Satya Nadella

Great to see the Fractal team at #NVIDIAGTC showcasing LLM Studio alongside Microsoft and NVIDIA. For those that may not know about Fractal - Fractal.ai is a global enterprise AI company and a Microsoft Solutions Partner that helps organizations turn data into actionable intelligence using Microsoft Cloud and AI technologies. Built on platforms like Azure AI and Azure OpenAI, Fractal delivers domain-specific AI solutions that help enterprises move from experimentation to production improving decision-making across areas such as supply chain, marketing, and operations. Enterprise AI is moving quickly, and platforms that help organizations build, manage, and operationalize domain-specific AI will play a huge role in turning experimentation into real business value. If you’re at GTC, stop by the Microsoft booth and connect with Louis Goldner, Chandramauli Chaudhuri, and the Fractal team to see how LLM Studio is helping organizations bring AI into production. And if you see Louis Goldner tell him your favorite dad joke he will love it. Sue McMahon | Richard Dalceno Chaves | Chris Longo | Mo'Shai Gibbs, MBA

8

Physical AI Builders's community partner Harvard Club of San Francisco and MIT Club of Northern California and Universal AI Services are hosting this pretty cool #Robotic and #PhysicalAI series of event on March 12. Talks from leaders building next-gen real-world AI systems Live robotic demonstrations (humanoids, robot dogs, edge AI in action, real-world perception, safety, and more) #physicalai is expanding, this is there #2 of the the 3 series. Check it out!

8

BIG news !!! NVIDIA buying Groq in a huge yet creative 20B USD cash licensing deal!!! Other than Google's TPUs, Groq is/was the most notable attempt on building an alternative that is AI-first to NVIDIA's #GPUs (using an architecture that allows for much more efficient scheduling). Grok's founder actually came from the Google TPU team. That's a strong addition to the Nvidia portfolio. Nvidia demonstrates clear commitment to maintain its dominance in the AI compute space, and an interesting signal for anyone working on an alternative... Less competition isn't great, but more options at scale is 🤷🏻‍♂️ On a personal note, I am a bullish investor in NVIDIA and an early investor in Groq - so quite happy 😀😇 https://lnkd.in/dt5qWuxa #compute #ai

64

1 Comment

The new NVIDIA–Accenture partnership is being called a true game-changer — giving Accenture access to NVIDIA’s cutting-edge AI infrastructure while filling a critical industry gap: helping enterprises scale AI when in-house expertise falls short. By combining Accenture’s deep consulting and implementation experience with NVIDIA’s advanced tech stack, this collaboration positions Accenture at the forefront of enterprise AI transformation — driving innovation, accelerating adoption, and unlocking real business value at scale. #AI #Innovation #DigitalTransformation #Accenture #NVIDIA https://lnkd.in/gwxS5agg

100

🚀🇺🇸 SF Compute Secures $40M to Expand AI Compute Marketplace 💡 San Francisco-based startup SF Compute has raised $40 million in a Series A funding round. The investment will fuel the growth of its marketplace designed to buy and resell unused GPU compute capacity. Here are the key details: 1️⃣ 💰 Major Funding Round The $40M round was led by DCVC and Wing Venture Capital, valuing the company at $300 million. 2️⃣ 🔄 Marketplace Model The platform allows buyers of GPU capacity to resell unused supply, aiming to increase market liquidity and efficiency. 3️⃣ ⚙️ Managed Scale While SF Compute owns no hardware itself, it reportedly manages over $100 million in GPU assets for its marketplace. 4️⃣ 🧠 Tech Focus The platform currently lists access to Nvidia H100 and H200 GPUs, with indications it will soon offer B300s. Quick question: Could a flexible compute marketplace be a key solution to the industry's supply and demand challenges? 🤔 (Date: December 2, 2025) #AI #Compute #Marketplace #VC #Startup #TechNews Invites You to # the 20th Annual China lDC Industry Ceremony (lDCC2025) and the Digital Infrastructure Technology Expo (DITExpo). https://difgc.idcnova.com/ 🔥See you on December 10-11 at Beijing Shougang Park to explore the future of computing power! #DITExpo #IDCC2025

1

Frontier models that use reasoning to "think" during inference are generating 5X more AI tokens per year. ✨ "Inference is now a thinking process. And in order to teach AI how to think, reinforcement learning and very significant computation was introduced into post-training," said NVIDIA CEO Jensen Huang in a recent keynote. Reinforcement learning is increasing computation demands across all AI scaling laws: pre-training, post-training, and test-time scaling. Learn more about reinforcement learning and AI scaling ➡️

21

2 Comments

Build high-performance, secure #AI applications with NVIDIA Nemotron RAG and Microsoft SQL Server 2025, announced at #MSIgnite. ✅ Improve performance bottlenecks ✅ Deploy AI models as simple, containerized endpoints ✅ Maintain security and flexibility 📝 Get started in our step-by-step guide: https://bit.ly/48tTo6q

2

“Data Center Cooling Stocks Sink After NVIDIA CEO’s CES Talk”. It’s the same mistake markets made with DeepSeek. When AI infrastructure advances, the physics do not change. Only the constraint shifts. At CES, Jensen Huang explained the cooling implication: “the power of Vera Rubin is twice as high… and yet… the water that goes into it is the same temperature, 45 degrees C. With 45 degrees C, no water chillers are necessary for data centers.” Markets heard “no chillers” and inferred “less cooling.” That inference is wrong. Every computing system must remove heat from silicon to the environment. That is non-negotiable. The only question is which layer of the cooling stack becomes the binding constraint. Historically, frontier AI halls were constrained by chilled water refrigeration. Mechanical chillers set the thermal design envelope. They added parasitic load, increased mechanical complexity, and often limited sites maximum scale. Vera Rubin shifts that design point. The system is architected around warm-water liquid cooling at a 45°C inlet. Jensen described the design succinctly: “80% liquid cool, 100% liquid cool.” The design change with Rubin: - chillers no longer set the thermal design envelope for frontier AI halls - parasitic refrigeration power is reduced - mechanical complexity and failure exposure decline Heat still must be rejected. Cooling systems still exist. Chillers may still be deployed for redundancy, mixed fleets, and extreme conditions. And mixed fleets are the baseline. Data centres will run Rubin alongside older GPUs, CPUs, and ASICs for years, many of which still benefit from colder coolant. The point is not that cooling goes away. The point is that the constraint moves. Specifically, chillers stop defining the hall design point, but operators will still chill water to operate within reliability margins, performance envelopes, and manufacturer spec’s. This outcome only exists because the system is redesigned end-to-end. Jensen was explicit about the mechanism: “unless we deployed aggressive, extreme co design… across the entire stack” And the infrastructure is engineered for AI load behaviour, including transients: “we now have power smoothing across the entire system” Vera Rubin does not reduce the need to move heat. It changes which subsystem becomes the bottleneck. Even in an all-liquid compute hall, residual heat from networking, cabling, and storage still lands in the datahall and must be handled at the facility level. Value shifts from refrigeration capacity toward liquid distribution, CDUs, heat-rejection hardware, controls, commissioning, redundancy, and integration. Capital risk and execution risk move with that shift. The market heard “no chillers” and priced “less cooling.” The correct read is “more liquid cooling, harder integration, and a different constraint.” Cooling remains. Heat remains. The constraint moved. In infrastructure, progress does not remove constraints. It relocates them. #ai

236

21 Comments