Nvidia and Accel’s $100M RadixArk Bet

Welcome back, EmergingNation. Today is packed with big shifts across AI, from courtroom battles to compute bottlenecks and billion dollar infrastructure moves.

• Musk vs Altman goes to court over AI
• Hidden compute bottlenecks drain AI efficiency
• Nvidia and Accel invest $100M in RadixArk open source AI
• DeepMind pushback on military AI use
• Microsoft pivots business model around AI

LATEST IN AI

So here is what is going on. Elon Musk has filed a lawsuit against Sam Altman and things are already moving in court.

Musk is pushing for stronger safety rules around AI development while Altman is saying that speed and big investment are necessary if the US wants to stay ahead in the global AI race.

It is getting a lot of attention because it brings two of the biggest names in AI into direct conflict and what happens next could shape how AI is built and controlled in the coming years.

🛜The Bottlenecks Slowing Down AI Compute

Most large AI training systems currently operate at just 35 to 45% MFU efficiency, meaning companies are paying for far more computing power than they actually use. Engineers at Lambda benchmarked Llama 3.1 models ranging from 8B to 405B parameters on NVIDIA Blackwell GPUs to identify where performance was being lost.

🚨Anthropic Teams Up With SpaceX for a Massive Compute Boost

Anthropic just locked in a huge deal with SpaceX for over 300 megawatts of AI compute and 220,000 NVIDIA GPUs. That means bigger limits for Claude users and more API power for developers. The real story? The AI race is now about who controls the hardware, not just the software.

🌐Nvidia and Accel Pour $100M Into RadixArk to Commercialize Open Source AI Engine

Nvidia and Accel have led a $100M seed round in RadixArk, a startup building on the widely used open-source inference engine SGLang. Founded by former xAI and Nvidia engineers, the company is turning a behind-the-scenes AI infrastructure tool into a major commercial platform backed by top tech investors.

1 I tested Jasper AI for generating marketing content like blogs, emails, and ad copy, and found it useful for creating content quickly with AI assistance.

2 Familiar with GitHub Copilot, an AI-powered coding assistant that helps generate code suggestions, automate repetitive tasks, and improve developer productivity inside the IDE.

3 Explored Llama 2, Meta’s open-source large language model designed for building AI applications, chatbots, and research-based generative AI solutions.

4 Familiar with Imagen by Google, a text-to-image diffusion model known for producing highly photorealistic images with strong language understanding.

1🛜 “Redraw the attached image in the most clumsy, scribbly, and utterly pathetic way possible. Use a white background, and make it look like it was drawn in MS Paint with a mouse. It should be vaguely similar but also not really, kind of matching but also off in a confusing, awkward way, with that low-quality pixel-by-pixel feel that really emphasizes how ridiculously bad it is. Actually, you know what, whatever, just draw it however you want.”

This GPT Image 2 prompt is going viral for turning normal photos into hilariously bad MS Paint-style sketches with chaotic, scribbly results.

2🧑‍💻A single prompt can now turn into a full video. Invideo AI lets creators generate Shorts, Reels, TikToks, and YouTube videos with AI-powered visuals, scripts, and voiceovers in just seconds.