There comes a time when a business acts more like an empire builder than a manufacturer. That moment may have already passed for Nvidia—quietly, without much fanfare, buried in a few press releases and a 13F filing that most people ignored. Nvidia invested $3.8 billion in two businesses in the first quarter of this year: Coherent, a manufacturer of photonics and networking hardware, and CoreWeave, an AI data center operator it had previously supported. When combined, the actions show something more intentional than a portfolio play. They resemble ownership of infrastructure disguised as investments.
Nvidia’s share of CoreWeave has almost doubled. The company currently owns about 47.2 million shares, worth about $3.66 billion, or about 11% of a company that basically uses Nvidia’s own GPUs. Companies developing AI systems, such as Anthropic, Perplexity AI, and Meta, which reportedly inked a $21 billion deal, rent computing power from CoreWeave. It reads like a who’s-who of the AI build-out, that customer list. And Nvidia sits behind it all, providing the hardware necessary for it to function. Investing more in CoreWeave is more than just a financial decision. Without the legal complications, it is more akin to vertical integration.
A different pressure point is addressed by the $1.8 billion Coherent investment, which was announced along with a partnership for research and development. The physical limitations of copper interconnects become a significant bottleneck as AI data centers become denser and more power-hungry. Photonics components, a type of light-based data transmission hardware that can transfer data more quickly and with less heat than conventional wiring, are produced by Coherent. Jensen Huang might have realized that photonics was a necessity rather than merely a nice-to-have when he looked at the roadmap for next-generation data centers. The Coherent deal begins to resemble Nvidia protecting its own supply chain rather than diversification.
Observing all of this, it seems as though the chip industry isn’t sufficient to fulfill the ambition here. Over 88% of Nvidia’s $215.9 billion in revenue in fiscal 2026 came from its data centers. Those figures are astounding. However, it appears that the company is aware of how vulnerable it is as a supplier to the entire AI ecosystem, relying on partners and customers to actually develop things quickly and on a sufficient scale. That equation is somewhat altered if you own some of those partners.
CoreWeave is currently trading at about $109, which is about 21% less than what analysts believe the stock is worth. Forecasts indicate a 5.2% annual decline in earnings over the next three years, so it’s still unclear if that gap represents true undervaluation or a market uneasy with the company’s debt load and earnings trajectory. That is a serious issue. Stronger relationships with Nvidia might make it easier for CoreWeave to obtain GPUs and lower supply risk, but they won’t automatically solve the profitability conundrum.
Here, the historical parallel is difficult to ignore. Microsoft made a big investment in OpenAI years ago, which at the time seemed strategic and ultimately changed the way the entire tech sector viewed AI collaborations. Nvidia’s actions seem consistent with that reasoning. Though not the same, they are connected. The business at the forefront of one technological wave uses its resources to influence the design of the subsequent layer. Execution, timing, and a few other factors that are currently unpredictable will determine whether that is profitable. However, at least the direction appears to be fairly obvious.
