Dassault & Vertiv Cut AI Factory Deployment by 50% with NVIDIA DSX

Key Takeaways

• Dassault Systèmes and Vertiv are partnering to cut AI factory deployment time by up to 50% using NVIDIA DSX reference designs.
• NVIDIA DSX POD, built around the Vera Rubin platform, provides a standardized blueprint for rack-scale AI system integration and transport.
• AI voice cloning gave a woman with ALS the ability to deliver her wedding speech in her own voice, highlighting the personal impact of the technology.
• Google Cloud identifies storage, not compute, as the primary bottleneck for scaling AI agents in production environments.
• OpenAI's Forward Deployed Engineering team helped John Deere and a semiconductor company apply AI in real-world, high-complexity scenarios.
• New Mexico won a $25 billion judgment against Meta and warned that more states may pursue similar legal action over misleading AI and privacy practices.
• Veeva launched Vault AI at 10 percent of subscription cost and Falcon agents for clinical trial tasks, but requires standardized systems to deploy them.
• EU AI Act does not mandate disclosure for all AI-generated content; context matters, according to EASA's Ludovic Basset.
• Yale's Dr. Joseph Cavallo emphasizes teaching radiologists data literacy so they can use AI tools without losing core clinical skills.
• A survey found 52% of security workers say AI attacks rarely or never affect their work, suggesting potential blind spots in cybersecurity readiness.

Dassault Systèmes and Vertiv partner on AI factory design

Dassault Systèmes and Vertiv are working together to speed up AI factory design. They combine the 3DEXPERIENCE platform with Vertiv infrastructure to validate NVIDIA DSX AI factories before building. The goal is to reduce field complexity and cut deployment time by up to 50%. The partnership also expands to Vertiv OneCore solutions.

NVIDIA DSX POD enables faster AI capacity deployment

NVIDIA DSX POD provides a design foundation for partners to install rack-scale AI systems in containers at the factory. It is built around the NVIDIA Vera Rubin platform and allows for factory integration, testing, and transport. The reference design covers compute, networking, power, liquid cooling, and mechanical requirements. This approach makes AI factory deployment more repeatable and predictable.

AI voice cloning gives woman with ALS her voice back

Monica O'Callaghan used AI voice cloning technology after ALS took away her ability to speak. The breakthrough allowed her to deliver an emotional wedding speech in her own voice. The technology helped her reclaim part of her identity. It shows how AI can preserve a person's voice even after they lose the ability to talk.

EU AI Act does not require all AI content disclosure

Ludovic Basset from EASA explained at a Romanian conference that the EU's AI Act does not require disclosure in every situation. Advertisers should consider context, asking how substantially AI was used and how significant it is to consumers. Disclosure is required for material uses like celebrity deepfakes but not for minor technical edits. Basset stressed that truth in advertising still applies regardless of labeling.

From AI prompt to 3D-printed armor in 18 hours

James Bricknell used Gemini 4.0 to create a web app that designs 3D-printable cosplay armor. He went from initial prompt to a finished 3D-printed object in under 18 hours. The process involved about 10 iterations to refine the app and ensure the models were 3D-printable. He used Layerling to edit the STL files before printing.

Google Cloud says storage is the real AI bottleneck

Google Cloud experts say storage is the main problem for scaling AI agents. Unlike old systems, AI agents work all day and use resources in unpredictable ways. Companies need secure sandboxes that can start and stop quickly to handle these bursts without wasting money. The new approach moves intelligence closer to where data lives instead of forcing all data into one central place. Security is also critical because agents can create new risks if they are not properly controlled and governed.

OpenAI FDE team helps customers use AI in real life

OpenAI created a Forward Deployed Engineering team to help customers use AI models effectively. This team works with clients to solve hard problems and close the gap between model capabilities and real-world use. For example, they helped John Deere use AI to give farmers personalized advice on using weed-control technology. The team also worked with a semiconductor company to make their testing process more efficient using AI tools. Their goal is to build systems that deliver practical value in complex environments rather than just showing off model features.

New Mexico AG warns AI industry after big Meta ruling

New Mexico Attorney General Raul Torrez warned the AI industry after winning a $25 billion judgment against Meta. He said the case showed that states are ready to sue tech companies for misleading consumers about privacy. Torrez told Silicon Valley that this is just the beginning and more companies could face similar legal actions. He plans to watch how AI products represent their capabilities and how they handle consumer data. The state used consumer protection laws to claim Meta deceived users about its privacy controls.

Yale radiologist discusses AI training and monitoring in medicine

Dr. Joseph Cavallo from Yale University School of Medicine discussed how to integrate AI into radiology safely. His department uses immediate feedback tools so radiologists can quickly report when AI misses something. They also use system macros to help explain false positives in reports to other doctors. Dr. Cavallo believes it is important to teach residents data literacy so they can use AI without losing their clinical skills. He warns that training must balance using new tools with keeping a strong foundation in traditional medical knowledge.

Veeva launches new AI pricing and agent tools for trials

Veeva announced new AI tools and pricing for clinical trials at its recent summit in Boston. The company introduced Vault AI, which costs 10 percent of a subscription fee with unlimited usage starting in the second year. Veeva also launched Falcon agents that can perform tasks like filing documents and drafting clinical study reports. These tools require standardization because agents cannot work with fifty different system configurations. The goal is to make clinical research more efficient by automating repetitive tasks and reducing the time to get drugs approved.

AI Security Threats Tools and Skills Gaps

AI has changed cybersecurity in important ways. Many security professionals do not deal directly with AI attacks. A survey found 52% of workers said AI attacks rarely or never affect their work. This may be because filters block many attacks or because staff have training to spot them. The article calls for a clear look at real threats and skills gaps instead of panic.

Sources

• Dassault Systèmes and Vertiv Team to Accelerate AI Factory Design and Deployment
• Achieve Productive AI Capacity Faster and More Predictably with NVIDIA DSX POD
• AI voice cloning helps woman with ALS get her voice back
• Does the EU's AI Act Require Disclosure of All AI-Generated Content? (via Passle)
• I Went From AI Prompt to 3D-Printed Armor in 18 Hours
• Beyond the model: The real AI bottleneck is storage
• What FDE means at OpenAI
• ‘Just the beginning’: New Mexico AG warns AI industry after $25B Meta ruling
• The Reading Room Podcast: Emerging Concepts with AI Integration in Radiology
• Veeva AI Pricing & Agentic Tools for Clinical Trials
• The State of AI Security: Threats, Tools, and Skills Gaps