AI Infrastructure Dashboard

Overview

I built this dashboard live, in about 15 minutes, during a VS Code workshop I hosted for the BIS club. The point of the workshop was to show members how to build and deploy a website end to end: build locally, push to a GitHub repo, deploy through Vercel, pull stock data from an API, connect that data to the site, and add an AI reasoning model to it.

The business case

A workshop needs a problem worth solving, and the one we brought was tracking the public companies supporting global AI growth. During the California gold rush, the people who made the most money were not the ones finding gold but the ones selling pickaxes, and the AI build-out has the same shape: whoever wins on models, every one of them needs chips, networking, data centers, power, and cooling.

The dashboard tracks 17 of those companies: Nvidia, Broadcom, Marvell, and Arm on silicon, Arista and Corning on networking, Super Micro on hardware, Equinix, Digital Realty, Iron Mountain, Applied Digital, and IREN on data centers, Constellation, Vistra, and Eaton on power, Vertiv on power and cooling, and CoreWeave on high-performance compute. Each company is tagged with one or two of seven segments, and the segment filter narrows the whole dashboard to one part of the build-out.

What we built

The stack is Next.js and TypeScript, deployed on Vercel. A server route pulls quotes and price history for all 17 stocks from Yahoo Finance through the yahoo-finance2 library, caches the result for 30 seconds, and serves the last good data if a fetch fails. The page refreshes it every 60 seconds.

The top row sums the group: total market cap, the average AI score, and how many stocks are up against how many are down. A performance chart ranks the filtered stocks by their one-day, one-week, or one-month move, or by AI score. The table lists every stock with its price, one-day, one-week, and one-month change, market cap, volume, 52-week high and low, and AI score, sortable by column. The AI score is a fixed number for each company set in the code, from 61 for Corning to 99 for Nvidia.

Clicking a stock opens a detail panel with its one-year change, a 90-day price chart, where the price sits in its 52-week range, volume against its three-month average, fundamentals including trailing and forward P/E, EPS, revenue, and gross margin, and analyst buy, hold, and sell counts with the average price target.

The panel also has an AI analysis section. It sent the stock's price and recent moves to Claude through my own Anthropic API key and streamed the written analysis back into the page. That key was only for the demonstration build and the walkthrough with club members, and it has since expired, so the AI analysis no longer returns anything. Everything else on the dashboard still runs on live data.

The public repository holds the first build. The detail panel and the AI analysis are on the live site but not in the repository.

Limitations

This was a brief demonstration build for a business case, made in 15 minutes, and it was never fully completed or audited. That includes the AI scores and the ticker list, neither of which was checked after the workshop.