A Detailed Look at Meta and Its $75 Billion Bet on the Future of AI

A Detailed Look at Meta and Its $75 Billion Bet on the Future of AI

Ondřej Barták
Ondřej Barták
Entrepreneur and Programmer
31. 10. 2025
3 minutes reading · 6 views
A Detailed Look at Meta and Its $75 Billion Bet on the Future of AI

In the third quarter of 2025, Meta spent 37% of its revenue on capital expenditures. This means that for every crown the company earns, more than a third goes toward building artificial intelligence (AI) infrastructure. By comparison, it was only 20% last year, and it is now the highest ratio in the company's history.

Between September 30 and October 27, 2025, Meta signed four major contracts totaling $75.5 billion (approximately CZK 1.74 trillion). These agreements include a partnership with CoreWeave worth $14.2 billion (approximately CZK 327 billion) for six or more years, with Oracle for roughly $20 billion (approximately CZK 460 billion) over several years, with Scale AI for $14.3 billion (approximately CZK 329 billion) for a 49% stake, and finally with Blue Owl/Hyperion for $27 billion (approximately CZK 621 billion) as part of a joint venture.

A Different Approach to Infrastructure Financing

Mark Zuckerberg talks about building "superintelligent" AI systems, but the key lies in how Meta finances these investments. For example, in the agreement with Blue Owl Capital for the Hyperion data center in Louisiana, Blue Owl holds an 80% ownership stake and Meta only 20%. The financing is provided through more than $27 billion in debt (approximately CZK 621 billion). Meta received $3 billion (approximately CZK 69 billion) in cash upfront and made a 16-year commitment through a residual value guarantee.

This model changes the old approach, in which companies purchased and directly owned servers. Meta now effectively leases infrastructure financed by private credit, enabling faster growth but carrying the risk of long-term payment obligations.

Meta CEO Mark Zuckerberg

What Does Meta Get for the Money?

The Hyperion data center, spanning 2,250 acres in Richland Parish, Louisiana, will have 2 gigawatts (2,000 megawatts) of power, enough to supply 1.5 million homes. Completion is scheduled for 2030, and Entergy is building a $1.2 billion (approximately CZK 27.6 billion) transmission line. This will make it possible to train GPT-4-sized models several times simultaneously, supported by hundreds of thousands of GPUs and petabytes of data.

Through CoreWeave, Meta is gaining access to Nvidia GB300 servers with 72 Blackwell GPUs per rack, available through December 2031 with an extension option. These GPUs offer 2.5 times the performance of the previous-generation H100 and greater energy efficiency.

The Oracle deal involves $20 billion (approximately CZK 460 billion) in cloud capacity for inference and distributed training. In addition, Meta signed an agreement with ENGIE for 1.3 gigawatts of solar power from four projects in Texas, including the 600-megawatt Swenson Ranch Solar project, which will be fully dedicated to Meta beginning in 2027.

The Economics Behind the Investments

Meta plans to invest $600 billion (approximately CZK 13.8 trillion) in U.S. data centers by 2028. In 2025, this will amount to $66–72 billion (approximately CZK 1.52–1.66 trillion), equivalent to 37% of revenue and up 44% from last year. Estimates for 2026 project $97 billion (approximately CZK 2.23 trillion).

Although AI is estimated to generate an additional $5–8 billion (approximately CZK 115–184 billion) in advertising revenue, infrastructure costs reach $80 billion (approximately CZK 1.84 trillion) annually, creating an imbalance. Meta is betting on improved advertising, new products, and the sale of excess capacity, but the risk lies in whether the investments will pay off in time.

Source: allenarch.dev

Advertisement

Content created with help from UpTier.

SEO and GEO on autopilot. UpTier’s multi-agent systems write and optimize content for search engines and AI answers.

Discover UpTier ↗

Category:AI
Did you enjoy this article?
Discover more interesting posts on our blog
Back to blog

Related posts

AI Fever Cools on Wall Street. Nvidia CEO’s Exit Sends AI Stocks PlungingAI Fever Cools on Wall Street. Nvidia CEO’s Exit Sends AI Stocks Plunging
AI stocks are tumbling after the Nvidia CEO’s unexpected departure, dragging down the entire market. U.S. restrictions on China and Nvidia’s upcoming earnings are adding to the jitters.
6 min read
18. 11. 2026
Trump Wants to Push Chinese AI Out of Global Markets. Nvidia, OpenAI and Anthropic Are His ChampionsTrump Wants to Push Chinese AI Out of Global Markets. Nvidia, OpenAI and Anthropic Are His Champions
Washington wants to turn Nvidia, OpenAI and other companies into national champions and use ready-made AI packages to push Chinese rivals out of global markets.
4 min read
18. 11. 2026
NVIDIA introduces Nemotron for Science to develop scientific AI agentsNVIDIA introduces Nemotron for Science to develop scientific AI agents
Nemotron for Science brings together an open model, a chemistry dataset and post-training recipes that researchers can use to build agents for specialized scientific tasks.
2 min read
18. 11. 2026
Přihlaste se k odběru našeho newsletteru
Zůstaňte informováni o nejnovějších příspěvcích, exkluzivních nabídkách, a aktualizacích.
CodedTrip

Operated by CodedTrip LLC, USA.

YouTube
TikTok