🔥Photonics concept stocks collectively surged, as the bottleneck of AI computing power is shifting from chips to "connectivity".

One of the most noteworthy changes this year is not just that $NVIDIA(NVDA.US) continues to be strong, but that the market is beginning to reprice another major theme in AI infrastructure: optical communication, optical modules, photonics, testing equipment, and high-speed interconnects.

This set of gains already shows what capital is looking at:

$Aehr Test(AEHR.US) +373%

$Applied Optoelectronics(AAOI.US) +295%

$Lumentum(LITE.US) +130%

$Ciena(CIEN.US) +121%

$Marvell Tech(MRVL.US) +95%

$Corning(GLW.US) +94%

$Coherent Corp.(COHR.US) +83%

$Macom Tech(MTSI.US) +66%

$Fabrinet(FN.US) +51%

$Credo Tech(CRDO.US) +29%

$Broadcom(AVGO.US) +21%

$Astera Labs(ALAB.US) +19%

In the past, when discussing AI, the focus was almost entirely on GPUs, HBM, servers, and data center power. But as model sizes continue to expand, the real bottleneck is starting to move to the network layer, transmission layer, and packaging layer.

An AI data center doesn't work on a single server; it's the constant exchange of data between thousands of chips. The stronger the computing power, the greater the pressure on data transmission; the larger the training cluster, the more critical low-latency, high-bandwidth, low-power connections become.

This is also why photonics-related companies are starting to be re-evaluated.

$Applied Optoelectronics(AAOI.US)、$Lumentum(LITE.US)、$Coherent Corp.(COHR.US) benefit more directly from the demand for optical communication and high-speed optical modules.

$Ciena(CIEN.US) represents network infrastructure upgrades.

$Marvell Tech(MRVL.US)、$Broadcom(AVGO.US)、$Credo Tech(CRDO.US)、$Astera Labs(ALAB.US) are closer to high-speed interconnects, SerDes, switching, and AI data center connectivity chips.

$Corning(GLW.US)、$Fabrinet(FN.US)、$Macom Tech(MTSI.US)、$Aehr Test(AEHR.US) correspond to materials, manufacturing, RF/optical communication devices, and testing segments, respectively.

This round of gains is not a single company story, but a valuation diffusion across the entire AI infrastructure chain.

Stage one, the market bought GPUs.

Stage two, the market bought servers, power, and cooling.

Stage three, the market is starting to buy high-speed networks, optical modules, photonics, and data transmission bottlenecks.

What really needs to be considered is:

If AI computing power continues to expand, the most scarce asset in the next round may not be "who can compute faster," but "who can make data transmit faster, more stable, and more power-efficient."

Are you more optimistic about the continued diffusion of the photonics theme, or do you think this round of gains has already front-loaded expectations?