AMD's data center division posted $6.7 billion in revenue for Q2 2026, a 107% jump from a year earlier, and CEO Lisa Su projects the segment will more than double again by 2027. AI inference demand, which AMD expects to overtake training workloads for the first time this year, drives the forecast — but shares still fell 8-9% after-hours on supply chain worries.
AMD's data center division generated $6.7 billion in revenue during Q2 2026, up 107% from the same period a year earlier. Total company revenue reached $11.54 billion, up 50% year-over-year, with data center sales now making up 58% of everything AMD sells.
CEO Lisa Su isn't stopping there. She projects data center sales will exceed $13 billion in Q3 2026, with the segment more than doubling again by 2027. AI inference workloads sit behind that forecast — AMD expects them to surpass training demand for the first time this year.
Inference reshapes the AMD pitch
AMD's server CPU revenue grew more than 70% year-over-year in Q2 2026, and its Instinct GPU lineup is gaining traction with hyperscalers seeking cost-effective inference options alongside Nvidia's chips. Inference workloads tend to be more price-sensitive than training, which opens room for AMD's value proposition.
New silicon and marquee customers
AMD is building its product roadmap around that shift. The Instinct MI350, MI400, and MI450 GPUs are all in various stages of development and deployment, and the MI400 series is expected to ramp volume in late 2026 into 2027. On the CPU side, the EPYC "Venice" and "Verano" processors are designed to pair with those GPUs in AI-optimized server setups, and the company also unveiled its Helios rack-scale platform for AI inference and training at data center scale.
The customer list backs up the ambition. Meta signed a multi-year GPU deal with AMD covering up to 6 gigawatts of compute capacity. Anthropic committed to 2 gigawatts of MI450 systems beginning in 2027.
Wall Street weighs the risk
Despite the numbers, AMD shares dropped 8-9% in after-hours trading following the earnings release. The concern appears to center on supply chain constraints: doubling data center revenue requires doubling production capacity, and AMD relies on TSMC for its most advanced chip manufacturing. Any bottleneck in wafer allocation, packaging, or high-bandwidth memory could open a gap between what AMD promises and what it delivers.
Nvidia still commands the lion's share of the AI accelerator market, and its upcoming Blackwell and Rubin architectures are designed to keep that lead. AMD isn't trying to dethrone Nvidia so much as position itself as the essential second supplier — a role hyperscalers actively want filled to avoid depending on a single vendor.
Lisa Su's roadmap has data center revenue more than doubling in 2027, with the Meta and Anthropic deals providing contractual backing. Converting those commitments into shipped silicon at scale is where chipmakers succeed or stumble.
Source: Crypto Briefing
Trading involves risk.
Risk Disclaimer
