Observed Signal · Mar 13, 2026 · Technical Release · Source: Trending Topics · Impact: 4/5 · Sentiment: Positive
Nvidia Invests $26B in Open-Source AI
Nvidia announced a plan to invest $26 billion over five years in developing open-weight AI models, as revealed in a financial filing and confirmed by company officials. The first result is Nemotron 3 Super, a 120-billion-parameter model optimized for autonomous AI agents, featuring a novel architecture with Mamba-2 layers, Transformer attention, and latent MoE, trained natively in NVFP4. The model claims to outperform OpenAI's GPT-OSS 120B and Alibaba's Qwen3.5-122B. Early adopters include Perplexity, Palantir, Cadence, and Siemens. The investment aims to defend Nvidia's CUDA ecosystem against rivals like AMD and its own customers developing custom chips, following an Android-like strategy of open models to drive hardware sales.
Nvidia's $26B investment and release of a powerful open-source model could significantly impact AI infrastructure and adoption, affecting advertising technology that relies on AI models.
Track NVIDIA Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Nvidia plans to invest $26 billion over five years in open-weight AI model development.
- The first model, Nemotron 3 Super, has 120 billion total parameters with 12 billion active, and a 1 million token context window.
- Nemotron 3 Super is natively trained in NVFP4, Nvidia's 4-bit floating-point format.
- The model claims to be 2.2x faster than OpenAI's GPT-OSS 120B and 7.5x faster than Alibaba's Qwen3.5-122B.
- Perplexity, Palantir, Cadence, and Siemens are already integrating the model into their workflows.
Connected Companies & Entities
14 Entities mapped“Nvidia plans to invest $26 billion in open-weight AI model development over five years....”
“Nvidia sells chips to Microsoft, which is also developing its own Azure AI chips....”
“Nvidia sells chips to Amazon, which is developing Trainium....”
“Nemotron 3 Super is 2.2x faster than OpenAI's GPT-OSS 120B....”
“Nvidia sells chips to Google, which uses its own TPUs....”
“Nvidia sells chips to Meta, which is developing MTIA chips....”
“Perplexity is integrating Nemotron 3 Super into its workflows....”
“AMD's MI355X offers 30% faster inference than Nvidia's B200....”
“Nemotron 3 Super is 7.5x faster than Alibaba's Qwen3.5-122B....”
“Siemens is integrating Nemotron 3 Super into its workflows....”
“Palantir is integrating Nemotron 3 Super into its workflows....”
“Jack Dorsey cut 40% of Block's workforce for AI initiatives....”
“AMD and Intel hardware can run models via OpenAI's Triton compiler....”
“Weights are available on Hugging Face....”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Windows 11 Search Upgrade, Copilot Gets File Access
Microsoft announced two upcoming updates for Windows 11: an enhanced search feature and expanded Copilot capabilities. The new search allows users to directly adjust settings like dark mode, take screenshots, and change volume from the search bar, with future ability to compose and send messages. Copilot's 'Hybrid Intelligence' feature, introduced at a Microsoft event, enables the AI to access local files and perform actions across the OS, such as finding, renaming, zipping documents, and attaching them to emails. Jacob Andreou, EVP of Copilot, demonstrated this in a tax return example. The rollout will be gradual, requiring user permission for file access. The search upgrade is expected later this year, with no exact date specified.
Musician sentenced for AI streaming fraud
A 54-year-old musician in the US has been sentenced to 18 months in prison for orchestrating a large-scale streaming fraud scheme. Using up to 10,000 bots, thousands of fake accounts, and thousands of AI-generated songs, he manipulated streaming platforms such as Apple Music, Amazon Music, and YouTube between 2017 and 2024. The fraudulent streams generated millions in royalty payments. At one point, his songs achieved 80.9 million streams in April, far surpassing Taylor Swift's 9.3 million. The court also ordered him to repay over $8 million. His defense argued that no musicians were directly harmed, but prosecutors countered that real artists and fans were deprived of royalties.
OAAA Launches Programmatic Center of Excellence for DOOH
AI is pushing advertising back into physical spaces due to declining trust in online content, the rise of AI answer engines that bypass websites, and the growth of agentic AI. Out-of-home (OOH) and digital out-of-home (DOOH) advertising are gaining renewed relevance, with US OOH revenue up 10.7% YoY to $3.16B in Q2 and DOOH up 18.5%, representing nearly 40% of category revenue. To guide this evolution, the OAAA has launched a Programmatic & Automation Center of Excellence led by COO Patrick Dolan, focusing on standardization, workflow automation, agentic AI, reporting/attribution, and omnichannel/retail media integration. The initiative aims to automate both programmatic and static OOH processes, addressing friction points. However, this trend raises concerns about consumer opt-out options and the over-commercialization of public spaces, as noted by Ezra Klein.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.