The open, owner-operated compute stack.
Worked backwards from AMD's real business — a data-center franchise now more than half its revenue, built on the ROCm software stack that answers CUDA's walled garden. Of every silicon vendor, AMD's open-software strategy and on-device Ryzen AI NPUs match the 🤫 thesis most closely: EPYC and Instinct for the AI-factory grid, Ryzen AI for the on-device agent, and ROCm as the open, inspectable, owner-controlled software layer. You build the open 'own your compute' hardware; we operate it, owner by owner, consent-first.
Written from public information, by AMD admirers. A proposal in pursuit — not a claim of partnership, endorsement, or investment by AMD.
Why AMD × 🤫 is obvious — in one look.
ROCm is your open answer to CUDA's walled garden, and Ryzen AI runs large local models on 128GB of unified memory.
We build Agent One's inference on ROCm and run it on Ryzen AI — owner-operated, open, no lock-in.
The on-device agent that proves ROCm + Ryzen AI beat the closed stack.
Data Center is now more than half your revenue; EPYC + Instinct are the growth engine.
Our garage and warehouse AI-factory grid is a distributed EPYC + Instinct buyer, running on ROCm.
A credible non-NVIDIA supply path with committed, distributed volume.
You already co-invest via aligned-incentive structures — the OpenAI warrant is the public template.
A milestone-based hardware-commitment and co-marketing arrangement for the distributed-grid rollout.
Precedented, honest alignment — grow the open-compute grid together.
How AMD wins, honestly.
AMD reports four segments — Data Center, Client, Gaming, Embedded — and the story is the data-center AI ramp (EPYC + Instinct) crossing half the company. Every figure is public (AMD FY2025 & Q1 2026 filings) and approximate.
Up 34% YoY, with ~$6.8B non-GAAP net income and ~52% non-GAAP gross margin.
Up 57% YoY — the first quarter Data Center exceeded half of AMD's total revenue.
EPYC reached a record ~46% of server-CPU revenue — the clear share-taker in the data center.
The XDNA-2 NPU, above the Copilot+ bar; Ryzen AI Max pairs it with up to 128GB unified memory to run large local models.
AMD's open-source (LLVM-based) answer to CUDA — framework-native, now across Windows and Linux.
A strategic partnership to deploy 6 GW of AMD GPUs — with a warrant aligning incentives, the template for co-investment.
EPYC, Instinct, Ryzen AI — and an open stack.
AMD's differentiation is credible data-center silicon plus an open-software ethos. These are the strengths a partnership builds on.
EPYC server CPUs
A record ~46% of server-CPU revenue — the workhorse of the modern data center and the CPU tier of any compute grid.
Instinct AI GPUs
The credible #2 to NVIDIA — MI350 shipping, MI450 + the Helios rack coming in 2H 2026 — anchored by OpenAI and (reported) Meta.
Ryzen AI (on-device)
50-TOPS XDNA-2 NPUs and Ryzen AI Max with up to 128GB unified memory — real hardware that runs large models on the machine the person owns.
ROCm — the open stack
Open-source, LLVM-based, framework-native (PyTorch/vLLM/JAX), with a CUDA-portability bridge — the antithesis of lock-in.
Chiplet leadership
AMD pioneered chiplet packaging (Infinity Fabric) — the cost and scaling advantage behind EPYC and the next-gen Venice redesign.
Xilinx / Embedded
The FPGA / adaptive-SoC franchise — flexible edge and embedded compute for specialized nodes.
Instinct MI350, Helios, and Ryzen AI Max.
What AMD shipped or announced most recently — the surfaces a partnership plugs into. Stated honestly, shipped vs. announced.
Instinct MI350 series
Shipping since mid-2025 — CDNA 4, 288GB HBM3E, the Q1-2026 Data Center growth driver. MI450 is sampling for 2H 2026.
Helios rack (2H 2026)
Announced — 72 MI455X accelerators + EPYC Venice CPUs + Pensando NICs, unified by ROCm: the productized 'supercomputing you own and operate.'
Ryzen AI Max ('Strix Halo')
16-core Zen 5 + 50-TOPS NPU + up to 128GB unified memory — runs quantized 100B-parameter-class models locally; the Ryzen AI Halo desktop undercuts rival dev boxes.
ROCm 7 / 7.2
Native MI350 support, FP4/FP8, day-zero PyTorch/vLLM, and expanded Windows + Linux support — the open software cadence accelerating.
What AMD is building toward.
Scale the data-center AI franchise
Lisa Su: Data Center is 'now the primary driver'; AMD guides Data Center growth >60% annually over 3–5 years.
Anchor mega-customers
The OpenAI 6 GW partnership (with a warrant) and a reported Meta 6 GW agreement — Instinct at hyperscale.
Open software (ROCm)
Close the CUDA gap with open, framework-native, Windows+Linux ROCm — openness and cost as the strategy.
On-device AI (Ryzen AI)
'AI everywhere, for everyone' — Copilot+-class NPUs across the client line and developer-focused Ryzen AI.
Rack-scale systems
Helios — moving from selling chips to selling integrated racks, competing at the AI-factory level.
Chiplet & process leadership
EPYC Venice on TSMC 2nm with up to 256 cores — the next-gen data-center CPU.
The people carrying it.
Our respectful reading of public roles and priorities — so any partnership makes each of their jobs easier, not harder.
Dr. Lisa Su
Chair & CEO
The architect of AMD's EPYC/Instinct turnaround — scale the data-center AI franchise on an open stack.
Jean Hu
EVP, CFO & Treasurer
Financial strategy and the disciplined capital allocation behind the AI ramp.
Mark Papermaster
EVP & CTO
The Zen/CDNA roadmap — Venice, MI450, and the chiplet and rack-scale future.
Forrest Norrod
EVP, Data Center Solutions
The EPYC + Instinct data-center P&L — AMD's growth engine.
Buy together. Build together. Sell together. Take risk together.
AMD silicon maps cleanly onto both halves of the 🤫 north star — the distributed AI-factory grid and the on-device agent — and its open ROCm ethos is the closest silicon-vendor match to our consent-first, owner-operated stance.
Build together — the factory-grid BOM
EPYC Turin/Venice CPUs + Instinct MI350/MI450 + Helios racks, unified by ROCm.
Standardize the 🤫 Warehouse / Factory One BOM on this stack — a credible non-NVIDIA supply path when the GPU grid is sold out.
AMD gets a distributed, owner-operated buyer; the human runs an open, owned AI factory.
Build together — the on-device agent on Ryzen AI
Ryzen AI Max (128GB unified, 50-TOPS NPU) runs large models locally.
Ryzen AI as the reference silicon for 🤫 Agent One running on the machine the person owns — built on ROCm, not a proprietary lock-in.
AMD proves on-device AI; the person owns their AI, data, and compute.
Sell together — ROCm-native reference systems
AMD pitches a cost/TCO advantage and an open stack.
Position Puppy One nodes as ROCm-native AMD reference systems; co-sell into edge + telecom communities seeking the lowest cost per watt.
AMD gets edge volume and an open-stack showcase; we get a credible, co-marketed spine.
Take risk together — aligned-incentive co-investment
AMD demonstrably co-invests via structures like the OpenAI warrant.
A milestone-based hardware-commitment and co-marketing arrangement for the distributed-grid rollout.
A precedented, honest way to grow the open-compute grid together.
Open compute, owned by the communities that run it.
AMD's open ROCm + open-rack posture is the closest silicon-vendor match to the 🤫 consent-first, owner-operated, open-compute stance — the honest technical spine for 'Own your AI. Own your data. Own your compute.' And 🤫 Agent One is free for life to every US federal and state government employee and every past or present service member who helps build the grid.
Buy together. Build together. Sell together.
The best business is fun when there is real money to be made and real value created for the people we both serve. That is the only partnership we will build — and we would be proud to build it with AMD.
What this is, and isn't.
This is our proposal, assembled from public information — not a claim of partnership, endorsement, sponsorship, or investment by AMD. AMD is not currently a partner of HushOne or Hushh Technologies. Market-share and per-token cost figures cited by third parties are estimates, not AMD-reported; MI450 / Helios / EPYC Venice are described as announced/sampling for 2H 2026, not shipping. AMD, EPYC, Instinct, Ryzen, ROCm, and related marks are trademarks of Advanced Micro Devices, Inc. Our own certifications are 'in pursuit' until a 3PAO assessment or agency ATO says otherwise.