AMD Posts $11.54 Billion Quarter, Lifts Datacenter Forecast on Helios Ramp
AMD reported record second-quarter revenue of $11.54 billion, up 50.1 percent, with datacenter sales more than doubling to $6.72 billion as Helios systems ramp with Meta and Microsoft.
- By
- Grace Kim
- Filed
- Channel
- AI & Compute
- Read
- 3 min read
What did AMD report for the second quarter?
AMD posted record second-quarter revenue of $11.54 billion, up 50.1 percent year over year. Operating income reached $1.99 billion, a sharp reversal from a $134 million operating loss a year earlier. Net income climbed by a factor of 2.6 to just under $3 billion, supported by mark-to-market gains on AMD's investment portfolio that lifted reported earnings above operating income for the period. AMD exited the quarter with $13.11 billion in cash and investments.
How fast is the datacenter business growing?
The Data Center segment generated $6.72 billion in revenue, up 107.3 percent year over year and 16.3 percent sequentially on continued Epyc adoption. Segment operating income hit $2.1 billion at a 31.3 percent margin, against a $155 million operating loss in the year-ago period. AMD does not break out Epyc CPU and Instinct GPU revenue separately, but third-party analyst estimates based on AMD's guidance place Epyc CPU sales at $3.35 billion (up 74 percent) and Instinct GPU sales at just over $3 billion (up 2.6 times).
When will Helios catch Nvidia's rackscale systems?
AMD expects its Helios rackscale systems, co-developed with Meta Platforms and Microsoft, to reach feature parity with Nvidia's "Oberon" rackscale systems by late 2026. Helios pairs AMD's MI455X "Altair" GPUs with "Verano" CPUs, a high-I/O mid-core variant of the sixth-generation Epyc "Venice" processors. AMD claims Helios carries 50 percent more HBM capacity than Nvidia's "Vera-Rubin" Oberon racks, delivers 15 percent more AI flops, and posts 30 percent better cost per token per watt on early benchmarks.
What did CEO Lisa Su commit to on supply and growth?
Chief executive officer Lisa Su said the Epyc CPU business will grow more than 80 percent in the second half of 2026 and more than 70 percent in 2027 against a tough compare year. She added that the overall datacenter business will more than double in 2027. Su said AMD has secured wafer, substrate, interposer, and HBM capacity to meet those commitments, with modest spare capacity for upside demand. The datacenter AI accelerator business, which now includes Helios rack components sold to ODMs and OEMs, will accelerate further in 2027 after growing in the second half of 2026.
How large is AMD's addressable market?
AMD's own market sizing projects the datacenter AI GPU/XPU TAM at $360 billion in 2026, rising to $630 billion in 2027, with growth cooling sharply after 2027. The datacenter CPU TAM is expected to double to $53 billion in 2026 and grow another 81.1 percent to $96 billion in 2027. Neither market declines through 2030 according to AMD's prognostications.
What is happening outside the datacenter?
PC and gaming chip revenue is under pressure as component prices rise and customers resist paying more for machines with less DRAM and flash storage. The embedded segment, built primarily on Xilinx FPGAs, remains AMD's most profitable unit at roughly 38 percent operating margins over the past two years, offsetting pressure on consumer chips. Analyst estimates put Pensando DPUs and AI NICs at roughly $275 million in revenue (nearly triple year over year) and datacenter FPGAs at about $95 million (up 1.7 times).
How concentrated is AMD's Epyc customer base?
Hyperscalers and cloud builders accounted for 73 percent of Epyc CPU sales in the quarter, up 75 percent year over year. Enterprises, telcos, academia, and sovereigns contributed $905 million in Epyc sales, up 71.4 percent. AMD is outpacing the GPU/XPU TAM through 2027 as Helios deployments scale. The competitive fight against Nvidia and the Arm collective will increasingly hinge on price and price-performance once both vendors reach feature parity in late 2026.
Source: The Next Platform
More from Grace Kim
Show full bio
Market editor covering people, launches and day-to-day practice in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
365 articles
Related articles
nvidia-commits-279-billion-to-suppliers-as-demand-outruns-capacity-392a136b
Nvidia Commits $279 Billion to Suppliers as Demand Outruns Capacity
intel-s-datacenter-sales-climb-59-on-ai-host-demand-129725fe
Intel's Datacenter Sales Climb 59% on AI Host Demand
raspberry-pi-h1-revenue-jumps-90-to-256-9m-on-memory-stockpile-5961ced7
Raspberry Pi H1 Revenue Jumps 90% to $256.9M on Memory Stockpile
amd-crosses-1-trillion-market-cap-on-ai-chip-demand-729596cc
AMD Crosses $1 Trillion Market Cap on AI Chip Demand


