Beyond Chips: The Hidden Battle Over Connectivity

In the fourth quarter of 2026, most of the electronics industry's attention is focused on computing-power chips and memory chips. Profit data also confirms the heat of these two lines: in the first three quarters, electronics industry profits grew 1.0-fold year on year, with profits in the integrated circuit industry growing especially prominently, contributing more than 80% of the electronics industry's profit growth. The numbers are impressive enough, and the story is clear enough. But if you shift your gaze away from computing-power and memory chips, you will notice another thread emerging. It does not directly produce computing power, yet it determines whether computing power can be "fed"; it does not manufacture memory, yet it affects whether data can be moved in time. This thread is called "connectivity."

Optical interconnect: the "internal blood vessels" of computing clusters

At the end of October, a company founded only two years ago completed two consecutive financing rounds, totaling several hundred million yuan. Lanqin Micro, a Tsinghua-affiliated optical interconnect chip company, focuses on interconnection within AI computing clusters, with products covering OIO chips and NPO and CPO solutions. It had previously launched a micro-ring optical I/O chip compatible with the OCI standard and featuring 3.2 Tb/s per-chip bandwidth.

A startup's financing news would not normally attract much attention. But viewed together with another piece of information, its significance becomes different. At GTC Berlin in October, Nvidia announced cooperation with six manufacturing partners to expand capacity for substrates, advanced packaging, testing, and optical packaging. Oracle, meanwhile, announced that it would expand deployment of Vera Rubin servers to several gigawatts and would also adopt self-developed network chips.

The larger the computing cluster, the more fatal the data transmission bottleneck between chips becomes. As the marginal improvement in single-chip performance narrows, the bottleneck in system-level efficiency shifts from "how fast it computes" to "how fast it transmits." Optical interconnect solves precisely this problem: using optical signals instead of electrical signals to transmit data between chips and racks, with higher bandwidth and lower power consumption. Lanqin Micro's financing and Nvidia's expansion of optical packaging capacity point to the same judgment: the next round of competition in computing infrastructure is not in the chips themselves, but between chips.

The "long-term lock-in" of memory: another form of connectivity

News of memory price increases has continued for more than half a year, but in October there was a noteworthy change: Chinese cloud vendors began seeking to sign ten-year long-term DRAM contracts.

According to South Korean media reports, Chinese cloud service providers such as Alibaba, Tencent, and ByteDance are in talks with Samsung Electronics and SK Hynix over long-term supply agreements for server DDR and LPDDR, with some customers hoping to extend the term to 2035, far beyond the industry's usual three to five years. At the same time, SK Hynix said the memory shortage will last until 2030, and the company has signed long-term agreements with 14 major customers, locking in orders exceeding half of its total capacity for 2027 to 2028.

This is no longer the logic of short-term stockpiling. The essence of a ten-year long-term contract is that downstream customers are using contracts to lock in future supply certainty. When memory changes from "a standard product that can be bought at any time" into "a scarce resource that requires planning ten years in advance," the power relationships in the industrial chain are shifting. In the past, memory manufacturers chased customers; now, customers chase capacity. For small and medium-sized enterprises not on the long-term contract list, this means that in the coming years, memory procurement costs and supply stability will increasingly depend on the depth of your relationships with upstream players rather than on market prices.

The "connectivity" of consumer electronics is also changing

The story of connectivity is not only in data centers. In late October, Huawei released the Mate 90 series, and one technology in it is worth noting: multi-device mutual-assistance communication sharing. Combined with NearLink technology, connectivity among phones, PCs, and cockpits is moving from "paired connection" to "seamless collaboration."

A more macro signal came from the "Top Ten Trends in Consumer Electronics" released at the CEIC2026 Consumer Electronics Innovation Conference. "From devices to agents, connectivity technology is accelerating its generational shift" was listed as one of the core trends. The judgment of Wu Hequan, an academician of the Chinese Academy of Engineering, was even more direct: in the agent era, consumer electronics devices are no longer isolated hardware, and NearLink will become the "neural network" connecting AI agents.

This means that the competitive dimension of consumer electronics is shifting from standalone performance to "collaborative capability between devices." No matter how good a phone's camera is, if it cannot seamlessly collaborate with your PC, earphones, and car system, the experience will be compromised. Connectivity technology has moved from a supporting role to a key variable in experience differentiation.

What this means for the industry

Value distribution in the electronics industry is shifting from "single-point performance" to "system efficiency." Progress in chip process technology remains important, but its marginal returns are diminishing. In its place, "connectivity" indicators such as interconnect bandwidth, certainty of memory supply, and device collaboration capability are becoming key factors determining the upper limit of system performance.

For small and medium-sized enterprises in the industrial chain, this trend offers a pragmatic angle for thinking: does your product and technology improve the system's "connectivity efficiency," or does it increase the system's "connectivity cost"? Demand for the former will grow as the scale of computing power expands, while the latter may be gradually replaced or squeezed. Optical interconnect, long-term supply agreements, and cross-device collaboration are essentially different manifestations of the same direction.