Nvidia (NASDAQ: NVDA) is deepening its push into photonics, the light-based technology that could become the next big battleground in artificial intelligence infrastructure.
The company has spent the past two months striking deals with suppliers that make the lasers, optical components and fibre connections needed to move data inside massive AI data centres.
The headline number began with $4 billion of investments in Lumentum and Coherent. But when Nvidia’s new Corning partnership is included, the commitment looks much larger.
The message is clear: Nvidia is not only selling the GPUs that train and run AI models. It wants more control over the pipes that connect them.
From $4B to $7B-plus in two months
Nvidia’s photonics push began on March 2, when it announced separate $2 billion investments in Lumentum and Coherent. Both companies make optical and photonic products used in high-speed data transmission.
The deals were not passive financial bets as Nvidia also secured multi-year purchase commitments and access to future capacity for advanced laser and optical networking products.
That matters because AI systems are no longer limited by chips alone. As models get bigger, the bottleneck increasingly shifts to how quickly thousands of processors can talk to one another.
Then came Corning. On May 6, Nvidia and the 175-year-old glassmaker announced a long-term partnership to expand US production of optical connectivity products.
Corning said it would increase US optical connectivity manufacturing capacity tenfold, boost fibre production capacity by more than 50%, build three new plants in North Carolina and Texas, and create more than 3,000 jobs.
The equity component gives Nvidia the option to build a stake worth roughly $3 billion.
That lifts the scale of the photonics strategy well beyond the $4 billion figure that dominated initial coverage.
Jensen Huang has framed the move as part of the “largest infrastructure buildout” of the AI era. In plain English, Nvidia sees optical technology as essential to keeping AI factories fast, efficient and scalable.
Why light beats copper
Today’s AI data centres still depend heavily on copper connections as copper is familiar, cheap and reliable over short distances.
But it has limits. As data moves faster and systems grow larger, copper creates more heat, consumes more power and becomes harder to manage at scale.
Photonics changes the medium. Instead of pushing electrical signals through metal wires, it uses light to move data through optical components and fibre.
The result is faster transmission, lower signal loss and better energy efficiency across large clusters.
The simplest comparison is a road system. Copper is a crowded two-lane highway, while photonics is a wider expressway that can carry more traffic while burning less fuel.
That is why co-packaged optics, or CPO, has become so important. CPO brings optical engines closer to the networking chip, reducing the need for separate pluggable transceivers and long electrical paths.
Nvidia says its Spectrum-X Ethernet Photonics switches integrate CPO directly onto the switch chip and are designed for million-GPU AI clusters.
This is not only about saving electricity. Faster interconnects can improve how quickly AI systems respond, especially during inference, when users expect near-instant answers.
In that sense, photonics is both an energy story and a speed story.
The post Nvidia is quietly betting $7B on this emerging tech for next phase of AI growth appeared first on Invezz
