Amazon custom chips get a boost from Meta, giving the cloud giant another path to win in AI

```json { "title": "Meta and AWS Graviton Deal Reshapes AI Chip Race", "metaDescription": "Meta signs a multi-year deal to deploy tens of millions of AWS Graviton5 cores, making it one of Amazon's largest custom chip customers and boosting AMZN shares 3.5%.", "content": "<h2>Meta and AWS Strike Major Graviton Deal, Signaling a Shift in AI Infrastructure</h2>\n\n<p>On April 24, 2026, Amazon Web Services and Meta announced a significant expansion of their long-standing partnership, with Meta agreeing to deploy tens of millions of <strong>AWS Graviton5</strong> processor cores to power its next generation of agentic AI workloads. The agreement, which will last at least three years and involves hundreds of thousands of Graviton chips, makes Meta one of the largest Graviton customers in the world — and gives Amazon's custom chip business one of its most prominent public endorsements to date.</p>\n\n<p>The news landed on the same day markets were already watching the cloud sector closely, coinciding with the close of Google Cloud Next. Amazon shares responded immediately, jumping 3.5% in the afternoon session following the announcement. Analysts at BMO, UBS, and Oppenheimer all increased their price targets for the stock.</p>\n\n<p>For a company that serves approximately 3.6 billion daily users across its family of apps — and that is simultaneously operating 32 data centers to handle that load — the infrastructure decisions Meta makes carry enormous weight across the industry. This deal signals that weight is increasingly falling on Amazon's custom silicon.</p>\n\n<h2>What the AWS Graviton Deal Actually Involves</h2>\n\n<p>The Meta-AWS agreement centers on the deployment of AWS Graviton5 processors at scale. According to Meta's official announcement, the deal brings tens of millions of Graviton cores into Meta's compute portfolio, with the flexibility to expand beyond that baseline.</p>\n\n<p>The Graviton5 chip is purpose-built for CPU-intensive workloads. Built on 3-nanometer chip technology, it features 192 cores and a cache that is five times larger than the previous generation. According to Amazon's official data, Graviton5 delivers up to 25% better performance than its predecessor and reduces delays in core-to-core communication by up to 33% — specifications that make it particularly well-suited to the demands of agentic AI, where real-time reasoning, code generation, and search tasks require rapid, low-latency compute.</p>\n\n<p>Meta's head of infrastructure, Santosh Janardhan, explained the rationale directly: <em>"Expanding to Graviton allows us to run the CPU-intensive workloads behind agentic AI with the performance and efficiency we need at our scale."</em></p>\n\n<p>From Amazon's side, the significance of landing Meta as a Graviton customer was not understated. Nafea Bshara, Vice President and Distinguished Engineer at Amazon Web Services, noted: <em>"Graviton is one of the most used platforms for pre training by a lot of foundation model companies, and Meta is now one the newest one."</em></p>\n\n<h2>Amazon's Custom Chip Business Is Now a $20 Billion Operation</h2>\n\n<p>The Meta deal does not exist in isolation. It is part of a broader, accelerating trend in which major AI players are diversifying away from exclusive reliance on Nvidia GPUs toward a wider range of chip architectures — and Amazon is positioning itself at the center of that shift.</p>\n\n<p>In his annual shareholder letter, Amazon CEO Andy Jassy stated plainly: <em>"Virtually all AI thus far has been done on Nvidia chips, but a new shift has started."</em> He also disclosed the scale of what Amazon has built: <em>"Our annual revenue run rate for our chips business (inclusive of Graviton, Trainium, and Nitro — our EC2 NIC) is now over $20 billion, and growing triple-digit percentages YoY."</em></p>\n\n<p>That figure encompasses three distinct chip products. Graviton handles general-purpose CPU compute and is now powering Meta's agentic AI workloads. Trainium is Amazon's purpose-built AI training chip. Nitro is Amazon's custom network interface chip for EC2 instances. Together, they represent a vertically integrated silicon strategy that AWS has been building quietly for years and is now monetizing at scale.</p>\n\n<p>The Trainium chip, in particular, has attracted its own landmark commitments. According to TechCrunch, Anthropic — the maker of the Claude AI model — agreed to spend $100 billion over 10 years running its workloads on AWS, with a particular focus on Trainium, while Amazon agreed to invest an additional $5 billion into the company, bringing its total investment in Anthropic to $13 billion. Separately, OpenAI agreed to use Amazon's Trainium chips as part of a $100 billion cloud deal earlier in 2026, according to GeekWire.</p>\n\n<p>UBS analyst Stephen Ju, who reaffirmed a Buy rating on Amazon stock and raised his price target to $304 from $301, expects AWS revenue to grow 38% in 2026 — significantly ahead of the Street's consensus estimate of 26%.</p>\n\n<h2>Context: Meta's Broader AI Infrastructure Spending Spree</h2>\n\n<p>The AWS Graviton deal is the latest in a series of major infrastructure commitments from Meta. Prior to this announcement, Meta had already made a combined $48 billion in AI infrastructure commitments with CoreWeave and Nebius in recent weeks. The AWS deal adds to that total and further diversifies Meta's compute supply chain.</p>\n\n<p>It is worth noting that Meta's cloud relationships have grown more complex over time. According to TechCrunch, last August Meta signed a six-year, $10 billion deal with Google Cloud, though Meta had until then primarily been an AWS customer that also used Microsoft Azure. The new Graviton agreement reinforces Meta's multi-cloud approach while signaling that AWS remains a central partner for its most demanding AI workloads.</p>\n\n<p>The timing of the announcement — made one day after Meta disclosed plans to lay off approximately 8,000 employees, representing 10% of its workforce — underscores the dual pressures the company is navigating: aggressively investing in AI infrastructure while simultaneously cutting operational costs. For hyperscalers building at Meta's scale, these two moves are not contradictory. Efficient custom silicon can reduce the per-unit cost of compute even as total spending rises.</p>\n\n<p>Nafea Bshara framed the broader ambition behind the partnership: <em>"This isn't just about chips; it's about giving customers the infrastructure foundation, as well as data and inference services, to build AI that understands, anticipates, and scales efficiently to billions of people worldwide."</em></p>\n\n<h2>What This Means for the AI Chip Landscape</h2>\n\n<p>The competitive dynamics between major cloud providers are sharpening. The Meta-AWS Graviton announcement was timed to coincide with the close of Google Cloud Next, and both Amazon and Meta are scheduled to report quarterly earnings after the closing bell on Wednesday, April 29, 2026 — alongside Microsoft and Alphabet. Those earnings calls will offer the next detailed look at how AI infrastructure spending is translating into revenue across the sector.</p>\n\n<p>For AWS, the string of high-profile chip deals — with Meta on Graviton, Anthropic and OpenAI on Trainium — represents a meaningful validation of its bet on custom silicon. Amazon is no longer positioning Graviton and Trainium as cost-saving alternatives to Nvidia; it is positioning them as purpose-built solutions for specific AI workload categories, with major AI companies publicly choosing them for exactly those reasons.</p>\n\n<p>The Graviton5's technical specifications matter here. The combination of 192 cores, a five-times-larger cache, and up to 33% lower core communication latency makes it a credible choice for the kind of fast, iterative CPU tasks that agentic AI systems require. As AI applications move beyond single-turn interactions toward multi-step reasoning agents that must call tools, retrieve information, and generate responses in sequence, the CPU layer of the stack becomes increasingly important alongside GPUs.</p>\n\n<p>Whether this shift away from Nvidia represents a durable structural change or a tactical diversification by large buyers remains an open question — one the earnings season beginning April 29 may begin to answer.</p>\n\n<p>For more tech news, visit our <a href=\"/news\">news section</a>.</p>\n\n<h2>What to Watch Next</h2>\n\n<p>Both Amazon and Meta will report Q1 2026 earnings on April 29, 2026. Those results will offer the first detailed public accounting of how AI infrastructure investments are showing up in each company's financials. For Amazon, the key metrics to watch will be AWS revenue growth and any additional commentary from leadership on the trajectory of the custom chip business. For Meta, analysts will be listening for more detail on how the Graviton deployment fits into its broader AI compute roadmap and whether the infrastructure investments are translating into measurable performance improvements for its AI products.</p>\n\n<p>The three-year minimum term of the Graviton deal, combined with Meta's flexibility to expand beyond the initial tens of millions of cores, means the financial and operational implications of this partnership will compound over time. With the Anthropic and OpenAI Trainium deals running on similarly long timelines, AWS's custom chip revenue trajectory has multi-year visibility that is unusual in a sector where vendor relationships can shift quickly.</p>", "excerpt": "Meta has agreed to deploy tens of millions of AWS Graviton5 processor cores in a deal lasting at least three years, making it one of Amazon's largest custom chip customers. The announcement sent Amazon shares up 3.5% on April 24, 2026, as analysts raised price targets and CEO Andy Jassy confirmed Amazon's chip business now generates over $20 billion annually. The deal reflects a broader industry shift away from exclusive reliance on Nvidia GPUs toward purpose-built silicon for specific AI workloads.", "keywords": ["AWS Graviton", "Meta AI infrastructure", "Amazon custom chips", "agentic AI", "cloud AI chips"], "slug": "meta-aws-graviton-deal-reshapes-ai-chip-race" } ```