When a company with Nvidia's cash reserves decides to write two billion-dollar cheques in a single morning, it is usually worth asking what problem it is trying to solve. In this case, the answer is light, quite literally.
Nvidia announced on Monday that it will invest $2 billion each in US photonics companies Lumentum and Coherent, a combined $4 billion bet on technology that uses pulses of light rather than electrical signals to move data inside the AI data centres that underpin the modern internet. The chip giant is looking to shore up research pipelines and supply chains to support the major AI infrastructure buildout.
Why light beats electrons
The pitch for photonics is straightforward, even if the engineering is not. Silicon photonics represents a fundamental reimagining of how data moves within computing systems: instead of pushing electrons through copper wires, the technology uses photons (particles of light) to carry information through silicon waveguides that function like nanoscale fibre optic cables, integrated directly onto chips. The practical upside is significant. Replacing electrons with photons can offer, according to estimates from Cambridge Consultants, a tenfold increase in energy efficiency, along with a 10 to 50 times bandwidth improvement over traditional computing.
For Nvidia, the urgency is real. Goldman Sachs projects a 160% increase in data centre power demand by 2030, reaching 945 terawatt-hours annually, equivalent to Japan's entire electricity consumption. Its own co-packaged optics switch systems already claim to deliver a 3.5 times leap in power efficiency compared to previous architectures, and a 10 times improvement in resiliency by reducing the number of overall optical components that may fail. Monday's investments are designed to guarantee Nvidia can actually procure those components at scale.
What the deals actually cover
The tie-ups include multi-billion-dollar purchase commitments from Nvidia and future capacity and access rights to advanced laser and optical networking products from both Lumentum and Coherent. For Lumentum, the agreement goes further: Nvidia is investing $2 billion in Lumentum to support R&D, future capacity and operations as the company builds out its US-based manufacturing capabilities in a new fab. Lumentum CEO Michael Hurlston said the company will invest in a new fabrication facility to increase capacity.
"Together with Lumentum, NVIDIA is advancing the world's most sophisticated silicon photonics to build the next generation of gigawatt-scale AI factories," said Jensen Huang, founder and CEO of Nvidia. Markets responded immediately: shares in Lumentum Holdings and Coherent Corp rose 8% and 6.5% respectively in premarket trading.
Supply chain, not just technology
There is a second rationale at work here beyond pure engineering. The funding is intended to support research and development as well as manufacturing operations within the US, aligning with broader efforts to localise advanced technology supply chains. That framing aligns Nvidia's spending with the political priorities of onshoring critical technology, a consideration that is increasingly hard to separate from pure commercial logic in the semiconductor sector.
The competitive pressure is also intensifying. Custom-chip maker Marvell Technology announced the acquisition of semiconductor startup Celestial AI last year in a $3.25 billion deal to tap into its work on photonics, which uses light rather than electrical signals to create connections between AI chips. The two deals could also help the AI chip giant extend its lead in the fast-moving AI hardware industry as many top cloud providers take to building custom silicon in order to match their personal AI needs. In other words, locking in supply from the best photonics producers is as much a strategic moat as a technology play.
The real limits of the bet
The counterargument to unbridled enthusiasm here is worth taking seriously. Nvidia did not disclose specific timelines or equity structures associated with the investments. The companies did not provide financial forecasts or quantify the expected revenue impact from the investment. Photonics manufacturing is notoriously difficult to scale, and the history of semiconductor supply chains is littered with bold commitments that took longer and cost more than anyone projected.
There is also the broader question of whether concentrating this much photonics capacity in two suppliers, and in one country, creates its own fragility. Diversity of supply matters. A single geopolitical disruption, a factory fire, or a patent dispute could create the exact kind of bottleneck Nvidia is spending $4 billion to avoid.
For Australian companies and research institutions that depend on AI infrastructure, the medium-term implications are mostly positive: faster, more energy-efficient data centres should eventually reduce the cost and carbon footprint of running large AI workloads. The CSIRO and Australia's growing cohort of AI-dependent enterprises will be downstream beneficiaries if Nvidia's gamble pays off. The Australian Bureau of Statistics has consistently recorded rising energy costs as a key operating concern for data-intensive industries, and any technology that materially shifts that equation commands attention.
The hype around photonics is real. But so are the engineering challenges. Nvidia is making a very large wager that it can solve both the physics and the logistics simultaneously. On past form, it would be unwise to bet against them. It would also be unwise to assume the path is straightforward.