SANTA CLARA, July 12 — The figure that matters in the AI arms race right now is 350,000. That is the number of Nvidia H100 graphics processing units Meta Platforms committed to acquiring in January 2025 as part of a plan to invest $60–$65 billion in AI-related capital expenditures for the year. The bottom line: Meta is spending more than $10 billion on these chips alone.
The H100, launched by Nvidia in 2022, carries a price tag of roughly $30,000 per unit. By the numbers, Meta’s GPU acquisition represents a staggering single-company bet on Nvidia’s hardware.
Nvidia, headquartered here in Santa Clara, reported $60.9 billion in data center revenue for fiscal year 2024. That figure was driven largely by demand from companies like Meta, which are racing to build the infrastructure for large-scale AI models. Meta’s 2025 capital expenditure target of $60–$65 billion marks a sharp jump from $37.2 billion in 2024.
The spending spree includes building a massive AI data center alongside the GPU haul. Investors will note that this is not a small pilot program; it is a full-scale industrial buildout designed to support Meta’s large language model LLaMA and its successor, LLaMA 2.
LLaMA and the competitive landscape
Meta first released LLaMA in February 2023, with LLaMA 2 following in July 2023. These models are central to the company’s strategy to compete with OpenAI — backed by Microsoft — and Google’s DeepMind. The H100 GPUs are specialized for AI training and inference, meaning they handle both the heavy lifting of teaching a model and the real-time work of running it once deployed.
The investment signals Meta’s commitment to staying in the race. The company is effectively buying its way into the top tier of AI infrastructure, matching the spending of rivals who have already poured billions into data centers and chip supply.
For Nvidia, the order book keeps growing. The H100 has become the default workhorse for AI computing, and Meta’s 350,000-unit order is among the largest single purchases publicly disclosed.
What to watch next
The question for investors and industry watchers is whether Meta can translate this hardware spending into a competitive advantage with LLaMA. The company has positioned its models as open and accessible, a contrast to the more proprietary approaches of OpenAI and Google. With the data center buildout underway and the H100s arriving, the next milestone will be the release of LLaMA’s next generation and how it stacks up against the competition.
Meta has placed its bet. Now the market waits to see if it pays off.
Sources
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “Meta’s AI Spending Hits $65 Billion, Boosts Investor Confidence”, “description”: “SANTA CLARA, July 12 — The figure that matters in the AI arms race right now is 350,000. That is the number of Nvidia H100…”, “datePublished”: “2026-07-12T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Ava Anderson”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}, “citation”: [{“@type”: “CreativeWork”, “url”: “https://en.wikipedia.org/wiki/Nvidia_H100”}]}
