The AI industry is entering a new era: Nvidia has unveiled solutions that will change the way we approach inference—the stage when a trained AI model analyzes new data and makes decisions. CEO Jensen Huang announced at the GTC conference in San Jose that the company has created a "secret sauce" that will accelerate and simplify AI workloads for real-world applications.
The new Blackwell and VeraRubin architectures and systems are a combination of hardware and software solutions, including GPUs, high-speed memory, and specialized Language Processing Units (LPUs). They will enable AI models to work with massive amounts of data, accelerating inference hundreds of times faster than the previous generation of Hopper GPUs.
Why this is important: Previously, Nvidia GPUs weren't always suitable for inference due to high power consumption and memory limitations. New platforms address these issues, making AI faster, more efficient, and more cost-effective. The company predicts that the inference chip market could reach $1 trillion by 2027.
What this will change in the future: Blackwell and VeraRubin technologies will accelerate the adoption of AI in chatbots, demand forecasting, autonomous systems, and enterprise applications. New computing standards will emerge, AI computing costs will decrease, and artificial intelligence capabilities will become more accessible to everyday use.
ORIENT
