Nvidia

Nvidia designs the accelerators most AI training and inference runs on, which makes its disclosures the closest thing the sector has to a demand gauge. Its filings, supply commentary and customer concentration move semiconductor equities, data-centre construction and, increasingly, the financing structures behind both.

What we track on this desk

Disclosed stakes and investments. Equity positions in customers and suppliers, which change how to read a revenue line.
Supply and allocation. Who gets parts, in what order, and what that says about which buyers are real.
Data-centre financing. The capital structures behind capacity, tracked with AI data centres.
Export controls and diversion. Where restricted parts turn up, and how firmly that is established.
Crypto-adjacent demand. Miners converting to AI hosting, where our mining coverage overlaps.

How to read an Nvidia story here
Two failure modes dominate this beat. The first is the unsourced number: a capacity, order-book or investment figure that originates with one outlet and is repeated everywhere without anyone adding reporting. Our stories carry a count of how many independent publishers actually carried a claim, so a figure resting on a single report is visibly a single report.
The second is the inferred customer. Supply-chain stories routinely name buyers who have confirmed nothing. Where a company has not said something itself and no document shows it, we attribute the claim to whoever made it rather than to the company.

Where to go next
See the AI desk for the model side, Semiconductors for the manufacturing chain, or the Verification Center for how publishers are counted.

Semiconductors
AI Data Centres
SpaceX
AI
Markets