AMD Smashes Q1 Targets With 38% Revenue Growth as Meta Commits $60 Billion to MI450 AI Chips

For years, the AI chip market has been a story about Nvidia. On May 5, 2026, Advanced Micro Devices wrote a new chapter. The chipmaker reported a record-breaking quarter, confirmed a $60 billion deal with Meta Platforms that is the largest AI chip contract in history, and sent its stock surging 8% in after-hours trading — all in a single evening.

Record Quarter: $10.3 Billion and 38% Growth

AMD’s first-quarter 2026 results beat Wall Street expectations across every key metric. Revenue came in at $10.3 billion, surpassing the $9.9 billion analyst consensus and growing 38% year-over-year. Adjusted earnings per share reached $1.37, above the $1.29 estimate. The company generated strong free cash flow, and gross margins held at 56.8% — a level that puts AMD firmly in the premium tier of chip economics.

The standout segment was Data Center, which posted $5.8 billion in revenue — up 57% year-over-year and a new quarterly record. That figure accounts for more than half of AMD’s total revenue for the first time in the company’s history, a milestone that reflects how thoroughly the AI buildout has reshaped the semiconductor industry’s value chain.

Looking ahead, AMD guided for second-quarter revenue of approximately $11.2 billion, plus or minus $300 million — well above the $10.5 billion Wall Street was expecting and implying continued sequential acceleration.

The Meta Deal: 6 Gigawatts and $60 Billion

Buried inside the earnings call was the announcement that has already upended AI chip market dynamics: AMD has secured a multi-year agreement with Meta Platforms to supply custom Instinct MI450 GPUs and sixth-generation EPYC “Venice” CPUs, deployed on AMD’s Helios rack-scale architecture. The deal is valued at approximately $60 billion over five years.

The scale is almost unprecedented. The agreement covers up to 6 gigawatts of compute capacity — a number that illustrates how AI infrastructure has become a power engineering challenge as much as a semiconductor one. The first deployment tranche of 1 gigawatt is scheduled to begin in the second half of 2026, with subsequent waves following as Meta’s AI compute requirements grow.

For context, Meta has been spending at the rate of $60 billion or more annually on capital expenditure across its operations, with AI infrastructure now the dominant driver. Selecting AMD’s MI450 for the next wave of that buildout — rather than defaulting to Nvidia’s Grace Blackwell architecture — is a decision that will be studied in semiconductor boardrooms for years.

Breaking Nvidia’s Structural Advantage

Nvidia has held near-monopoly status in AI training and inference workloads since 2022. Its CUDA software ecosystem, which has been in development for over a decade, creates switching costs that go beyond hardware. Developers write code for CUDA; rewriting it for a competitor’s stack requires time, money, and risk.

AMD has spent billions building out ROCm — its open-source answer to CUDA — and the Meta deal suggests that investment is paying off at scale. Whether AMD can sustain ROCm parity as Nvidia continues to advance CUDA is the central technical question, but the existence of a $60 billion contract indicates that at least one of the world’s most sophisticated AI infrastructure teams has concluded the gap is manageable.

“Customer engagement around MI450 Series and Helios is strengthening, with leading customer forecasts exceeding initial expectations,” said AMD CEO Lisa Su on the earnings call. “We’re seeing a growing pipeline of large-scale deployments across cloud and enterprise customers.”

Su was careful not to oversell the competitive narrative, but the numbers speak for themselves. AMD’s Instinct data center GPU revenue has grown at a compounded rate exceeding 100% over the past six quarters.

The Helios Architecture: Rack-Scale AI

The MI450 deal is not just about individual GPUs. AMD’s Helios platform is a rack-scale architecture that integrates compute, memory, and networking into a unified system optimized for large AI model training and inference. By offering a complete rack-level solution rather than discrete components, AMD is competing in the same tier as Nvidia’s DGX systems — the gold standard for enterprise AI infrastructure.

Helios incorporates AMD’s Infinity Fabric interconnect, which allows MI450 GPUs to share memory across multiple chips with lower latency than traditional NVLink alternatives. For the workloads Meta is running — large language models, recommendation systems, video generation at scale — that interconnect efficiency translates directly into throughput and energy savings.

What It Means for the AI Chip Market

The Gartner forecast puts global semiconductor revenue above $1.3 trillion in 2026, the highest growth rate in two decades, with AI chips projected to represent 30% of total revenue. AMD’s results and the Meta deal confirm that this market is now large enough to support more than one dominant supplier.

That does not mean Nvidia is threatened in the near term. The company’s upcoming Rubin architecture and its CUDA moat remain powerful advantages. But the era of a single-vendor AI chip market — where every major hyperscaler had no practical alternative to Nvidia — appears to be ending.

For enterprise buyers, the emergence of a credible second supplier has direct implications on pricing, negotiating leverage, and supply chain resilience. Hyperscalers have been quietly investing in AMD’s roadmap for precisely this reason. Meta’s bet validates that investment in the most public way possible.

AMD’s stock, which had already surged 74% in April as investors positioned ahead of earnings, traded up another 8% in after-hours markets on May 5 — evidence that even a well-anticipated beat can surprise when the magnitude is large enough. The AI chip war has a genuine second front. It opened at $60 billion.