--- title: "More Than Just GPUs: Meta Platforms Bets Billions on Amazon Graviton as AWS Self-Developed CPUs Enter AI Computing Core" type: "News" locale: "en" url: "https://longbridge.com/en/news/284007175.md" description: "Meta Platforms and Amazon have reached a multi-billion dollar agreement to lease tens of millions of Graviton CPU cores to support its AI agent business. As AI enters the 'post-training' and agent era, CPUs are regaining mainstream attention due to their collaborative computing advantages. Meta Platforms is building a diversified computing power system by partnering with NVIDIA, Arm, and Amazon to break the single-dominance pattern of GPUs" datetime: "2026-04-24T12:59:02.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/284007175.md) - [en](https://longbridge.com/en/news/284007175.md) - [zh-HK](https://longbridge.com/zh-HK/news/284007175.md) --- # More Than Just GPUs: Meta Platforms Bets Billions on Amazon Graviton as AWS Self-Developed CPUs Enter AI Computing Core Meta Platforms and Amazon have reached a chip procurement agreement spanning several years and worth billions of dollars, under which Meta Platforms will use Amazon Web Services' (AWS) Graviton CPU chips to support its AI agent business. This move marks a quiet shift in the long GPU-dominated AI chip landscape. According to the agreement, **Meta Platforms will lease tens of millions of AWS Graviton chip cores, most of which will be deployed within the United States.** Neither company disclosed specific financial terms, but Nafea Bshara, Vice President and Distinguished Engineer at Amazon, stated that **the agreement has a term of three to five years.** This collaboration will place Meta Platforms among the top five largest Graviton customers on AWS. This transaction continues Meta Platforms' recent series of major infrastructure investments. Over the past few weeks, the company has signed agreements with CoreWeave and Nebius, totaling $48 billion, primarily to lease NVIDIA GPU resources. Meanwhile, on Thursday, Meta Platforms announced plans to lay off approximately 8,000 employees, representing 10% of its workforce, to offset the continuously high costs of AI investments. For AWS, this agreement further validates the commercial feasibility of Graviton chips in AI workloads. Later that same week, Amazon added a $5 billion investment in Anthropic, an agreement that also includes provisions for large-scale Graviton core usage by Anthropic. Amazon's pre-market stock price rose nearly 2%. ## Deal Details: Massive Scale Covering Core AI Agent Scenarios According to information disclosed by both parties, **Meta Platforms will use hundreds of thousands of Graviton physical chips under this agreement framework, collectively accessing tens of millions of chip cores.** Bshara noted that Meta Platforms had previously used Graviton on a small scale, making this a significant leap forward in expansion. Meta Platforms is purchasing **AWS's latest generation product, Graviton5**—a 3-nanometer process CPU chip based on the Arm architecture. Bshara stated that on the AWS EC2 platform, Graviton offers the best cost-performance ratio among equivalent computing options, **with energy consumption approximately 60% lower than similar products.** "Meta Platforms has extremely rich choices on the supply side, but they chose Graviton5 precisely because of its balance of price and performance," Bshara said. Santosh Janardhan, Head of Infrastructure at Meta Platforms, stated in a declaration: "Expanding to Graviton enables us to run CPU-intensive workloads behind AI agents at our scale with the required performance and efficiency." ## CPUs Regain Favor as AI Agents Drive Demand Reversal In the past few years, the narrative of AI computing power was almost entirely dominated by GPUs, with CPUs relegated to supporting roles. However, with the rise of AI Agents, this landscape is quietly changing. Brendan Burke, Director of Semiconductor Research at market research firm Futurum Group, pointed out that **CPUs and GPUs exhibit strong synergy in AI workloads—CPUs handle running specific applications and feeding tasks back to GPUs, with both types of chips jointly supporting AI agents in executing diverse tasks.** CPUs also play a critical role in the "post-training" phase of large language models—the process of fine-tuning models for specific objectives after pre-training is complete, where CPUs undertake important computational functions. Intel CEO Lip-Bu Tan confirmed this trend during an analyst meeting on Thursday, stating that his Xeon server chips are currently in short supply. "In the past few years, the narrative for high-performance computing was almost exclusively centered around GPUs and other accelerators. But in recent months, we have seen clear signals that CPUs are once again becoming an indispensable foundation for the AI era," he said. Burke believes, "For the most cutting-edge AI labs, there is virtually no upper limit to their demand for CPUs." ## Meta Platforms' Diversified Chip Strategy Continues to Expand This cooperation between Meta Platforms and AWS is part of Meta Platforms' ongoing strategy to diversify chip procurement throughout the year. **In addition to this Graviton CPU agreement, Meta Platforms has already reached cooperative agreements with NVIDIA, Advanced Micro Devices (AMD), and Arm Holdings earlier this year.** Meta Platforms stated that this new agreement reflects the company's philosophy of diversified infrastructure layout and indicates that no single chip architecture can efficiently handle all computing tasks. Meta Platforms' ambition in AI agents also provides direct impetus for these demands. The company acquired AI agent startup Manus in December last year for over $2 billion; Manus specializes in developing AI agent products capable of executing complex tasks. Additionally, early this month, Meta Platforms released Muse Spark, its first new AI model in a year, and indicated plans to release more models subsequently. The cooperation between Meta Platforms and AWS dates back to around 2016, but previously focused mainly on core cloud services, the use of Amazon Bedrock platforms, and leasing GPU clusters from AWS. Bshara noted that Meta Platforms began leasing NVIDIA GPUs from AWS starting in 2017. ## AWS Leverages Momentum to Consolidate Graviton Commercial Footprint For AWS, securing Meta Platforms as a landmark client is a significant milestone in the large-scale commercialization process of Graviton chips. Bshara is a co-founder of Annapurna Labs, a chip company acquired by Amazon in 2015, which has since become the core force behind internal chip R&D at AWS. **Graviton has attracted adoption by well-known companies such as Adobe, Apple, and Snowflake, and is currently further consolidating its market position in the AI infrastructure sector.** Bshara stated, "Graviton is one of the most commonly used pre-training platforms for many foundational model companies; Meta Platforms is the latest important customer to join." At the same time, AWS has not completely focused on a self-developed route in the chip field. In March this year, AWS announced a partnership with Cerebras, an AI inference chip startup, to deploy Cerebras inference chips within its data centers, further broadening the diversity of its chip ecosystem. Burke believes that this agreement between Meta Platforms and AWS will further strengthen market recognition of Graviton's value in the AI era. "This is a highly significant proof-of-concept case," he said. ### Related Stocks - [AMZN.US](https://longbridge.com/en/quote/AMZN.US.md) - [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md) - [META.US](https://longbridge.com/en/quote/META.US.md) - [XDAT.US](https://longbridge.com/en/quote/XDAT.US.md) - [AMZU.US](https://longbridge.com/en/quote/AMZU.US.md) - [CLOU.US](https://longbridge.com/en/quote/CLOU.US.md) - [CRWV.US](https://longbridge.com/en/quote/CRWV.US.md) - [NBIS.US](https://longbridge.com/en/quote/NBIS.US.md) - [ARM.US](https://longbridge.com/en/quote/ARM.US.md) - [INTC.US](https://longbridge.com/en/quote/INTC.US.md) - [AMD.US](https://longbridge.com/en/quote/AMD.US.md) - [ADBE.US](https://longbridge.com/en/quote/ADBE.US.md) - [AAPL.US](https://longbridge.com/en/quote/AAPL.US.md) - [SNOW.US](https://longbridge.com/en/quote/SNOW.US.md) - [NVD.DE](https://longbridge.com/en/quote/NVD.DE.md) ## Related News & Research - [Meta’s loss is Thinking Machines gain](https://longbridge.com/en/news/284053848.md) - [Anthropic Commits $100B to AWS as Amazon Bets Up to $25B on AI Partner. Is AMZN a Buy Now?](https://longbridge.com/en/news/284008126.md) - [Meta Platforms, Inc. $META Shares Sold by Bridgewater Advisors Inc.](https://longbridge.com/en/news/283953573.md) - [Key facts: Amazon $200B AI Capex; Supplies Graviton to Meta; Backs SMRs](https://longbridge.com/en/news/284049294.md) - [Meta Platforms (META) Eyes Major May Layoffs Amid Mounting AI Expenses](https://longbridge.com/en/news/283200664.md)