Corning reported before Tuesday’s open and the print was, by any normal standard, excellent. Core sales rose 17% to $4.74 billion against roughly $4.63 billion expected. Core EPS rose 30% to $0.78 against $0.76. Optical Communications — the segment that is the entire AI story here — grew sales 32% to $2.07 billion, with Enterprise Networks up 65% and segment net income up 77% to $438 million. On a GAAP basis the company booked $4.51 billion in revenue and $0.64 a share.
The stock fell about 19% to near $116, on pace for its worst single session since October 8, 2002, when it dropped 17.3%.
What did it? The guide. Corning told investors to expect third-quarter core sales of $4.9 billion to $5.0 billion — roughly 16% growth — and core EPS of $0.85 to $0.89, about 28% growth. FactSet consensus for revenue was $5.0 billion. That is not a miss. That is a company promising 16% growth and topping out at precisely what analysts already had in the model. In this tape, that is the same thing.
Two words on the call did the damage
Underneath the numbers, management flagged a sharper-than-expected deceleration in carrier network spending as telecom operators defer fiber deployments — the boring, non-AI half of the optical business that was supposed to carry the base while data centers carried the upside. Then they flagged rising memory prices as a headwind for the consumer electronics market through 2026, which matters because Corning sells cover glass into phones and memory costs are already pushing handset prices up. One AI-adjacent business decelerating, one consumer business getting squeezed by AI’s own supply chain.
The rest of the optical complex went with it. Ciena fell 5.9%, Coherent 5.7%, Lumentum 4.7%, with Marvell and AXT down harder. Every one of those names is priced off the same assumption: that data-center fiber demand compounds fast enough to make today’s multiple look cheap in eighteen months.
The context that makes this brutal
Corning hit an all-time high of $271.78 on June 30, capping a run that included record closes after Amazon agreed to buy billions of dollars of optical fiber for its US data centers. Near $116, the stock is roughly 57% below that high — in four weeks, and with a beat in hand. Management’s Springboard plan still targets a $20 billion annualized sales run rate by the end of this year, $30 billion by end-2028 and $40 billion by end-2030. Nobody sold that today. They sold the slope.
Our take: This is the clearest evidence yet that the AI supply chain has stopped being priced on growth and started being priced on acceleration. Corning grew 17%, guided to 16%, and lost a fifth of its value — because 16 is smaller than 17. When a market pays for the second derivative, every deceleration is a downgrade regardless of profit, and a beat buys you nothing. That is a much more fragile setup than a high multiple, because a company can control its earnings and cannot control the shape of its own curve forever. The AI infrastructure names reporting from here don’t need to miss to get hit. They just need to stop speeding up.
What to watch
- Hyperscaler capex guidance this week. Corning’s Amazon and Nvidia agreements came with customer prepayments. If the buyers slow, the prepayments are the first thing to renegotiate.
- Whether carrier fiber deferrals show up elsewhere. If Ciena and Lumentum confirm the same telecom pause, this is an industry datapoint, not a Corning one.
- The memory pass-through. Corning is now the second major company this week to name memory pricing as a consumer headwind. Watch the phone makers.
- The Fed on Wednesday. A rotation out of AI infrastructure behaves very differently depending on what the front end does the next morning.