Meta spends billions of dollars to tie up with Amazon AWS! Snapping up hundreds of thousands of Graviton5 chips to counter NVIDIA's AI computing monopoly

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯1374 words

Meta has reached a multi-billion dollar deal with AWS, expected to sweep up hundreds of thousands of AWS's latest self-developed Graviton5 CPUs for its data centers to fully support its AI infrastructure and inference workloads. This is not only a key strategy for Meta to optimize its capital expenditure, which is set to exceed $100 billion in 2026, but also proves the strong competitiveness of AWS's self-developed chips in the battle among top AI giants. (Context: AI data centers have become carbon bombs: annual greenhouse gas emissions exceed those of entire countries, causing the carbon reduction commitments of OpenAI, Meta, Microsoft, etc., to collapse) (Background: White House OSTP calls out China: Industrial-scale AI distillation attacks violate safety protocols, four-step counter-sanctions to be launched) In 2026, as the global AI arms race intensifies, tech giants are scrambling to break their fatal reliance on single chip suppliers. Meta (parent company of Facebook) and Amazon Web Services (AWS) have recently officially reached a multi-year, multi-billion dollar strategic infrastructure agreement. Meta will adopt Amazon's self-developed chips on a massive scale to inject new computing power into its vast AI empire. The core of this agreement is that Meta will deploy hundreds of thousands of Amazon's custom-designed Graviton series processors (specifically the latest generation of Arm-based Graviton5 CPUs) across its global data centers and AI workloads. Although AWS executives did not disclose the exact contract amount, they confirmed that the scale is massive. These chips will be primarily used for general-purpose computing, AI inference, and some training tasks. Compared to traditional x86 architectures, the Graviton series offers significant advantages in "price-performance ratio" and "energy efficiency," which can substantially reduce the cooling and power burdens of large data centers. In fact, this is not the first time the two sides have worked together. Meta and AWS have had deep cooperation in cloud services and the deployment of Llama open-source models in the past; however, Meta's decision to "explicitly and massively" buy into Amazon's self-developed hardware is undoubtedly a major upgrade in their strategic partnership. Why did Meta place this huge order? The answer lies in the increasingly uncontrollable AI computing costs and supply chain risks. - Optimizing massive capital expenditure: Meta is at the peak of its AI investment, with its 2026 capital expenditure (CapEx) projected to soar to a staggering $115 billion to $135 billion. Introducing energy-efficient Graviton chips can effectively lower the operating costs of AI inference and infrastructure expansion. - Diversifying sources of computing power: Faced with NVIDIA's absolute monopoly and expensive pricing in the market, Meta is striving to diversify risks. This batch of AWS chips will perfectly complement Meta's self-developed MTIA chips and solutions from partners like Broadcom and AMD. For Amazon, this contract is the strongest endorsement of its self-developed chip strategy (Graviton plus Trainium chips designed specifically for AI). Currently, AWS's custom chip business has an annualized revenue exceeding $20 billion and is growing rapidly. Securing top-tier AI demand from a "hyperscaler" client like Meta thoroughly proves the strong performance and commercial competitiveness of AWS hardware. From Meta allying with AMD and Broadcom, to Anthropic making huge commitments to AWS, and now Meta's massive procurement of Graviton5, all of this reflects a clear industry trend: in this life-or-death AI battle, giants are frantically diversifying their supply chains, attempting to seize the dominant power of computing that will determine the winner between performance and cost.

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching6 items

2026-04-24

AI data centers become carbon bombs: annual greenhouse gas emissions exceed entire countries, OpenAI, Meta, Microsoft... decarbonization pledges collapse

Similarity 120%關鍵字 meta同分類 zh

2026-04-24

Jensen Huang sends internal memo: 10,000 NVIDIA employees mandated to switch to OpenAI Codex, GPT-5.5 to run on GB200 chips

Similarity 120%關鍵字 nvidia同分類 zh

2026-04-24

Jensen Huang sends all-hands email embracing OpenAI Codex: over 10,000 NVIDIA employees are already using it, with GPT-5.5 running on GB200

Similarity 120%關鍵字 nvidia同分類 zh

2026-04-23