We’re onsite at @OpenComputePrj’s 2026 Global Summit!

@LisaSu is taking the keynote stage to share how AMD is building the next era of AI infrastructure. Follow our thread for live highlights from her #OCPSummit26 keynote. 🧵

Oct 12, 2026 · 11:20 PM UTC

37
43
438
37,563
Replying to @AMD
AI is helping solve some of the world's most important challenges. AMD is accelerating AI from silicon to solutions, powering innovation across industries and workloads.
2
1
52
4,669
Replying to @AMD
For 15 years, AMD and the OCP community have helped advance open infrastructure together.

The next era of AI will be built the same way: through open collaboration and ecosystem
innovation. #OCPSummit26
2
5
43
3,337
Replying to @AMD
AI demand is accelerating fast. The world now consumes more than 40 quadrillion inference tokens per month, up 95x in just two years. As inference and agentic AI scale, demand is growing across both CPUs and GPUs.

#OCPSummit26
1
0
13
1,878
Replying to @AMD
Open hardware. Open networking. Open software.

Co-engineered with @Meta and the OCP community, AMD Helios brings together more than 20 open components, high-performance networking and ROCm software to scale AI.
2
3
16
1,682
Replying to @AMD
@Meta Open architecture. Leadership performance.

AMD Helios delivers up to 30% more tokens/$ on leading AI models.

And, AMD Helios delivers leadership throughput on DeepSeekR1, with 17,000 tokens/second/GPU.
2
2
23
1,460
Replying to @AMD
@Meta AMD Helios is on the move with customer deliveries to @OpenAI, @Meta, @AnthropicAI, @Microsoft, and @OracleCloud. We’re going behind the scenes with the teams building, testing and delivering our rack-scale AI systems.
2
1
18
1,388
Replying to @AMD
@Meta @OpenAI @AnthropicAI @Microsoft @OracleCloud From model builders to hyperscalers, the AMD ecosystem continues to grow.

Helios is moving beyond architecture into real-world deployment across the leaders shaping the future of AI.

Together, we’re advancing AI on open, rack-scale infrastructure.
#OCPSummit26
1
2
18
1,717
Replying to @AMD
@Meta @OpenAI @AnthropicAI @Microsoft @OracleCloud Introducing AMD Helion: a new flexible, open rack-scale architecture designed for the agentic era. Built to support diverse AI workloads across hyperscale and enterprise environments.
3
7
31
3,931
Replying to @AMD
@Meta @OpenAI @AnthropicAI @Microsoft @OracleCloud AMD Helion will support CPU-only hyperscale configurations and all-in-one enterprise configurations with AMD EPYC CPUs + AMD Instinct MI355X GPUs.

EPYC “Venice” racks are available today; the new Agentic AI Rack is planned for 1H 2027.
1
4
12
1,647
Replying to @AMD
@Meta @OpenAI @AnthropicAI @Microsoft @OracleCloud Open innovation means building with the community and contributing back. AMD is donating the Helion Agentic Rack to OCP to help advance the future of open AI infrastructure. #OCPSummit26
2
2
14
2,546
Replying to @AMD
It has become clear: AI needs more compute and more from the CPU.

AMD EPYC “Venice” spans GPU hosts, agent sandboxes and general-purpose compute.

Now, Dr. Lisa Su just revealed 6th Gen AMD EPYC "Verano," the next step in AMD's CPU roadmap for the AI era.
1
11
101
25,204
Replying to @AMD
That's a wrap on the AMD #OCPSummit26 keynote.

From open innovation and rack-scale AI to Helios momentum, AMD Helion and the future of AI infrastructure, read more all of today's announcements on the AMD Newsroom: newsroom.amd.com/news/ocp-2026-…
1
2
23
6,298
RepliesRelevant
From the VA experts: get your own Grok Bot AI assistant, set up for you in one go and supported for as long as you need us. Use it to back up your VA or to take the busywork off your plate. Book a 15-min call: adminops.com/grok-bot-setup
0
0
1
10,689
Replying to @AMD
@AMD @OpenComputePrj @LisaSu WHERE IS ZEN 6?
0
0
0
97
Replying to @AMD
@AMD @OpenComputePrj @LisaSu FFS no one cares about fake AI and the ridiculous uses by buying up all the ram and gpus. Actual AI won't exist u til we move away from the silicon process, wasting everything to chase fake AI.
0
0
0
148
Replying to @AMD
@AMD @OpenComputePrj @LisaSu Meaning, They are going to follow Nvidia and not caring about gamers. I hope if or when AI crashes, everyone remembers. These companies Before, Today, and in the future, they don't care about anyone but the shareholders. Thank Dodge.
0
0
0
149
Replying to @AMD
@AMD @OpenComputePrj @LisaSu verano is the one i'm watching tbh
0
0
0
87
Replying to @AMD
@AMD @OpenComputePrj @LisaSu 🇺🇸👏👏👏👏👏
0
0
0
50
Replying to @AMD
@AMD @OpenComputePrj @LisaSu interested to see how the interconnect scales past previous generations
0
0
0
109
Replying to @AMD
@AMD @OpenComputePrj @LisaSu Lisa Su on the OCP keynote stage again. AMD keeps showing up where the rack-scale crowd actually listens.
0
0
0
61
Replying to @AMD
@AMD @OpenComputePrj @LisaSu 리사수 눈나!!!
0
0
0
9
Replying to @AMD
0
0
0
35
Replying to @AMD
@AMD @OpenComputePrj @LisaSu Lisa Su 又主讲 OCP 峰会了,我去年追直播追到凌晨三点,就为看她讲 MI 系列的路线图。今年这波 AI 基础设施估计又要放猛料。
0
0
0
60
Replying to @AMD
@AMD @OpenComputePrj @LisaSu We don't care we want neuronal lightning in our games
0
0
0
22
Replying to @AMD
@AMD @OpenComputePrj @LisaSu Thanks for sharing this
0
0
0
15
Replying to @AMD
@AMD @OpenComputePrj @LisaSu AMD onsite at OCP with Lisa Su keynoting the next era of AI infrastructure, that open rack-scale push is looking ready to scale hard.
0
0
0
19
Replying to @AMD
@AMD @OpenComputePrj @LisaSu OCP 峰会传递的信号是竞争维度变了:从单芯片跑分打到机架级(rack-scale)方案,拼的是互联带宽、整柜功耗和客户 TCO。Keynote 如果只谈单卡性能、不谈大客户的整柜采购意愿和交付节奏,那就是在避重就轻。
0
0
0
56
Your team reviews every change in Git. Except the API collections. Bruno fixes that: export from Postman, import, and collections become plain-text files you branch, diff, and merge like code. No account. No cloud sync. Free and open source.
16
27
360
2,300,260
Replying to @AMD
@AMD @OpenComputePrj @LisaSu @grok Any catalysts in there
1
0
0
52
Replying to @AMD
@AMD @OpenComputePrj @LisaSu Running Luna, my 100% local assistant, on a single RX 9070 XT. Moving from Ollama to llama.cpp with ROCm took one LLM call from 5.95s to 0.51s. Great to see AMD pushing AI infrastructure. #AMD #ROCm #LocalLLM #llamacpp #OCPSummit26
0
0
1
67