Nvidia's Capital Web Tightens as Vera Rubin Ramps Across Five Clouds

Nvidia backs AI startups and cloud providers, ramps Vera Rubin, but faces memory shortages and China export limits.

The most telling signal about Nvidia's strategy no longer comes from its earnings calls — it comes from the list of companies it is writing checks to. Over the past week, the chipmaker's name has surfaced in a string of funding rounds across the AI ecosystem, marking a decisive shift from supplier to financier of the very infrastructure that consumes its products.

The pattern crystallized with reports that Nvidia is in talks to back a Perplexity funding round that would value the AI search startup at more than $30 billion. That followed news of a $1.5 billion private placement commitment tied to SB Energy's planned US listing, a deal notable for the fact that SB Energy and OpenAI both depend heavily on Nvidia chips for their Ohio data center project, according to the IPO prospectus. Days earlier, Nvidia had joined a $125 million financing round for iPronics, a Spanish startup developing optical networking chips for data centers.

A Self-Reinforcing Cycle

The strategic logic is becoming clearer by the week. Nvidia is effectively underwriting demand for its own silicon while simultaneously bankrolling the supply chain that supports it. Anthropic's multibillion-dollar cloud contract with Lambda — a $35 billion deal for a Texas data center, confirmed by Bloomberg and the Wall Street Journal — adds another layer to this web, given Lambda's own Nvidia backing.

Yet questions are mounting about how much of this activity represents genuine end-market demand rather than circular value creation. The Wall Street Journal reported that Nvidia has paused some revenue-sharing agreements with AI cloud providers that were part of a financing initiative. Nvidia responded that the underlying program remains active and is "continuously evolving" due to high demand.

Vera Rubin Goes Live

On the product front, Nvidia's next-generation Vera Rubin platform has moved from announcement to production. The racks are now running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius, with the company also reporting full production of the Groq 3 LPX accelerator for interactive AI inference. SpaceXAI will deploy Nvidia's Vera CPUs for agentic AI workloads, while the deepened MediaTek partnership extends into AI infrastructure, edge computing and automotive platforms.

The numbers behind this expansion remain formidable. Revenue more than doubled year-over-year in the latest quarterly report, with the company guiding to $108 billion for the current quarter, plus or minus 2 percent. The stock closed Thursday at €196.48, up 1.4 percent on the day, and has gained 4.6 percent over seven days and 23 percent since the start of the year. It now sits roughly 3 percent below its May peak of €202.50.

The Memory Crunch Bites

Not everything is running smoothly. Nvidia warned that memory chip shortages could constrain growth, and the impact is already visible in the consumer segment. The planned refresh of the RTX 50-series "Super" line was scrapped in December 2025, while the RTX 60-series has been pushed back to 2028. The culprit is the global DRAM and HBM shortage that industry watchers have dubbed "RAMageddon" — Nvidia is directing scarce memory supplies toward far more profitable AI accelerators rather than gaming cards.

Lenovo and Acer are still expected to ship the first Windows PCs with Nvidia's RTX Spark chip in October, part of the company's effort to carry its AI narrative beyond the data center. But for now, the consumer side remains a secondary consideration to the data center bonanza.

China Remains Off the Table

One market stays firmly out of reach. Chinese antitrust authorities issued a preliminary finding last September that Nvidia's compliance with US export controls — which mandate weakened products for Chinese customers — was discriminatory and violated conditions tied to the Mellanox acquisition. By the end of fiscal 2026, Nvidia was effectively shut out of the Chinese data center market, unable to offer a competitive product there. Its latest guidance accordingly assumes zero revenue from that segment.

That leaves North America, Europe and other Asian markets to carry the growth burden — a constraint that has yet to slow the company but remains a structural vulnerability if geopolitical tensions persist. The next earnings report, due November 17 after market close, will show whether the Vera Rubin ramp is meeting expectations and how deeply the memory shortage cuts into consumer margins.