Nvidia CEO Jensen Huang’s keynote handle at CES 2025 was stuffed with AI bulletins, with the 61-year-old head of the chip big unveiling its newest line of gaming chips, an AI mannequin to coach robots and self-driving automobiles, its first desktop laptop, and extra.
Taking the stage earlier than a packed viewers in Las Vegas, US, Huang delivered a two-hour presentation that outlined Nvidia’s 2025 sport plan and set the tone for the key annual tech convention that may run from January 7 to 10.
Nvidia’s rocketing worth has been pushed by demand for its processors which might be powering the continued AI revolution. Forward of Huang’s CES keynote, the corporate’s shares closed at a document excessive $149.43 with a market cap of $3.66 trillion, which means that Nvidia remains to be the second-most invaluable firm on this planet, trailing behind Apple.
Right here’s a take a look at all the brand new GPU {hardware} and AI software program launched by Nvidia at CES 2025.
RTX 50-series GPUs
Nvidia formally introduced its newest line of GPUs referred to as the RTX-50 sequence at CES 2025 on Tuesday. The following-generation RTX GPUs will run on the corporate’s flagship Blackwell structure. The RTX 40-series of GPUs had been based mostly on its Ada Lovelace structure.
Huang unveiled 4 next-gen GPUs, particularly: RTX 5070, RTX 5070 Ti, RTX 5080, and RTX 5090 with costs starting from $549 to $1,999. The RTX 5090 and RTX 5080 GPUs will likely be accessible from January 30 onwards, adopted by the discharge of the opposite two GPUs in February this yr.
Nvidia’s newest GPU sequence include a brand new design that features two double flow-through followers, a 3D vapour chamber, and GDDR7 reminiscence. All 4 of the graphic playing cards are geared up with DisplayPort 2.1b connectors to drive shows as much as 8K and 165Hz.
The scale of the RTX 5090 GPU is considerably trimmed down from its predecessor (the RTX 4090) and is reportedly able to becoming inside a small-form issue PC. Nonetheless, the entire graphics energy (575 watts) of the RTX 5090 is 125 watts greater than that of the RTX 4090.
Nvidia’s RTX 50-series of GPUs may even be built-in into laptops, that are anticipated to be made accessible by numerous PC makers from March onwards.
Throughout his keynote, Huang confirmed off the capabilities of the RTX-50 GPU sequence by way of a real-time rendering demonstration of key options comparable to RTX Neural Supplies, RTX Neural Faces, text-to-animation, and even DLSS 4.
DLSS 4 with multi body era
Nvidia introduced that it’s upgrading its Deep Studying Tremendous Sampling (DLSS) know-how with new neural rendering capabilities comparable to Multi Body era on techniques working the RTX-50 sequence. It should generate “as much as three further frames per historically rendered body, working in unison with the whole suite of DLSS applied sciences to multiply body charges by as much as 8X over conventional brute-force rendering.”
“The brand new era of DLSS can generate past frames, it could predict the long run. […] We used GeForce to allow AI, and now AI is revolutionizing GeForce,” Huang stated in his keynote.
The DLSS 4 improve will work on current RTX GPUs as nicely.
Cosmos basis fashions
Nvidia has introduced a brand new set of AI fashions referred to as Cosmos that can be utilized to generate photo-realistic movies to be able to practice robots and autonomous autos.
Customers, for example, will have the ability to immediate Cosmos to generate a video of an atmosphere that follows the legal guidelines of physics.
The synthetically generated knowledge is anticipated to assist robotics firms and AV makers lower down the prices of gathering standard, real-world knowledge. “We’re going to have mountains of coaching knowledge for autonomous autos,” Huang stated, including that knowledge assortment from “automobiles on the street” will proceed to be wanted.
Nvidia is releasing Cosmos beneath an “open licence”. “We actually hope (Cosmos) will do for the world of robotics and industrial AI what Llama 3 has achieved for enterprise AI,” Huang stated.
Challenge Digits
Nvidia introduced its first-ever desktop laptop referred to as Challenge Digits that has been designed for AI researchers, knowledge scientists, and college students. It prices $3,000, runs on a Linux-based working system, and packs a brand new GB10 Grace Blackwell Superchip that delivers a computing efficiency meant for working and fine-tuning AI fashions as much as 200 billion parameters in measurement.
“[Project Digits] runs your complete Nvidia AI stack — all of Nvidia software program runs on this. […] It’s a cloud computing platform that sits in your desk … It’s even a workstation, in the event you prefer it to be,” Huang stated onstage.
Challenge Digits will grow to be accessible in March 2025.
Uncover the Advantages of Our Subscription!
Keep knowledgeable with entry to our award-winning journalism.
Keep away from misinformation with trusted, correct reporting.
Make smarter choices with insights that matter.
Select your subscription bundle
