Three AI infrastructure leaders (Nvidia, Alphabet, Meta) emerge as Nasdaq hits all-time high — each betting differently on who wins the AI buildout

Yahoo Finance AI · April 25, 2026

AI Summary

•Nvidia expanded beyond GPU chips to become a full AI infrastructure provider: it now sells CPUs (general processors), DPUs (data-handling chips), and LPUs (language processing chips acquired from Groq), allowing it to package complete server solutions for specific AI tasks like training, inference, and agent AI. This positions Nvidia to capture revenue across the entire stack rather than losing customers to specialized competitors.
•Alphabet designed its own custom TPUs (tensor processing chips) a decade ago and optimized all its hardware and software around them — the company just released the eighth generation in two versions: one for training AI models and one for running them in production. This custom hardware gives Alphabet cost advantages that competitors cannot match, meaning it can train its Gemini model more cheaply and reinvest savings into better ads and search results.
•Meta redirected billions from its money-losing metaverse project toward AI infrastructure and custom chips, using AI to keep users on Facebook, Instagram, and Threads longer by showing them more relevant content — which then lets Meta serve more and higher-priced ads. Meta is now starting to place ads on WhatsApp (its messaging platform with 2 billion users), creating new revenue streams unavailable to competitors.

Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.