content for >2x total. The Street models 5-7% share; management targets teens. The cash flow gap is real but is a lagging indicator of prepayments against backlog the company can already see. The structural pricing power is the edge, not the narrative.
Verdict: LONG — Conviction: MEDIUM
Catalyst: First half 2027 Rushmore 1.6T production revenue, confirmed through an initial design win and accelerated one full quarter from mid-2027. Washington TIA and Annapurna revenue starting 2027, volume 2028.
Key Risk: Debt-funded wafer prepayments against cancellable orders. If hyperscaler optical CapEx pauses one quarter, MXL absorbs prepayments on the balance sheet while revolver balances the debt. Infrastructure is 57% of revenue with NVIDIA and Google named as the primary US ramp customers.
The Tell: Kishore, answering a double-ordering question, said: "nobody's signing real long-term agreements." He intended to prove supply discipline. He also just told you the multiyear backlog through 2028 rests on cancellable POs, not commitments, while his CFO expands the revolver to $200M to fund wafer prepayments against that same backlog.
Friction Level: MODERATE_FRICTION — Both sides agree on the facts: debt-funded wafer prepayments, foundry DSP bottleneck, hyperscaler concentration. Bull reads prepayments as Capital Conviction on visible demand; bear reads same action as Backlog Quality risk against cancellable orders. Same $50M prepaid asset increase, opposite reads.
2026Q3 · Oct 22, 2026
5 of 5 readings agreed5 LONGRead the full call
72% Revenue Growth, Negative Operating Cash Flow
MXL sits on the constrained side of the industry's real bottleneck: foundry DSP wafer capacity. Seendripu stated plainly that fab capacity is not coming online, lead times are 26-28 weeks with orders beyond that window, and DSP ASPs are doubling from 800G to 1.6T with TIA stacking content for >2x total. The Street models 5-7% share; management targets teens. The cash flow gap is real but is a lagging indicator of prepayments against backlog the company can already see. The structural pricing power is the edge, not the narrative.
Kishore, answering a double-ordering question, said: "nobody's signing real long-term agreements." He intended to prove supply discipline. He also just told you the multiyear backlog through 2028 rests on cancellable POs, not commitments, while his CFO expands the revolver to $200M to fund wafer prepayments against that same backlog.
Revenue beat at $218.1M (29% QoQ, 72% YoY), gross margin hit 60% against 58% prior, raised optical data center outlook to $250-270M, Q4 guide of $240-260M midpoint up 16% QoQ despite broadband and connectivity declining seasonally.
Seendripu stated DSP ASPs are "more or less doubling" going 800G to 1.6T, with TIA content stacking on top for more than 2x total ASP expansion. First time in history the speed-transition price increase is sticking, per his own words.
CFO: "booking products throughout 2027 into 2028" with 26-28 week lead times and customers ordering beyond that window to secure supply. Raised full-year optical outlook with bookings visibility extending through 2027 into 2028.
$50M sequential increase in prepaid wafer assets, revolver expanded $70M to $200M explicitly to fund prepayments for backlog through 2027. CFO expects prepayments to continue rising with revenue. Management is spending liquidity to lock supply.
Seendripu directly addressed double ordering ("I don't think there's any double ordering"), named the binding constraint ("we're really constrained by foundries on the DSP supply"), quantified TIA attach at 100%, and sized the AEC opportunity at 20-30% share.
Seendripu admitted "nobody's signing real long-term agreements" in the same answer where he confirms no double ordering. Multiyear visibility rests on cancellable purchase orders funded by debt-financed prepayments, not firm commitments.
CEO: "fab capacity is not coming online" and "we're really constrained by foundries on the DSP supply." MXL is fabless. TSMC and Samsung own their cost structure. Gross margin expansion to 60% depends on mix holding while foundry pricing is outside their control.
Keystone (800G) is the only product in volume production. Rushmore, Big Sky, Annapurna, Washington, RackCommander, and Panther are all pre-revenue or early for 2027-2028. Seendripu: "we're constrained in how many we can activate at the same time" in qualification cycles.
Zero hedging language across the entire call. "Extreme satisfaction and joy," "neck and neck with one of our two competitors," share "in the teens for sure." No floor language on any forward statement.
Days of inventory dropped from 123 to 111, presented as goods shipping faster than building. Combined with foundry capacity that is not expanding, this is a utilization ceiling: any demand acceleration hits supply they cannot quickly add.
+85
CONFIDENT"Seendripu volunteered a share target nobody asked for ("in the teens for sure"), accelerated 1.6T timeline unprompted, and attached hard numbers to every question: 100% TIA attach rate, 20-30% AEC share in one or two data centers, ASP doubling quantified. No hedging language anywhere on the call."
Specific numbers throughout: revenue by segment, gross margin bridge, lead times, ASP multiples, share trajectory, cash flow detail. Management directly addressed double ordering and supply constraints. Gap: no segment-level profitability, no signed backlog detail.
First half 2027 Rushmore 1.6T production revenue, confirmed through an initial design win and accelerated one full quarter from mid-2027. Washington TIA and Annapurna revenue starting 2027, volume 2028.
Debt-funded wafer prepayments against cancellable orders. If hyperscaler optical CapEx pauses one quarter, MXL absorbs prepayments on the balance sheet while revolver balances the debt. Infrastructure is 57% of revenue with NVIDIA and Google named as the primary US ramp customers.
The verdict is public. The case behind it is Pro: the forensic reading of guidance and language, the bull and bear cases, the red flags, and the play. It opens to everyone on Jan 20, 2027.