Nvidia just reported a massive 122% jump in Q2 revenue to $30.04 billion, driven by explosive 154% growth in its data center division. However, its gross margin dipped slightly to 75.1% due to Blackwell chip design tweaks, sending shares down over 7% after hours as sky-high investor expectations proved tough to beat.
NVIDIA just released NVIDIA Blueprints to help businesses build and deploy custom generative AI workflows. Each blueprint packages reference code, documentation, and Helm charts alongside NIM microservices, starting with customer service digital humans, drug discovery, and multimodal PDF extraction. It basically lets enterprise developers skip the plumbing and turn internal data into self-improving AI workflows right away.
Cerebras just launched its new AI inference service powered by the CS-3 system, clocking an eye-watering 1,800 tokens per second on Llama 3.1 8B. That is 20 times faster than hyperscale Nvidia GPUs while retaining full 16-bit precision at just 10 cents per million tokens. Why it matters: lightning-fast generation is critical for multi-step AI agents to finally feel real-time rather than sluggish.
Nvidia just introduced its first on-device small language model, Nemotron-4 4B Instruct, to its ACE digital human suite at Gamescom. Distilled from a 15B model, it is optimized to run locally on RTX GPUs with minimal memory and fast response times. Why it matters: Gamers can now talk to dynamic NPCs in games like Mecha BREAK entirely on local hardware, without relying on cloud servers.
NVIDIA just opened applications for its 24th Graduate Fellowship Program, offering Ph.D. researchers in AI, robotics, and computing up to $60,000 in funding alongside mentorship. Applicants must have completed at least their first year of doctoral studies and be available for an in-person internship in summer 2025. The deadline is Sept. 13, 2024.
Nvidia has been scraping massive amounts of video from YouTube, Netflix, and other platforms to train its AI models, according to leaked internal documents obtained by 404 Media. Employees raised copyright concerns, but managers insisted the project—code-named Cosmos—had approval from top leadership. Nvidia defended the practice as fully compliant with copyright law, while quietly developing 3D world generators and digital humans behind closed doors.
Nvidia just unveiled a suite of simulation tools and NIM microservices that let developers teleoperate humanoid robots using Apple's Vision Pro. By tracking hand gestures through the headset, Nvidia's MimicGen service generates synthetic training data from minimal demonstrations, slashing the time and cost needed to train models like Project Groot.
Nvidia has reportedly delayed its next-gen Blackwell B200 chips by at least three months due to a design flaw discovered late in production. Massive shipments to heavyweights like Microsoft, Meta, and Google are now pushed into early next year. Nvidia says production will still ramp in the second half, but declined to comment on rumors.
NVIDIA just showcased James, an interactive digital human built on its ACE platform and ElevenLabs voice tech that connects with users using humor and emotion. The company also previewed Maxine 3D, which turns 2D video feeds into real-time 3D avatars for telepresence. Virtual brand ambassadors are quickly shifting from uncanny bots into expressive, lifelike agents.
Shutterstock and Getty Images just rolled out major visual AI upgrades powered by Nvidia’s Edify architecture. Shutterstock launched a commercial beta for Generative 3D that creates editable assets and 16K HDRi lighting maps from prompts. Meanwhile, Getty doubled its generation speed to four images in six seconds and added fine-tuning, both leveraging commercially licensed data.
NVIDIA just announced a full suite of services, models, and computing platforms for humanoid robot developers at SIGGRAPH. Key additions include new NIM microservices for synthetic motion data, an OSMO orchestration cloud service that cuts development cycles from months to under a week, and a teleoperation workflow using Apple Vision Pro to turn tiny amounts of human demonstration into massive training data. Early access is now open via NVIDIA's developer program.
NVIDIA just broke down how its AI Foundry service helps companies build custom generative AI models using proprietary data. Powered by DGX Cloud, the NeMo platform, and support for Llama 3.1, it lets businesses forge and deploy tailored models as NIM microservices without building infrastructure from scratch.
Elon Musk just fired up xAI's Memphis Supercluster, packing 100,000 liquid-cooled Nvidia H100 GPUs tied together on a single RDMA fabric. Musk boasts it is the world's most powerful AI training cluster and aims to deliver the strongest AI on every metric by December, though local utility contracts for its massive 100-megawatt power appetite are still not finalized.
Sequoia Capital partner David Cahn just calculated that the AI industry needs to generate $600 billion per year to pay off its massive hardware and data center investments. Even with aggressive revenue assumptions for tech giants, there remains a staggering $500 billion gap. Why it matters: Big tech is pouring fortunes into GPUs, but actual software monetization is falling far behind.
OpenAI co-founder Andrej Karpathy took the stage at UC Berkeley, framing modern LLMs as a brand-new operating system akin to 1980s personal computing. He revealed that GPT-4 began as an unglamorous internal Reddit chatbot, urging young developers to embrace the 10,000-hour rule and let tiny hackathon side projects snowball into real breakthroughs.
Elon Musk just revealed that xAI plans to launch Grok 2 this August, promising a major leap in real-time knowledge and reasoning. Musk noted the team did heavy lifting to purge junk from Internet training datasets, while teasing that Grok 3—powered by 100,000 Nvidia H100 GPUs—is slated for the end of the year.
Google just officially released Gemma 2, introducing ultra-efficient 9B and 27B open models for developers. The 27B variant punches above its weight by matching models twice its size while running full precision on a single Nvidia H100, and the 9B edges out Llama 3 8B. It is available now under a commercial-friendly license.
AI startup Etched just raised $120M to build Sohu, a specialized ASIC hardwired solely for Transformer models. By stripping away general-purpose GPU flexibility, the company claims it outperforms Nvidia's H100 by 20x. Why it matters: It could drastically lower LLM serving costs, but the chip hasn't taped out yet and is useless for any other AI architecture.
Nvidia just dethroned Microsoft to become the world's most valuable public company, reaching a $3.34 trillion market cap after shares rose 3.6% on Tuesday. The chipmaker commands around 80% of data center AI processors, pushing its stock up over 170% this year alone. It confirms that the AI boom's biggest winner remains the hardware engine powering it all.
Pharma and tech giants like Eli Lilly and Nvidia see generative AI taking over drug discovery within a few years. By treating molecules and proteins like words in a sentence, AI can generate candidates in minutes that would take labs a year to synthesize, cutting research timelines from years to months. The real bottleneck remains human clinical trials.
