AI capital expenditure “spills over” from GPUs to the entire industry chain! Corning (GLW.US) and Qualcomm (QCOM.US) won big orders on the same day, and confidence in data center spending was strongly endorsed

The Zhitong Finance App learned that as the boom in AI capital spending spreads from Nvidia's GPUs to the wider industry chain, two very different companies — glass manufacturer Corning (GLW.US) and mobile chip giant Qualcomm (QCOM.US) — announced two heavyweight deals on the same day that are redefining the boundaries of AI infrastructure investment.

On Tuesday (September 8), Corning announced that it has signed a multi-billion dollar fiber supply agreement with Verizon that will last until 2032; Qualcomm revealed that it has issued warrants worth 4 billion US dollars to Amazon as part of the AWS server chip purchase agreement of up to 60 billion US dollars. Together, the two deals send a clear signal: the capital expenditure of AI data centers is being completely spilled from the “chip layer” to the “network transmission layer,” and the list of beneficiaries is rapidly expanding.

Corning: From glass manufacturer to AI optical networking “key sir”

Corning's agreement with Verizon covers more than 80 million miles of high-density fiber and connectivity solutions from 2027 to 2032. Verizon will use this order to simultaneously advance two major strategies: one is to expand home fiber-optic broadband coverage to 40 million to 50 million access points; the other is to build a long-distance backbone network connected to AI hyperscale data centers. Verizon Business CEO Kyle Malady said the company is deploying a “converged architecture” to unify mobile connectivity and broadband into a seamless experience while building the high-capacity, low-latency backbone required by AI hyperscale customers.

This is the fourth major AI infrastructure order that Corning has recently locked in. Previously, Corning had reached a multi-year fiber supply agreement with Meta of up to 6 billion US dollars; cooperated with Nvidia to expand US optical connectivity manufacturing capacity, and received a commitment from the latter to invest up to 3.2 billion US dollars to build three new optical fiber manufacturing plants; and signed a multi-billion dollar data center fiber supply agreement with Amazon. Corning CEO Wendell Weeks previously revealed that the deal size with two unnamed hyperscale customers was even “larger” than Meta's $6 billion agreement.

Behind these orders is a key piece of data: AI data centers require about 16 times the amount of fiber required for a single node than traditional switches. From cloud vendors to chip companies to today's leading telecom operators, AI's demand for optical fiber is forming an unprecedented structural growth cycle. Boosted by this, Corning's stock price rose 8% on Tuesday, with a cumulative increase of 90% during the year.

Qualcomm: A “critical leap forward” from mobile phone chips to AWS custom chips

If Corning's story is “AI demand spreads from chips to the periphery,” then Qualcomm's story is “a mobile phone chip company trying to tear apart a hole in an AI data center.”

According to regulatory documents submitted by Qualcomm, the company has issued warrants to Amazon, allowing AWS to purchase up to 25 million Qualcomm shares at a price of $161.26 per share, with a potential shareholding value of approximately US$4 billion. Warrants will be vested in batches based on the amount of orders placed by Amazon against Qualcomm, and the total amount of related orders can reach up to 60 billion US dollars. The deal is part of a “multi-generation” custom chip partnership between Qualcomm and Amazon. The two sides will jointly develop customized chips for AI inference and extend the joint development rate to 1.6T and higher optical interconnection solutions.

Qualcomm Chief Financial Officer Akash Palkhiwala confirmed at the Goldman Sachs conference on Tuesday that revenue from manufacturing chips for Amazon will be realized starting in the December quarter and will be a “core component” for the company to achieve its $15 billion data center revenue target for fiscal year 2029. At the same time, he revealed that Qualcomm “is promoting cooperation with another data center customer in a similar way” — implying that following Meta, Qualcomm's data center customer list is still expanding.

In June of this year, Qualcomm released the Dragonfly C1000 processor for data centers, focusing on agent-based AI workloads. Bank of America predicts that the global CPU market may more than double from US$27 billion in 2025 to reach US$60 billion by 2030. Qualcomm's transformation is betting on the macro trend of repricing CPU demand in the AI reasoning era.

The “spillover effect” of AI infrastructure investment: from Nvidia to the entire technology industry chain

It was no accident that both deals were announced on the same day. Together, they point to an accelerating structural trend: the beneficiaries of AI data center spending are spreading across the board from GPU manufacturers to the broader infrastructure industry chain.

After the market on Tuesday, Intel (INTC.US) and AMD (AMD.US) rose 9% and 6% respectively, HPE Technology (HPE.US) rose 8%, and photonics company Coherent (COHR.US) rose 7%. Since this year, the market capitalization of both HPE Technology and AMD has doubled, and the market value of Intel has almost tripled. Goldman Sachs raised expectations for the optical module industry on the same day, and Coherent closed up 7.24%; Deutsche Bank covered the AI hardware sector for the first time, listing Coherent and Lumentum as the “highest confidence” AI hardware targets.

The macro background of this diffusion trend is that AI capital expenditure is moving from “whether you want to invest” to the stage of “invest as much as you can”. The capital expenditure of the four major US CSPs (Microsoft, Google, Amazon, Meta) increased 86% year-on-year in the second quarter, and is expected to exceed 886.7 billion US dollars in total by 2026. TrendForce has raised the total capital expenditure estimates of the five major US CSP companies in Canada and China to about US$830 billion.

Worries Underneath Prosperity: 71% of Americans Oppose It

However, the fanatical expansion of AI infrastructure is not without costs. According to Gallup's May 2026 poll, 71% of Americans oppose building data centers in their communities. This social resistance has directly translated into financial risk — it is reported that Anthropic is preparing to list “the public's negative perception of AI and data centers” as a risk factor in its upcoming IPO prospectus.

At the same time, the continued rise in global bond yields is putting substantial pressure on the financing costs of AI capital expenditure. The 10-year US Treasury yield has risen to around 4.81%, and the 30-year term is approaching 5.25%, which means that future AI infrastructure debt financing costs are rising systematically.

Summarize

The two deals announced by Corning and Qualcomm on the same day provided the market with the latest evidence that AI infrastructure investment is still spreading at an accelerated pace. From glass optical fibers to custom chips, from telecom operators to cloud giants, the chain of benefits from AI capital expenditure is expanding at an unprecedented speed. As Corning CEO said, the 175-year-old glass company is “leading the AI revolution at the speed of light.” However, 71% of public opposition and rising financing costs are also reminding the market that this feast of AI infrastructure is not without a ceiling.