AMD
Lisa Su (Chair, President & CEO)
Data verified as of Sep 18, 2026
Summary
AMD, founded in 1969 by Jerry Sanders and seven Fairchild Semiconductor colleagues, is a fabless designer of CPUs, GPUs, FPGAs, DPUs and rack-scale AI systems, with production outsourced mainly to TSMC. Under Lisa Su, CEO since 2014, it rebuilt x86 share with Ryzen and EPYC, bought ATI (2006), Xilinx (2022) and Pensando (2022), and is now the clearest alternative to Nvidia in data-center AI. Fiscal 2025 revenue was a record $34.6 billion (+34%), with Data Center at $16.6 billion. The mix then flipped: Q2 2026 revenue hit $11.5 billion (+50% YoY), Data Center $6.7 billion (+107%) and 58% of the company, and management guided Q3 to about $13 billion. Instinct MI350-series GPUs are the current AI volume product; Helios, AMD's first rack-scale AI system (72 Instinct MI455X GPUs, 18 EPYC Venice CPUs, Pensando networking, ROCm), entered full production on July 23, 2026, with volume shipments targeted for late Q3 into Q4. Named Helios customers include OpenAI, Anthropic, Meta, Microsoft, Oracle, HUMAIN, Tensorwave, Vultr and Cirrascale. Multi-year GPU commitments now include up to 6 GW with OpenAI (first 1 GW of MI450 in H2 2026, plus a warrant for up to 160 million AMD shares at $0.01), up to 6 GW with Meta including a custom MI450-based GPU, and up to 2 GW with Anthropic in Helios from H1 2027 alongside an AMD commitment to invest up to $5 billion in Anthropic. Oracle is set to offer the first public Helios supercluster, 50,000 MI450-series GPUs, from calendar Q3 2026. Client Ryzen remains a share-gaining franchise; gaming revenue fell 31% in Q2 on lower semi-custom console SoCs and higher board prices. The stock closed September 17, 2026 at $545.09, a market cap of about $890 billion. The open question is whether Helios and MI400 actually ship at gigawatt scale in late 2026 and 2027, against Nvidia's already-shipping Vera Rubin platform and custom silicon from Google, Amazon and Broadcom.
Company OverviewRead the long-form profile of AMDMain Products
AMD's first rack-scale AI system: 72 Instinct MI455X GPUs in 18 Open Rack Wide compute trays, 18 6th Gen EPYC Venice CPUs, Pensando Vulcano 800 AI NICs, UALink-over-Ethernet scale-up and Ethernet scale-out, running ROCm. A rack is specified at up to 2.9 exaflops FP4, 1.4 exaflops FP8, 31 TB of HBM4 and 1.7 PB/s of memory bandwidth. AMD claims up to 30% more inference tokens per dollar, 15% more peak FP4, and 50% more HBM than Nvidia's Vera Rubin NVL72.
Declared in full production on July 23, 2026; volume shipments targeted for late Q3 into Q4 2026. Named deployers include OpenAI, Anthropic, Meta, Microsoft, Oracle, HUMAIN, Tensorwave, Vultr and Cirrascale. OEM/ODM partners include HPE, Lenovo, Supermicro, Bull, Sanmina and Wiwynn.
AMD's data-center GPU line. MI350-series (CDNA 4), led by MI355X (288 GB HBM3E, 8 TB/s, up to 10.1 PFLOPS MXFP4, 1,400 W TBP), is the generation shipping into 2026 and the volume driver of Instinct revenue. MI400-series (CDNA 5), launched July 23, 2026, splits into MI455X for frontier AI and AI factories (432 GB HBM4, 23.3 TB/s, up to 40 PFLOPS FP4, 320 billion transistors) and MI430X for HPC and sovereign AI (up to 288 TFLOPS FP64). ROCm and ROCm.ai are the software stack.
MI355X is in production and live at customers including HUMAIN (Aug 31, 2026) and in AMD's MLPerf Inference 6.1 submission (Sep 16, 2026). MI400/MI455X ramps with Helios from late Q3 2026. MI500 series is on the 2027 roadmap.
6th Generation EPYC 9006-series server CPUs on Zen 6, launched at Advancing AI 2026 as the host CPU for Helios and as a general-purpose and agentic-AI server line. Top SKU EPYC 9996 is specified at 256 cores / 512 threads. AMD positions them as the densest x86 CPU for agentic workloads and the CPU that keeps Instinct GPUs fed. Meta is a lead customer; Azure is adding Venice-powered HDv2 and HXv2 VM series.
Launched July 23, 2026. Ramping with Helios in H2 2026. Next EPYC generations on the public roadmap are Zen 7 (Florence / Ferrara / Fidenza) in 2028 and Zen 8 Ravenna in 2030.
AMD's client CPU and APU brand, from desktop AM5 chips through Ryzen AI notebooks and Ryzen AI Halo agentic PCs. Client revenue was a record $10.6 billion in 2025 and $3.1 billion in Q2 2026. Recent additions include Ryzen AI PRO 400 Series, Ryzen AI Halo developer systems, 3D V-Cache on Ryzen PRO workstations, and a commitment to support Socket AM5 through 2029.
Shipping. Q2 2026 client revenue $3.062B (+23% YoY). Ryzen AI Halo systems and expanded Ryzen PRO 9000 SKUs were called out in the Q2 highlights; AM5 support extended through 2029.
RDNA 4 desktop GPUs launched in 2025 as RX 9070 XT ($599 SEP) and RX 9070 ($549 SEP), later joined by RX 9070 GRE and RX 9060 XT. FY2025 gaming revenue was $3.9 billion, but Q2 2026 gaming fell to $779 million (-31% YoY) on lower semi-custom console SoC sales and higher board prices that Lisa Su said weighed on demand.
Shipping. Gaming is a shrinking slice of AMD: $779 million in Q2 2026 versus $3.1 billion of client CPUs in the same quarter. RX 9070 GRE global availability was expanded in the Q2 2026 highlights.
What's Next
Ship Helios volume in late Q3 and Q4 2026
Lisa Su said at Advancing AI 2026 that Helios is in full production, with shipments starting at the end of Q3 and ramping through Q4. Q3 guidance of about $13 billion is the first earnings print that can show whether rack-scale revenue is real. Watch named customer deliveries versus reference-design status.
Bring OpenAI's first Helios gigawatt online
OpenAI has a binding commitment for the first 1 GW of MI450-series GPUs, the first tranche of a 6 GW multi-generation deal. At AAI 2026 it said Helios would come online through deployment partners beginning in Q4 2026, accelerating through 2027. The first warrant tranche vests with that 1 GW delivery.
Report Q3 2026 results against the $13 billion guide
On August 4 AMD guided Q3 revenue to about $13 billion, plus or minus $300 million (~41% YoY, ~13% QoQ) and non-GAAP gross margin of about 56%. Consensus around the print (Zacks estimates November 3, 2026) will be the first look at whether Helios is in the numbers.
Start Anthropic's first gigawatt of Helios
Anthropic will deploy up to 2 GW of MI450-series GPUs in Helios, with the first gigawatt beginning in H1 2027. AMD's up-to-$5 billion equity investment is tied to that compute partnership. Watch whether the first GW date holds as Anthropic's other Nvidia, Google and Amazon capacity comes online.
Track Record
4 of 4 announced dates hit
Operations & Revenue
AMD reported Q2 2026 revenue of $11.536 billion on August 4, 2026, up 50% year over year, with Data Center at $6.718 billion (+107%) and 58% of sales. Instinct MI350-series GPUs and EPYC CPUs drove that print. Helios, combining 72 MI455X GPUs with EPYC Venice CPUs and Pensando networking, was declared in full production at Advancing AI 2026 on July 23, with Lisa Su saying shipments start at the end of Q3 and ramp through Q4. OpenAI expects Helios online in Q4 2026 as the first slice of its 6 GW deal; Meta is validating Helios for a custom MI450 GPU and up to 6 GW; Anthropic's 2 GW starts in H1 2027; Oracle's 50,000-GPU public cluster is dated to Q3 2026; Microsoft will deploy Helios on Azure for frontier inference. Gaming remains the soft spot ($779 million, -31% YoY). Q3 guidance of about $13 billion is the near-term test of whether Helios revenue shows up before year-end. Shares closed September 17, 2026 at $545.09 (~$890B).
Revenue Streams
Data Center (Instinct, EPYC, networking, FPGAs)
AI accelerators, server CPUs, DPUs/AI NICs, and adaptive SoCs for hyperscalers, AI clouds, enterprises and HPC. $16.635B in FY2025 (+32%) and $6.718B in Q2 2026 (+107%), now 58% of company revenue.
Desktop and notebook CPUs and APUs, including Ryzen AI and Ryzen AI Halo agentic-PC platforms. Client revenue was $10.640B in FY2025 (+51%) and $3.062B in Q2 2026 (+23%).
Key Metrics
- Employees
- 31,000 (as of Dec 27, 2025)
- Est. Annual Revenue
- $34.639B (FY2025, ended Dec 27, 2025, +34% YoY; GAAP net income $4.335B). H1 2026 added $21.789B. Q3 2026 guided to ~$13B plus or minus $300 million
- Latest quarterly revenue
- $11.536B (Q2 2026, ended Jun 27, 2026, +50% YoY, +13% QoQ); Q3 guided to ~$13B plus or minus $300 million
- Announced GPU deployments
- Up to 6 GW OpenAI + up to 6 GW Meta + up to 2 GW Anthropic, plus Oracle's 50,000-GPU Helios cluster and HUMAIN's path to 1 GW by 2030
Timeline
On February 3, 2026, AMD reports 2025 revenue of $34.639 billion, GAAP net income of $4.335 billion and non-GAAP EPS of $4.17. Data Center is $16.6 billion (+32%), Client and Gaming $14.6 billion (+51%). Results include about $440 million of MI308 China export-control inventory charges.
On February 24, 2026, AMD and Meta announce plans to deploy up to 6 gigawatts of Instinct GPUs, with the first 1 GW powered by a custom MI450-based GPU. Meta is also a lead customer for 6th Gen EPYC Venice and Verano. AMD's later 10-Q says it issued Meta a warrant for up to 160 million shares on the same $0.01, milestone-vesting structure as OpenAI.
On July 22, 2026, AMD and Anthropic agree that Anthropic will deploy up to 2 gigawatts of Instinct MI450-series GPUs in Helios racks, with the first gigawatt beginning in H1 2027, building on existing MI355X use. AMD commits to a strategic equity investment of up to $5 billion in Anthropic, and the companies will use Claude to optimize ROCm while AMD adopts Claude internally.
On July 23, 2026, AMD puts Helios in full production and launches Instinct MI400-series GPUs (MI455X for frontier AI, MI430X for HPC and sovereign AI), 6th Gen EPYC Venice CPUs, ROCm.ai, Ryzen AI Embedded X100 and the Kria AI robotics platform. Helios is claimed to deliver up to 30% more inference tokens per dollar than Nvidia's Vera Rubin NVL72. OpenAI says it expects to bring Helios online in Q4 2026.
On August 4, 2026, AMD reports Q2 revenue of $11.536 billion (+50% YoY), GAAP EPS $1.38 and non-GAAP EPS $1.66. Data Center is $6.718 billion (+107%) and 58% of revenue. Client is $3.062 billion (+23%); gaming $779 million (-31%). Q3 is guided to about $13 billion, plus or minus $300 million, with non-GAAP gross margin about 56%.
On August 6, 2026, AMD announces a definitive agreement to buy Toronto-based Taalas, a 2023-founded inference-chip startup, to fold specialized inference dataflow hardware into Instinct and Helios. The deal is subject to customary closing conditions.
On August 31, 2026, AMD, Cisco and HUMAIN say Instinct MI355X systems with EPYC CPUs and Cisco Silicon One networking are in production in the Kingdom. The JV still targets up to 250 MW of MI400-series infrastructure from 2027 and up to 1 GW by 2030.
On October 6, 2025, AMD and OpenAI announce a multi-year, multi-generation deal for 6 gigawatts of Instinct GPUs, with the first 1 GW of MI450 series starting in H2 2026. AMD issues OpenAI a warrant for up to 160 million shares at $0.01, vesting on purchase and stock-price milestones up to $600 a share. CFO Jean Hu says the partnership is expected to deliver tens of billions of dollars in revenue.
