AMD

AMD designs CPUs and AI accelerators including the MI300 series, competing directly with Nvidia in data-center AI workloads. MI300X launched late 2023; Microsoft, Meta, and Oracle confirmed customers. CPU market share gains in server continue.

AMD's MI300X is the only credible alternative to Nvidia's H100 at scale, backed by major hyperscaler adoption.

amd.com AI AcceleratorsUSApublic · AMD

At a glance

Total raised

$7.5M

last round ipo

Headcount

—

no public estimate on record

Headquarters

Santa Clara

USA

Founded

1969

57 years active

Status

Public

Ticker · AMD

Patents on file

USPTO grants

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Founding

50% off

/month

91 of 100 spots left

Full

$10

/month

Available once all 100 Founding Member spots are claimed.

Get full access

Already subscribed? Sign in →

Funding history

Total raised

$7.5M

Rounds tracked

Latest stage

IPO

Date	Stage	Amount	Valuation	Lead investors
Sep 27, 1972	IPO	$7.5M	—	—

Stock performance

How they make money

Hardware (one-time)Pattern

AMD operates a hardware sales model: customers pay a one-time purchase price for CPUs and AI accelerators, with revenue recognized at shipment to OEMs, hyperscalers, and enterprise data-center operators. The company designs x86 server and client processors (EPYC, Ryzen) and GPU-based AI accelerators (MI300 series), selling directly to cloud providers such as Microsoft, Meta, and Oracle as well as through channel partners. Server CPU sales are recognized on delivery to system integrators or direct customers, while AI accelerator revenue follows a similar point-in-time model for chip purchases, though some large hyperscaler deals include multi-quarter delivery schedules that spread revenue recognition across shipments. The economics benefit from a fabless model — AMD outsources manufacturing to TSMC, converting wafer costs into variable COGS rather than fixed plant investment. Gross margins structurally improve as each new architecture (e.g., MI300X, Zen 5) packs more compute per wafer, but the one-time nature of each sale means growth depends entirely on unit volume and average selling price per generation. Software (ROCm) is given away free to drive hardware attach, not monetized directly.

Generated by deepseek-v4-flash · June 2, 2026Analytical summary · not investment advice

Founder backgrounds

Jerry Sanders

Co-founder (former CEO)

Graduated with a B.S. in electrical engineering from the University of Illinois Urbana-Champaign in 1958. Began his career at Douglas Aircraft, then moved to Motorola before joining Fairchild Semiconductor in 1961, where he rose to head worldwide sales and marketing. When new Fairchild management clashed with his style, he was dismissed in 1968. In 1969, Sanders co-founded Advanced Micro Devices in Sunnyvale alongside seven other engineers, serving as CEO for 33 years. He is credited with the 1982 licensing deal that let AMD clone Intel's x86 processors, and the 1999 Athlon CPU that first challenged Intel at the high end of the market.

AI bull / bear

Bull case

AMD has locked in the two most critical ingredients for winning in AI infrastructure: a credible multi-generational roadmap and hyperscaler conviction. The MI300X is already deployed at Microsoft, Meta, and Oracle. Now Oracle is first in line for the Altair MI450 and Helios racks, while Supermicro is building NVL72-class systems around those same MI450 accelerators. That is OEM and end-customer alignment Nvidia used to own exclusively. The MI400 series and Helios are on track for H2 2026, giving enterprise planners a visible upgrade path. AMD's EPYC CPU momentum in servers also means it can pitch complete GPU-plus-CPU racks, not just discrete chips. The Helios system-level play — combining MI400-class GPUs, EPYC, and unified networking — directly targets the Nvidia DGX rack model that hyperscalers have been forced to adopt. If AMD delivers on its MI400 roadmap and keeps closing the ROCm software gap, it captures a durable 20-30% share of the $100B+ data-center AI market.

Bear case

AMD's MI300X shipped in late 2023, but Nvidia has already moved to Blackwell and is shipping Rubin systems by 2026 — exactly when AMD's MI400 series arrives. AMD is perpetually a generation behind on performance and a decade behind on software. ROCm still cannot match CUDA's library depth, debugging tools, and enterprise support, which means every hyperscaler deployment requires painful internal engineering workarounds. The MI450 and Helios look strong in press releases, but Nvidia's GB200 NVL72 delivers real benchmark results today, not promises. AMD also faces a structural disadvantage on supply: TSMC's CoWoS advanced packaging capacity is constrained, and Nvidia has the purchasing power to lock up the majority of it. Even with Oracle and Supermicro on board, AMD has not demonstrated it can scale to match Nvidia's shipment volumes. If the MI400 slips even one quarter, Nvidia will have another full generation of dominance before AMD's alternative even reaches the rack.

Generated by deepseek-v4-flash · June 2, 2026Refreshes quarterly · see disclosures

In the news

Watch

Leadership

Jerry Sanders

Co-founder (former CEO)

—

Lisa Su

Chair & CEO

𝕏

Jean Hu

EVP, CFO & Treasurer

—

Mark Papermaster

EVP & CTO

in 𝕏

Products

Instinct MI300X / MI325X / MI350 Series
Data-center AI accelerators built on the CDNA architecture and targeting Nvidia's H100/H200 in training and inference. The MI300X pairs CDNA 3 compute with up to 192GB of HBM3 memory in a chiplet package; the MI325X and 2025-era MI350 series extend memory capacity and bandwidth. Adopted by Microsoft Azure, Meta, and Oracle, the line is AMD's primary entry in the hyperscale AI-compute market and anchors its ROCm software stack.
EPYC Server Processors
Data-center x86 CPUs ('Zen' architecture) that drove AMD's server-market-share gains against Intel Xeon. The latest 'Turin' generation scales to 192 cores per socket, delivering high core density, memory bandwidth, and energy efficiency for cloud, virtualization, and AI-host workloads. EPYC powers a large share of hyperscaler fleets and frequently serves as the head node feeding AMD Instinct and Nvidia GPU clusters.
Ryzen Processors (incl. Ryzen AI)
Consumer and commercial CPUs and APUs for desktops and laptops, built on 'Zen' cores. The Ryzen AI mobile line integrates an XDNA neural processing unit (NPU) on-die for on-device generative-AI and Copilot+ PC workloads, alongside Radeon integrated graphics. Ryzen is AMD's flagship client-compute brand and a core driver of its PC-market position.
Versal Adaptive SoCs (Xilinx)
Adaptive system-on-chips combining FPGA programmable logic, Arm CPUs, and AI engines on a single device, originating from AMD's 2022 Xilinx acquisition. Versal targets edge AI, communications infrastructure, aerospace and defense, and automotive, where reconfigurable hardware and deterministic latency matter. It anchors AMD's adaptive-computing and embedded portfolio.

Most recent patents

13 patents on file, but none with both an extractable figure and an abstract on Google Patents yet.

AMD

At a glance

Subscribe to see the full profile

Funding history

Stock performance

How they make money

Founder backgrounds

Jerry Sanders

AI bull / bear

In the news

AMD ‘had to re-engineer’ the Ryzen 7 5800X3D for a re-release — 10th Anniversary Edition chip had ‘a whole body of engineering work’ put into it

Chip Industry Technical Paper Roundup: Jun. 2

AMD Radeon RX 9070 GRE review: thoroughly midrange

Supermicro shows off Vera Rubin NVL72 rack with all-new type of coolant — company claims coolant offers 1,000 times higher electrical impedance over standard cooling

AMD promises 13% uplift with new EXPO ‘Ultra Low Latency’ overclocking on DDR5 DIMMs — automatic memory overclocking delivers 4% improvement over standard EXPO, says AMD

Junctionless Transistors Show a New Path to 3D Chips

Watch

Leadership

Products

Instinct MI300X / MI325X / MI350 Series

EPYC Server Processors

Ryzen Processors (incl. Ryzen AI)

Versal Adaptive SoCs (Xilinx)

Most recent patents

Compare AMD with