Journalist |
Nvidia is urgently hunting for AI base station suppliers in China to develop 6G AI-RAN base stations—equipment that not only handles communications but also powers AI computations on site.
On August 4, we exclusively learned from a key insider in the telecom supply chain that Nvidia is accelerating its push into the telecom carrier market, actively seeking Chinese base station manufacturing partners to build 6G base stations tailored for overseas markets.
This source described Nvidia’s quest as looking for the “next Zhongji Innolight” in edge computing—because Nvidia believes that as computing power shifts from data centers to countless AI devices and end users, it will ultimately rely on wireless transmission through base stations.
The insider added that Nvidia’s timeline for this project is “fairly tight,” with hopes of entering trial networks by 2027 or 2028.
We also exclusively confirmed that Shenzhen-based Jiaxian Communication is one of the base station partners Nvidia is building relationships with. An insider at Jiaxian confirmed to us that Nvidia first reached out back in 2025, and the two companies are now co-developing 6G AI-RAN base stations based on Nvidia’s CUDA ecosystem. The collaboration has been running for over six months already.
“Jiaxian Communication and Nvidia have already set up a dedicated joint working group. Nvidia has over a dozen people involved, spread across Singapore, Hong Kong, and mainland China. Right now, the group is focused on researching the open-source ecosystem built around Nvidia’s CUDA system,” a Jiaxian representative told us.
Founded in 2011 and headquartered in Shenzhen, Jiaxian Communication is an AI-RAN wireless communications product and solution provider. On August 4, Tongyu Communication announced a 300 million yuan acquisition of roughly 25% of Jiaxian’s equity.
We’ve learned that the Jiaxian-Nvidia partnership has already moved into the technical R&D phase, and the possibility of setting up a joint commercial entity for overseas expansion hasn’t been ruled out.
“On the technical front, we’ve already started working together,” the Jiaxian representative said. In this collaboration, Nvidia doesn’t directly manufacture or sell complete base stations. Instead, it participates through its chips and technology ecosystem—meaning the base station maker purchases Nvidia GPUs and builds 6G AI-RAN base stations on the CUDA ecosystem, then sells the finished products to overseas markets.
“We expect the AI base station prototype developed with Nvidia to be ready by the end of this year. After validation, it should hit the market within a year or two,” the source added.
For Nvidia, AI base stations represent a business opportunity just as lucrative as AI data centers.
In the AI 1.0 phase, computing power was concentrated in data centers, and Nvidia sold massive volumes of GPUs. Now that we’re entering the AI 2.0 phase of application deployment, edge computing offers even broader potential. Nvidia wants its chips used for edge and on-device computing too, so positioning itself in AI base stations is really about preparing for the coming wave of edge computing.
In other words, a chip company can supply GPUs and model inference capabilities, but to push computing power out to the network edge and deliver it directly to terminals and users, you still need base stations and carrier networks to carry the load.
Nvidia’s ambitions in the communications space are urgent and unmistakable.
On October 28, 2025, Nvidia announced a $1 billion investment in new Nokia shares, securing roughly 2.9% ownership. As part of the deal, Nokia will develop commercial AI-RAN products based on Nvidia’s platform and integrate them into its existing radio access network product lines, enabling carriers to deploy AI-native 5G-Advanced and 6G-ready networks.
In February of this year, Nvidia joined forces with Nokia, Ericsson, T-Mobile, Deutsche Telekom, SoftBank, and others to push for building 6G on an open, software-defined, AI-native platform. Then in March, Nvidia, T-Mobile, and Nokia unveiled AI-RAN testing progress, experimenting with base stations and mobile switching centers that deliver 5G connectivity while simultaneously running edge AI tasks.
Industry analysts point out that AI infrastructure will gradually expand from centralized computing hubs into distributed computing networks made up of base stations, edge facilities, and local nodes. This is especially relevant overseas, where some markets lack the conditions to build massive centralized data centers—making distributed computing a vital complement.
“Apart from China and the U.S., virtually no other country has the capability to build large-scale data centers. Terminal and edge computing will be where global demand truly explodes,” the source noted.
We also learned that Jiaxian Communication plans to unveil a 6G AI-RAN prototype built on Nvidia’s CUDA ecosystem by the end of this year. If the product moves into large-scale deployment, the company’s total sales volume across the 6G era could reach one million units. At roughly 200,000 yuan per unit, that translates to a market opportunity of about 200 billion yuan. Analyst firm Omdia projects that the AI-RAN market will cumulatively exceed $200 billion by 2030.
Nvidia’s decision to source AI base station suppliers in China isn’t accidental—there are very practical reasons behind it.
The global base station supply chain, much like the optical module industry, is already heavily concentrated in China. According to commercial research firm IndexBox, China accounts for an estimated 60% to 70% of global 5G base station hardware manufacturing in 2026, with major production clusters in Shenzhen, Dongguan, Shanghai, and Chengdu.
Having chips and a software ecosystem doesn’t mean Nvidia can independently develop base stations on its own. Base stations also involve RF, microwave, antennas, AAUs, complete unit manufacturing, production testing, and carrier network adaptation—all of which depend heavily on decades of accumulated engineering expertise in the communications industry.
We spoke with telecom supply chain insiders who emphasized that the base station industry is highly specialized. Once an industrial cluster forms, it’s extremely difficult for other regions to cultivate the same talent from scratch. Additionally, training engineers in this field takes a very long time. The base station industry’s talent pool—especially hardware engineers and production capabilities—is predominantly concentrated in China.
“That’s why Nvidia has no choice but to find manufacturers in China—it’s the only way to get the cheapest, most cost-effective, and best-quality base stations,” the supply chain source said. Even overseas equipment giants like Nokia and Ericsson have their production capacity based in China, with companies like Luxshare Precision and Foxconn handling contract manufacturing.
In his view, the base station industry has essentially disappeared overseas—foreign markets can no longer handle full-scale production, assembly, testing, and hardware R&D, all of which remain concentrated in China. “This industry may have a huge market and high output value, but it’s deeply rooted in China. Other electronics industries might find alternatives overseas, but the base station industry is very hard to replicate elsewhere.”
A Jiaxian Communication representative told us that because major domestic telecom equipment players like Huawei and ZTE have their own proprietary chips and technology ecosystems—and given the broader China-U.S. tech environment—Nvidia’s options for independent base station partners in China are quite limited.
We also gathered from carrier channels that telecom operators are accelerating AI base station adoption. Current 5G pricing is relatively low, but if operators can charge “token” fees, prices could rise substantially.
In May 2026, China Telecom launched token-based plans, while China Mobile has been experimenting with combining data traffic with large model tokens. China Telecom’s Ningxia branch has a “Token Factory” centralized procurement valued at approximately 17.4 billion yuan. Carriers are pivoting from selling voice and data to selling model invocation and inference services. Core telecom supply chain analysts also predict that 6G base stations will kick off another round of trillion-yuan communications infrastructure investment.
As the large models trained in computing centers continue to mature, AI edge and distributed computing infrastructure is rapidly moving up the priority list.
Just as QualcommPresident and CEO Cristiano Amon predicted earlier this year: “Whoever wins edge AI is more likely to win the entire AI era.”
