CES 2025 Tech Trends
Consumer Electronics Show
TechChurn Special Edition
February 3, 2025
Artificial Intelligence was, of course, everywhere at the Consumer Electronics Show (CES 2025) as this year’s top AI products and consumer devices offered no shortage of amazing reveals.
Featuring everything from self-driving cars to more Copilot+ laptops, other big names such as NVIDIA shaped many of these trends, as a driving force and main contributor.
Tech Churn has rounded up a few trends in commercial products and AI from the show.
Agentic AI: The Next Move for Generative Bots
The last half of 2024 seemed dominated by the buzz of Agentic AI and it was the clear frontrunner at CES this year. Agentic AI ties together multiple actions by several generative AI services to automatically perform tasks a human worker can take hours or days to complete.
GPU Tools and AI Dev
NVIDIA’s Blueprints for agentic AI are pre-built packages of NIM micro-services and technologies from AI partners. (Think of LangChain as a good example.) NIM leverages industry-standard APIs for app integrations.
LangChain uses its own LangGraph, plus Llama 3.3 70B NVIDIA NIM micro-services to generate reports. The agent searches the web and interprets the user’s request to create and provide the report in a particular format.
NVIDIA's partner sees agentic AI as useful for personalizing patient care in clinical trials, managing inventory, and troubleshooting industrial equipment problems. The company partnered with NVIDIA on its AI Refinery platform for deploying agents in business environments. Karthik Narain, group chief executive of technology and chief technology officer at Accenture, said in a press release that “Advancements in digitizing knowledge, new AI models, agentic AI systems and architecture enables enterprises to create their own unique cognitive digital brains.”
It's hard to believe the first CES was held in June 1967 in New York City. It was a spin-off of the Chicago Music Show which, until then, had served as the main event for exhibiting consumer electronics.
As ever, new electronics innovations are making significant inroads for technical advancements.
Next-generation GPUs revealed
The chips powering generative AI training and inference and processors for laptops and PCs were at the top of my list at CES 2025. The major processor announcements were:
NVIDIA’s GeForce RTX 50-series.
AMD’s Radeon 9000 series and Ryzen AI series.
Intel’s Core Ultra 200V series.
Qualcomm’s Snapdragon X.
The World's Smallest AI Supercomputer
The GeForce RTX 5090 GPU is a beneficiary of NVIDIA’s top-of-the-line Blackwell architecture. Developers can also look at the $3,000 Project DIGITS, which NVIDIA calls a desktop supercomputer. Project DIGITS uses the one-pentaflop NVIDIA GB10 Grace Blackwell Superchip for prototyping, tuning, and deploying generative AI models.
Tech Stocks Take A Hit
NVIDIA stock recently dropped for several days, due to introduction of smaller AI models and some obviously steep competition. China's new upstart known as DeepSeek caused a monumental disturbance — sending tech stocks into a nosedive.
Now, OpenAI says Chinese startups are recreating AI models done by U.S. companies, and are painfully similar to ChatGPT. It claims rivals are persistently attempting to replicate existing AI technologies, adding that OpenAI and its partner Microsoft have been banning accounts suspected of such moves.
The company didn’t directly mention DeepSeek in its statement, but we would be remiss if we didn't also mention OpenAI's admission of copyright infringement last year. Several media companies cried foul on OpenAi for its inappropriate use of training data.
It's important to note that the 5090 is a hard to find for some strange reason.
Despite availability online commentary is abundant. Enthusiastic geeks and gamers are exccited about the GeForce RTX 50-series GPUs. But, it’s generating some web reviews which are less than impressive. With so much speculation on the interwebs, one can only guess that maybe scarcity is a driver for demand?
Project DIGITS
The Project DIGITS desktop can run a petaflop of AI computing and takes up about as much space as a Mac mini on your desk.
You're probably wondering what heck is a petaflop? To explain, we first need to take a look at FLOPS (also called Floating-Point Performance). A FLOP is a specification that defines graphics card performance metrics.
The first thing that springs to mind is usually megahertz or gigahertz when folks talk about computer processor speeds. A FLOP -- or floating point operations per second -- is a specification that expresses how fast your computer can do mathematical calculations.
It involves a range of really large numbers, as well as smaller and even fractional numbers. Typically you see it broken down in 32 and 64 bit for a single number. Computers are typically limited to a finite amount of space in combination with processing operations for things like exponents values, changing decimal numbers and rounding, making it hard to put them into basic bit calculations and standard integers. In short, it's not cut and dried.
As you by now have guessed, scientific research or certain kinds of digital modeling fall into this category. To put it into perspective, much of the math needed to generate images on your screen uses vectors to create shapes, lines, and their positions. Super computers rely heavily on floating point numbers, but the average user probably won't need to be worry about using a high powered GPU with the highest amount of FLOPS.
What does that mean?
More power is obviously better for your gaming graphics (i e. GPU power), camera Megapixels, computer memory and processing CPU or megahertz allocations. Unless you're working with big data analytical research, or a complex AI research project you won't need a supercomputer or high powered workstation with that kind of power.
The Concept
Projects Digits is powered by a 20-core GB10 Grace Blackwell Superchip on 128GB of LPDDR5X memory with a 4TB NVMe solid-state drive (SSD) — that is just mind-blowing.
“Fusing AI-driven neural rendering and ray tracing, Blackwell is the most significant computer graphics innovation since we introduced programmable shading 25 years ago,” NVIDIA CEO Jensen Huang said in a press release.
Nvidia's Cosmos promises to be the underlying technology that could accelerate all of them into the future. Robotics manufacturers and car makers could use Cosmos to intelligently deal with real-time situations, ranging from navigating uneven or cluttered environments to unexpected obstacles.
This could be the AI platform that enables innovators at CES for years to come. Cosmos is what Nvidia CEO Jensen Huang called "the world's first world foundation model," a previously missing link that will allow robots and autonomous vehicles to become much more capable.
It solves the problem of the robots need for vast amounts of training data in order to make them more useful. Cosmos can simulate that data with AI, taking digital models of roads, factories, homes and other physical spaces; transforming them into simulations that innovators can use to create the robots of tomorrow.
Best TV or home theater: LG G5 OLED TV
New TVs always steal the show, and this one is no exception. The LG line of TV’s unveiled its latest flagship OLED home theater — the breathtaking LG G5 OLED.
Samsung has also revealed its new ode to aesthetically pleasing design. The Frame Pro TV it was nothing short of impressive.
Hisense also launched the rather expansive mini-LED TV in the form of an imposing 116-inch behemoth. The new ‘XXL-size premium’ TV also comes in a consumer-ready 163-inch micro-LED version — another beast of a TV. It really make the home theatre experience come alive.
With it's sleek lines and Hisense’s Hi view AI Engine X processor. It's a surprising addition, since they usually present more budget-friendly versions. According to the company, this new processor is faster than the one used in its 2024 TV lineup, and has specific AI functions to optimize picture, sound, and energy consumption.
Hands-off AI laundry
Most days, laundry is often a boring, unavoidable chore. The inevitable question pops up at the wrong time… do you run a small load when you need just two or three things? Should you waste laundry detergent, and water? Or do you wait until more dirty clothes accumulates to justify dealing with it? (Don’t even mention drying and folding all of it.) Tenet's AI Laundry Robot prototype tackles everything. For small loads, the egg-shaped unit washes and then hang-dries clothes inside the machine. (We hear it can even fold the laundry.) The bigger front-loading version does the same with larger jobs.
Needless to say — we can’t wait for next year’s CES 2026 in Las Vegas.