Lisa Su's AMD Posts $5.8B AI Data Center Quarter as Meta Signs 6GW Deal

AMD's data center unit pulled in $5.8 billion in Q1 2026, a 57% year-over-year increase that put the company on a trajectory to more than double its 2024 data center revenue within a single fiscal year, according to AMD's Q1 2026 earnings release. AI accelerators from the Instinct MI300 series accounted for roughly $4.2 billion of that quarter, approximately 73% of the entire data center segment, a share that has risen sharply over the past two years.

How AMD's Data Center Became a $16.6 Billion Business

Full-year 2025 marked a structural break for AMD. Data center revenue hit $16.6 billion, up 32% from an estimated $12.6 billion in FY2024, according to AMD's FY2025 earnings filing with the SEC. Total company revenue reached $34.6 billion for the year, up 34% from $25.8 billion in 2024, both figures representing all-time records.

The shift has been rapid by semiconductor standards. In Q1 2024, data center accounted for roughly 38% of AMD's total revenue. By Q1 2026, that share reached 59%, a 21-percentage-point move in two years, per CNBC's reporting on the earnings call. This was not a cyclical swing: AMD deliberately refocused capital, software investment, and engineering resources on the Instinct GPU line, while simultaneously taking server CPU share from Intel through its EPYC product line.

Lisa Su described the Q1 2026 result as "a clear inflection in our growth trajectory and a structural shift in our business," per the AMD investor relations press release. Free cash flow for the quarter reached $2.566 billion, a record quarterly figure for AMD, indicating that the revenue surge is generating real cash rather than requiring growth-funded spending to sustain it.

Who Is Buying, and What They Are Running

The most consequential customer disclosure in AMD's Q1 2026 reporting cycle involved Meta. The social media company committed to a multi-gigawatt Instinct GPU deployment, with the first gigawatt of capacity powered by custom MI450-based chips, with shipments due in the second half of 2026, per Data Center Dynamics. Meta has already moved all live Llama 405B inference traffic to the MI300X, a public signal of confidence in AMD's ROCm software stack for production inference at scale.

Oracle Cloud Infrastructure made a separate independent bet. OCI's latest Supercluster compute instances run on MI300X with ROCm, making Oracle the first major cloud provider to publish MI300X capacity as a headline offering, according to Constellation Research. Microsoft Azure has committed to MI400-series instances once the next generation ships. The customer list has expanded from a handful of early adopters to a tier-1 hyperscaler roster within 18 months of the MI300X commercial launch.

At CES 2026, OpenAI's Sam Altman appeared alongside Su for the MI400 announcement, according to Tech Insider's CES coverage, in a move that signalled OpenAI's interest in broadening its compute relationships beyond Nvidia. AMD disclosed that OpenAI is working with AMD on the MI450 architecture. The significance is not lost in the context of AI supply chain dynamics: see our Sam Altman profile for the broader picture of OpenAI's hardware strategy.

The MI400 Roadmap and AMD's Growth Targets

AMD unveiled five MI400-series accelerators at CES 2026 on January 5, positioning them as full rack-scale solutions for large-scale training and distributed inference, per Tech Insider. The MI400 line succeeds both the MI300X and MI325X, and is designed to compete directly with Nvidia's Blackwell-era products. The custom MI450 variant, the chip Meta will deploy in its first gigawatt-scale facility, represents a step toward the hyperscaler-specific customisation that Nvidia has historically offered through its own rack-scale Grace Blackwell systems.

AMD has been explicit on its financial ambitions. Management outlined a path to data center revenue growing at more than 60% annually over the next three to five years, and Su has stated AMD expects to scale AI revenue to "tens of billions" in fiscal 2027, per the Data Center Dynamics earnings summary. AMD also forecasts the server CPU market reaching $120 billion by 2030, with AMD's own share of that growing at approximately 35% per year.

Q2 2026 guidance of $11.2 billion in total revenue, cited in CNBC's Q1 coverage, would represent AMD's highest single-quarter revenue ever. If the data center revenue share holds near 59%, that implies a Q2 data center result approaching $6.6 billion, another potential record. AMD has not broken out an AI-specific sub-line within data center for the Q2 guide; the segment-level total is the only disclosed figure.

What It Means

AMD enters the second half of 2026 as something its leadership could not credibly claim two years ago: a genuine second source for hyperscale AI compute at the rack level. The financial metrics confirm the transition. Data center now represents 59% of total AMD revenue, AI accelerators are 73% of that segment, and free cash flow hit a quarterly record. The open question is whether AMD's ROCm software ecosystem can close the gap with Nvidia's entrenched CUDA platform enough to win greenfield deployments rather than complementary capacity. Meta's decision to route 100% of its Llama 405B inference traffic through MI300X is the clearest signal yet that at least one top-tier operator has concluded it already can.

Sources

• AMD Investor Relations: AMD Reports First Quarter 2026 Financial Results
• CNBC: AMD's stock soars 16% as data center growth pushes revenue and guidance past estimates
• SEC (AMD Form 8-K): AMD Q4 and Full Year 2025 Financial Results
• Data Center Dynamics: AMD posts Q1 2026 data center revenue of $5.8bn, forecasts $120bn server CPU income by 2030
• Constellation Research: With Oracle Cloud win, AMD MI300X gains traction as Nvidia counterweight
• Tech Insider: AMD MI400 Series AI GPU Challenging Nvidia (CES 2026 coverage)

Editorial standards: every claim is sourced. Tips: [email protected]