Fujitsu boosts oneDNN performance on ARM processors

Hani Daou gives an overview of the high scaling of copper interconnects for AI networks, the extreme amounts of testing required for the ramp up, and how this lays the foundation for scalable #CPO and #NPO systems in the near future. Watch here: https://bit.ly/ml-ocp-25

OCI’s massive GPU fabrics demand rock-solid L1 reliability. Our 18-month journey with Stephen M. on Zero Flap Optics is paying off. Check out the video to learn about eliminating the “bucket”!

AI at scale breaks when time drifts. Every distributed system — whether orchestrating thousands of GPUs or serving real-time inference — relies on a single assumption: that every machine agrees on when something happened. But in reality, they don’t. Milliseconds of clock skew between nodes means: stale data, misordered logs, inconsistent state, and unpredictable performance. That’s where Clockwork comes in. We deliver software-defined clock synchronization that brings your entire distributed fabric — GPUs, containers, VMs — into perfect temporal alignment. The result? Deterministic behavior. Consistent observability. Real-time reliability at massive scale. See how Clockwork’s time foundation keeps distributed AI infrastructure in lockstep. 📍 Meet us at Ray Summit 2025 — San Francisco, Nov 3–5. https://lnkd.in/gjvKdvFF

An S&C team including Nader Mousavi and Mark Schenkel is advising OpenAI in its strategic partnership with AMD to deploy six gigawatts of GPUs to power OpenAI’s next-generation AI infrastructure. Learn more: https://lnkd.in/gJ5MRem8

We just ingested 117,659 nodes and 290,674 relationships into our knowledge graph—on commodity CPU, with no GPU assistance—in just 237 ms. The system remains instantly queryable through real-time validation and region-aware tiering, which keep the hottest data fast while efficiently managing the rest. This is a major step toward self-hosted semantic memory at production scale. #AI #SemanticMemory #EdgeAI #KaiCore *MLCommons Benchmarking is still pending. But we are pushing hard for commonly recognized standards.

We’re incredibly proud to announce a multi-year, multi generation strategic partnership with OpenAI that puts AMD compute at the center of the global AI infrastructure buildout. ✅ 6GW of AI infrastructure ✅ Initial 1GW deployment of AMD Instinct MI450 series GPU capacity beginning 2H 2026 ✅ Enabling very large-scale AI deployments and advancing the entire AI ecosystem More on the news: https://bit.ly/3KzsnFk.

Some breaking news in the #semiconductor space; AMD and OpenAI announce partnership, with deployments beginning H2 2026. AMD's share price is currently up around 26% during pre-market trading. AMD's press release is linked below.

WOW 😲 ‘Shares of AMD jumped more than 23% in premarket trading. Deal includes warrant for OpenAI to buy up to 10% AMD stake at 1 cent per share AMD expects more than $100 billion revenue from OpenAI, others as a result of deal.’ Reuters

Big Big News in AI + Compute ! I’m thrilled to see AMD and OpenAI announce a bold and ambitious partnership: a 6-gigawatt deployment of AMD GPUs over the coming years, beginning with a 1 GW rollout of the Instinct MI450 series in H2 2026. This is a game changer — here’s why it matters (to me, and to all of us watching where AI infrastructure is headed): 🔍 Why this is so significant: 🔹 Massive scale — 6 GW is a huge commitment in the world of AI infrastructure. 🔹 Multi-generation collaboration — AMD and OpenAI will co-design across hardware and software stacks, aligning roadmaps and optimizations. (Advanced Micro Devices, Inc.) 🔹 Strategic alignment & incentives — AMD is issuing OpenAI warrants (up to 160M shares) tied to milestones and share-price goals. (Advanced Micro Devices, Inc.) 🔹 Ecosystem acceleration — Moves like this help accelerate innovation throughout the AI stack (chips, cooling, datacenters, software) 🔹 Vision for the future — It signals that compute is no longer a limiting factor (if all goes to plan), unlocking new possibilities in generative AI, simulation, and more. 💡 My take / what I’m excited about 1. This kind of “compute scale commitment” is a statement: we are in a phase where compute is infrastructure — like power or networking — and needs to be reliably available. 2. The deep hardware-software collaboration could drive optimizations that benefit many (beyond just OpenAI) — from more efficient deployments to better AI performance per watt. 3. It might trigger ripple effects: more demand for data center innovation (cooling, power, facility design), better chip yields, more competition among hardware vendors. 4. It’s a bold risk — but one with huge upside, if AMD and OpenAI execute well. If you’re working anywhere at the intersection of AI, infrastructure, or high-performance compute, this is a moment to watch closely. 👇Happy to hear thoughts — what do you all think the challenges will be (cooling, cost, supply chains?) and the biggest opportunities? 🔗 More on the news: https://bit.ly/3KzsnFk. #AI #Infrastructure #Compute #AMD #OpenAI #HighPerformanceComputing #TechStrategy #AMDBrandAmbassador #TogetherWeAdvance

◆ Core Message AMD has entered a multi-year, multi-generation agreement to supply AI chips to OpenAI, committing to deliver 6 gigawatts of GPU capacity over time. The initial deployment will begin in 2H 2026 with AMD’s Instinct MI450 series. As part of the agreement, OpenAI is granted a warrant for up to 160 million AMD shares, tied to performance milestones. --- ◆ What This Really Means This deal marks a serious elevation of AMD’s role in the AI infrastructure supply chain. Rather than incremental sales, AMD is now a core strategic compute partner to OpenAI across multiple GPU generations. The share-based incentive aligns interests between AMD and OpenAI, and the scale (6 GW) suggests that AMD expects significant future revenue and influence in training and deployment of large AI systems. --- ◆ Key Insights The agreement is explicitly multi-year and multi-generation — AMD and OpenAI will collaborate on future GPU architectures, not just one product cycle. The first deployment will be 1 gigawatt of AMD Instinct MI450 series, starting late 2026. OpenAI receives a warrant for 160 million shares of AMD, vesting in tranches linked to deployment and share price targets. The total scale (6 GW) is substantial: in public statements, it’s estimated that 1 GW of AI capacity costs tens of billions of dollars in chips + infrastructure. Though financial terms weren’t disclosed, AMD executives have indicated that the deal could produce “tens of billions” in revenue over its lifetime. --- ◆ Why This Matters For sales / BD: This is a credibility boost. You can point to AMD as a deep compute partner (not just a parts supplier) in conversations with large AI customers. For engineering / product: AMD’s roadmap must support not just current performance, but long-term integration with AI model requirements, power envelopes, and scaling. Strategically: This deal positions AMD more directly against NVIDIA in AI infrastructure. It changes how compute supply is contracted, potentially shifting some leverage to AMD in AI compute bidding and ecosystem partnerships. --- ◆ Final Takeaway This multi-year agreement with OpenAI is a pivot point. AMD is stepping into the AI compute arena more deeply than ever, aligning its future GPU roadmap with AI use cases. The shared stake and long-term deployment scale indicate high confidence—and high ambition—in becoming a foundational compute backbone in the AI era. --- ◆ Most Relevant Website AMD’s official press release on the partnership: AMD and OpenAI Announce Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs