Nvidia (NASDAQ:NVDA) just wrote two very large checks. It will invest $2 billion each in Lumentum and Coherent, plus make multiyear, multibillion-dollar purchase commitments.
This is not a “sprinkle some venture dust” moment. It is a supply-and-physics moment. Nvidia wants optical interconnects to scale AI clusters. It also wants co-packaged optics, where light engines sit right next to switching silicon.
The logic is plain: secure capacity, cut supply risk, and move more bits with less power.
Networking already matters more in AI than most people admit. Nvidia’s own results show networking is becoming a much bigger revenue line.
The long game is about data-center economics, U.S. photonics scale-up, and who controls the next bottleneck.
Capacity Security & Supply-Chain De-Risking Through Owning The Bottleneck
If Nvidia were the ACQUIRER and Lumentum or Coherent were the TARGET, the biggest synergy is simple: certainty. These deals already include capacity rights and purchase commitments.
An acquisition would push that further. It could turn “preferred access” into “planned access.” That matters when AI builds are measured in racks, not servers. It also matters when a missing laser can idle a whole network shelf.
There is a second angle. It is bargaining power across the stack. Nvidia sells GPUs, systems, and networking. It also coordinates with foundries and packaging partners. An optical components unit inside that machine could smooth shocks. Think fewer emergency spot buys. Think more predictable qualification cycles. It could also reduce duplicate engineering work. The same team could tune optics for Nvidia’s switch road map.
There is a third angle that investors often miss. Supply risk is now a competitive weapon. Nvidia’s CFO has talked about inventory and purchase commitments extending well out.
Optics becomes another place to “pull forward” capacity. If rivals face optical shortages, their clusters scale slower. That does not require sabotage. It is just math. In that world, owning a key optics supplier can be a quiet moat. It can also keep pricing steadier over time.
Bandwidth-Per-Watt Wins That Reshape AI Networking Economics
AI data centers hit power walls before they hit floor-space walls. That makes “bits moved per watt” the new obsession. Optical links help because light moves data with lower loss over distance than copper.
Now add co-packaged optics. You shorten the electrical path. You reduce heat and signal problems. Nvidia has already pushed this direction in its AI networking line.
So what is the acquisition synergy? It is co-design. Nvidia likes to own hard problems and then ship them at scale. Optics is full of hard problems. Lasers, alignment, reliability, yield, and test all bite. A combined company could iterate faster. It could also align product cycles. Switch silicon and optical engines would launch in tighter sync. That reduces integration friction for customers.
The long-term economic effect is subtle but big. If optics improves network utilization, you waste fewer GPU cycles waiting on data. That changes effective cost per training hour. It also changes cost per inference token. Those savings are not only “hardware savings.” They are “time savings.” And in AI, time is a real expense. Better networking can mean fewer GPUs needed for the same throughput. That shifts budgets away from “more boxes” and toward “better boxes.”
U.S. Photonics Manufacturing Scale-Up As A Strategic Asset
Both partnerships explicitly point toward U.S. manufacturing build-out.
Lumentum referenced a new U.S. fab in Nvidia’s announcement.
That is not just patriotic window dressing. It is about control, resilience, and lead times. Optics supply chains are global. They also depend on specialized tools and materials. When demand spikes, capacity does not appear overnight.
If Nvidia acquired a key supplier, it could help underwrite expansion with its scale and planning discipline. It already does this across its broader supply chain.
That could mean faster ramp of laser and optical component output. It could also mean tighter process control. In photonics, yield matters a lot. Small yield shifts can swing component availability. A deeper integration could improve forecasting and capex timing. It could also reduce “bullwhip” swings.
There is also a policy and customer angle. Sovereign AI and regulated industries care about where key components are made. Nvidia has spoken about sovereign demand as a meaningful growth vector.
U.S.-scaled photonics could become a selling point for certain deployments. It could also reduce import risk and compliance complexity. None of this guarantees smoother supply. But it changes the odds. And it creates an ecosystem gravity in the U.S. for optics talent, packaging, and test.
Competitive Dynamics: Turning Optics Into The Next Platform Edge
In the last decade, Nvidia turned GPUs into a platform. Then it turned networking into a platform, helped by Mellanox.
Optics looks like the next layer. The market is already moving. Rivals and partners are investing in photonics and optical switching.
So the real question is: who “owns” the standards, the interfaces, and the capacity?
An acquisition could shift leverage versus other chip and network vendors. Hyperscalers are also building custom silicon.
That makes Nvidia’s full-stack pitch more important. If Nvidia can offer compute plus networking plus optics that is pre-validated, it reduces customer integration work. Customers still get choice. The deals are nonexclusive today.
But ownership could pull more of the roadmap into Nvidia’s orbit. That can matter in procurement decisions.
There is a counterweight too. Vertical moves can spook partners. Optical component ecosystems are interconnected. Some customers may prefer multi-vendor optics. Regulators may also scrutinize consolidation. And competitors could respond by funding alternative suppliers. Still, from a strategic view, optics is becoming a gating item. In a world where networking revenue is already huge for Nvidia, controlling optics could protect that profit pool.
Final Thoughts
Nvidia’s twin $2B investments, plus multibillion-dollar purchase commitments, look like an effort to get ahead of the next choke point.
Optical interconnects and co-packaged optics can improve bandwidth per watt. They can also reduce wasted GPU time.
The deals also aim to expand U.S. photonics capacity, which could help resilience.
And they may reshape competitive dynamics in components and networking.
There are real tradeoffs. Optics manufacturing is complex. Ramps can slip. Integration can create partner tensions. And customers still want optionality. The deals being nonexclusive is a clue that Nvidia sees that.
On valuation, the market still prices Nvidia like a special asset. Using your March 2, 2026 table, Nvidia trades around ~20.29x LTM EV/Revenue, ~32.89x LTM EV/EBITDA, and ~37.24x LTM P/E. That is rich by most large-cap standards. It also reflects how central Nvidia is to AI build-outs. These optics moves support that narrative. They do not remove execution risk. They mostly shift where the next hard work lives.
Disclaimer: We do not hold any positions in the above stock(s). Read our full disclaimer here.
