Based on the latest Mercury Research data (sent over by AMD this afternoon), AMD hits record 41.3% Server Revenue Share in Q4 2025 🚀 For the first time, AMD has surpassed the 40% mark in server revenue share. What’s even more interesting is the "Quality of Share", they aren't

Got the new Dell XPS 16 with the Intel Core Ultra 3 series (Panther Lake) today. Perfectly pairs with the UltraSharp 52” 6K display. Don’t mind if I do… And yes, using the display’s KVM and PbP feature for a Mac running OpenClaw🦞. 😂 Testing soon from Signal65.

Love to see the aggressive and direct statements. Looking forward to attending Advancing AI this summer to see what the AMD team has been building.

Six Five Media is heading to NVIDIA GTC 2026 in San Jose with analysts Patrick Moorhead, Daniel Newman , @BrendanBurke, Ryan Shrout , and Anshel Sag . From AI factories and agentic workflows to inference at scale, we’ll be unpacking what’s moving from roadmap to real-world

Meta and Nvidia announced a multi-year, multi-generation partnership that will span millions of Nvidia GPUs (Blackwell now, Rubin next), plus expanded CPU deployment with Grace and a potential large-scale Vera CPU rollout starting in 2027. (A big deal for Arm here as well,

I don’t know what the punishment is for robo-calls and telemarketing and the likes, but it’s not severe enough.

FWIW I’ve had the best success with Claude and creating complex decks based on any degree of complexity. One shot. Based on a research paper. Based on a transcript. Anything. I do have to ask it to radically reduce the text. And not awesome at graphics. -ChatGPT tells me it

Blackwell inference economics are starting to look very different. NVIDIA says Baseten, DeepInfra, Fireworks AI, and Together AI are seeing up to 10x lower cost per token vs Hopper by pairing open-source frontier models with optimized stacks on Blackwell. Also yes, that’s me

DigitalOcean announced new GPU Droplets powered by AMD Instinct MI350X, positioned specifically for “agentic inference” workloads, where latency and throughput determine if apps to feel either instant or clunky. DigitalOcean is talking about compute-bound prefill (the heavy

I can't wait for our team Signal65 to get hands on with Maia 200 and really see how it stands up to other silicon in the market on specific workloads. Looks like Microsoft has definitely built the Maia 200 to be an inference beast for its ecosystem.

Hey, do you remember when everyone said DeepSeek was trained in two weeks with a toothpick in a pack of gum as the compute?

Let me repeat. H/T Ryan Shrout $IBM

AMD $AMD and Meta $META just announced a $100B and 6-gigawatt over 5-year, multi-generation partnership to deploy custom AMD Instinct GPUs and CPUs across Meta's AI infrastructure. It was great to catch up with AMD CEO Lisa Su on this. Here is what the deal includes:   •GPU:

A definitive agreement: AMD and Meta are expanding our existing partnership, aligning roadmaps across silicon, systems and software to deliver AI platforms purpose-built for Meta’s workloads. ✅ 6GW of AMD Instinct GPUs ✅ Custom AMD Instinct GPU derived from MI450 for Meta

There are a lot of interesting takes on this announcement, especially when paired with the previously announced, and structurally very similar, OpenAI deal. This Meta 6-Gigawatt deal isn't just about silicon; it’s about alignment. 1/ The Cementing Strategy: By issuing 160M

Need some fisheye rack server in your feed this morning?

I think this view of what NVIDIA's business model is incredibly interesting, and really showcases the power and influence the company has on the market as a whole. All eyes on earnings tomorrow...

Intel is getting in on the AI announcements today as well, in partnership with SambaNova. I still have a lot of digging to do on what makes these chips unique, but the claims are impressive. and the partnership with Intel means the potential to scale quickly and broadly.

AI inference isn’t just about GPUs. The host CPU is critical. On this Signal65 podcast, Ryan Shrout, Russ Fellows, and Mitch Lewis speak with AMD’s Madhu Rangarajan and Curt Waltman on how AMD EPYC EPYC CPUs improve enterprise AI inference performance. Full episode: