Nvidia's 120W APU leak could mark the beginning of the end for laptop GPUs
A new, credible leak suggests Nvidia is cooking up something that could transform gaming laptops as we know them: A 120W Arm-based APU that combines CPU, GPU, and NPU into a single chip.
If early reports are accurate, this could deliver RTX 4070–level performance in a slimmer, cooler, and more efficient package and see Nvidia take pole position in laptop chip primacy.
Let's be clear—this wouldn't just be another performance play that we frequently see with new hardware iterations, but a potential design revolution that could change everything. Or nothing.
The leak, first reported by YouTube channel Moore's Law Is Dead and later verified by Notebookcheck, shows an engineering sample of the chip surrounded by eight LPDDR5X memory modules. The chip integrates a custom Arm CPU and a next-gen Nvidia GPU on the same die, likely codenamed Blackwell. There's also a powerful NPU onboard, positioned to handle heavy AI workloads.
What's particularly impressive is that this unified chip reportedly runs at a 120W TDP and still manages to compete with the RTX 4070 Laptop GPU in benchmarks.
Sources claim Nvidia is internally comparing its new APU to a 65W-tuned 4070, suggesting near parity in gaming performance with far better power efficiency.
According to Moore's Law Is Dead, the chip is scheduled to hit the market between Q4 2025 and early 2026, and Nvidia is reportedly working with Dell to bring this architecture to life in future Alienware laptops.
If Nvidia's APU lives up to its potential, the entire idea of separating CPU and GPU in a gaming laptop could become obsolete.
Combining the two allows for tighter integration, lower latency, and shared access to high-bandwidth memory. Thermal design also benefits since there's only one hot zone to manage instead of two.
This could mean fewer fans, thinner chassis, and better battery life, all without sacrificing AAA gaming performance. Those in the know will recognize this as a similar design philosophy to Apple's M-series chips, but unlike Snapdragon or Apple silicon, this chip might bring actual desktop-class frame rates to the table.
Nvidia's closest competitor here is AMD, which is preparing its own monster chip: The Strix Halo APU.
AMD's APU is expected to pair up to 16 Zen 5 cores with 40 RDNA 3.5 GPU compute units, pushing as much as 120W as well. In theory, it should be a beast.
But Nvidia may have a leg up with its more efficient GPU architecture and early adoption of on-package memory. Notebookcheck notes that Nvidia's sample includes LPDDR5X soldered around the die, enabling high memory bandwidth without a separate graphics memory pool.
That unified design is a big win for energy efficiency, which matters more than ever as gaming laptops strive to balance power and portability.
A 120W chip still requires serious cooling, so we're not talking about fanless devices. But without the need for a separate GPU and VRAM modules, OEMs have a chance to rethink their internals. One configuration might resemble a MacBook Pro in weight but pack gaming performance that rivals today's bulkier rigs.
Keep in mind that we're just speculating and thinking about what could eventually be. It would be surprising to see a dedicated Nvidia gaming laptop SoC right out of the gate, but the rumored APU could well be a starting point. That said, anything can happen, and the quoted sources do directly reference the APU as being earmarked for gaming laptops.
The most likely initial application for the APU is AI. With the likes of upscaling, frame generation, real-time voice effects, and creative tools becoming more common, a strong on-chip AI engine could be the killer feature that separates Nvidia's solution from AMD and any future moves by Intel (though Intel is conspicuously absent from the APU race at the moment, with nothing in its roadmap suggesting a full-on AI or gaming APU.)
With sources claiming that the chip includes an NPU powerful enough to run modern generative models and assist with GPU-accelerated workloads without the cloud, it's hard to think of any applications where it wouldn't shine, gaming or otherwise.
For Dell and other partners, this is essentially an invitation to innovate, perhaps eventually leading to the first wave of 'all-Nvidia' laptops, with unified software and hardware tuned for peak efficiency.
AMD Strix Halo APU could make low-end discrete GPUs obsolete — Nvidia's RTX 4060 is on notice
Nvidia's affordable RTX 5060 is coming soon, but that's not the GPU I have my eyes on
Dell's new laptop ditches the GPU for a discrete NPU — here's why that's a big deal
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Tom's Guide
a few seconds ago
- Tom's Guide
Don't upgrade yet — the M6 MacBook Pro sounds like a game-changer and ditches the most annoying feature
All of the best MacBooks have been relatively predictable for the last few years. The 'new' design, which debuted in 2021, doesn't feel all that new nowadays, and one of its more contentious inclusions is the notch. Housing the camera, it's always felt like Apple making the best of a bad situation. Now, it appears the Cupertino crew could be preparing to retire the controversial feature and replace it with something else entirely, alongside a new OLED panel in time for the M6 versions of the machine (expected in 2026 after this year's M5 models). According to new reports, Apple's M6 MacBook Pro could feature a tandem OLED panel, similar to what is currently available in the M4 iPad Pro. And, backing up previous rumors, this could spell the end for the notch. Notebook Check has suggested that the MacBook Pro webcam could be moved to a punch-hole (similar to phones from other manufacturers), freeing up the space around it for the Dynamic Island to make its debut on Mac. This clever feature arrived with iPhone 14 Pro and Pro Max, and is still going strong in the iPhone 16 lineup — allowing Live Activities to occupy a smaller space on your screen. Using your music player? Tap the Dynamic Island to make it full-screen, or use it to track your takeaway delivery. Lending more credence to the idea of Dynamic Island debuting on Mac is that Live Activities have landed on the Mac in the upcoming macOS Tahoe. At the moment, they pop up in your Menu Bar, but it's not a wild idea to think we could see them in a Dynamic Island on Mac soon. Get instant access to breaking news, the hottest reviews, great deals and helpful tips. Follow Tom's Guide on Google News to get our up-to-date news, how-tos, and reviews in your feeds. Make sure to click the Follow button.
Vox
a minute ago
- Vox
Mark Zuckerberg's unbelievably bleak AI vision
is a senior correspondent and head writer for Vox's Future Perfect section and has worked at Vox since 2014. He is particularly interested in global health and pandemic prevention, anti-poverty efforts, economic policy and theory, and conflicts about the right way to do philanthropy. The future is Zuck, and all of us, using AI on our sunglasses, apparently. Allison Robbert/Bloomberg via Getty Images Of all the many famous Steve Jobs stories that tech industry folks like to share, perhaps the single most famous is his 1983 pitch to then-Pepsi president John Sculley to join Apple: 'Do you want to spend the rest of your life selling sugared water or do you want a chance to change the world?' Like many things Jobs said, the pitch was wildly arrogant, self-important and self-aggrandizing, but ultimately correct. What Sculley did at Apple (mostly after firing Jobs) to sell the Macintosh and popularize personal, graphics-centered computing changed the world more than his invention of the Pepsi Challenge had. There really was a huge difference between selling Macs and selling sugar water. After listening to Meta CEO Mark Zuckerberg lay out his vision of how AI 'superintelligence' would change the world, though, my main reaction was: man, this guy just wants to sell us sugar water. 'Personal superintelligence'? Maybe just 'superintelligence,' it's cleaner In an Instagram video (of course) posted last week, Zuck explains that Meta's goal is to develop 'personal superintelligence for everyone,' accessed through devices like 'glasses that can see what we see, hear what we hear, and interact with us throughout the day.' 'A lot has been written about the scientific and economic advances that AI can bring,' he noted. 'And I'm really optimistic about this.' But his vision is 'different from others in the industry who want to direct AI at automating all of the valuable work': 'I think an even more meaningful impact in our lives is going to come from everyone having a personal superintelligence that helps you achieve your goals, create what you want to see in the world, be a better friend, and grow to become the person that you aspire to be.' This story was first featured in the Future Perfect newsletter. Sign up here to explore the big, complicated problems the world faces and the most efficient ways to solve them. Sent twice a week. The main reaction to this pitch I've seen from smart AI observers is: are you kidding? 'Superintelligence,' by definition, means a system that performs better than a human, sometimes vastly better, across most if not all domains. And the most ambitious thing Zuck can think of to make with that is… VR glasses? As Fortune's Sharon Goldman put it, while Steve Jobs called his computers 'a bicycle for the mind,' 'Zuckerberg, by contrast, imagines superintelligence as a pair of Ray-Bans that help you…be a better friend?' The lack of ambition in Zuckerberg's rhetoric is all the more striking when one considers the extreme ambition of his spending on AI. This year alone, he's hired former GitHub CEO Nat Friedman and veteran AI founder Daniel Gross; Scale AI CEO Alexandr Wang (as part of a quasi-purchase of Scale, a massively important company whose training data is used by just about every AI company); Apple AI chief Ruoming Pang; and ChatGPT co-creator Shengjia Zhao, among several others. His hiring spree, and the gargantuan amounts he's willing to pay top talent, have roiled the sector for weeks now. At one competitor (former OpenAI CTO Mira Murati's Thinking Machines Lab), Zuckerberg reportedly sent offers to more than a dozen of the company's 50 staffers, one of which was for over $1 billion over a few years, while the rest ranged from $200 million to $500 million over four years. Even for a company as rich as Meta, billion-plus offers for talent are unheard of. ($1 billion is how much Zuckerberg paid for all of Instagram in 2012.) It's a pretty vivid sign that Zuckerberg sees AI as the future of his business. But what does that future look like? I've seen the best digital minds of my generation wasted on Reels One could, very charitably, reason that Zuckerberg knows that a world of vastly superintelligent AI systems would lead to massive, far-reaching social ramifications that are not adequately summarized as 'you get smart sunglasses,' but he has concluded that most of his investors and customers aren't in a place to understand the gravity of those changes. Hence, talk about glasses. That could be what's happening, and I have some sympathy for his position if so. Trying to game out what a post-superintelligence world looks like is in fact extremely difficult, not least to those of us limited to mere human intellects. And it's usually scary — even if the changes ultimately prove positive. For all the uncertainty, there is no plausible world where people have access to 'personal superintelligence' and they and businesses do not use that to automate huge numbers of tasks, and there are a number of conceivable scenarios where that leads demand for human wage labor to totally collapse. Other scenarios see wages skyrocket. It's a tough situation for a CEO to message. But it's also worth considering the possibility that Zuckerberg means exactly what he's saying: that the AI systems his team is building are not meant to automate work but to provide a Meta-governed layer between individual human beings and the world outside of them. Facebook and Instagram are, in a sense, very crude versions of that layer, synthesizing and compressing the outside world into a digestible and addictive form people can consume throughout their days, and Zuckerberg's earlier obsession with the metaverse seemed a logical continuation. This approach has been immensely profitable. (Though, not so much the metaverse.) Imagine how much more profitable it'd be if a digital mind much smarter than Zuck's was designing it. Conversations like Zuckerberg's with the business writer Ben Thompson in May give credence to this interpretation. Zuckerberg sees four opportunities with AI: improving his products' recommendation algorithms to better target advertising, driving greater engagement on 'consumer surfaces' like Instagram Reels, 'business messaging' (i.e., businesses doing transactions through WhatsApp and Messenger, using AI), and lastly direct AI use à la ChatGPT. The promise of AI, to Zuckerberg, is that it can help him sell you more ads and convince you to spend more time watching Instagram brainrot. My reaction to that pitch was the same as AI writer Zvi Mowshowitz's: 'It was like if you took a left wing caricature of why Zuckerberg is evil, combined it with a left wing caricature about why AI is evil, and then fused them into their final form. Except it's coming directly from Zuckerberg, as explicit text, on purpose.' At least the sugar water from Pepsi tastes good. That the sixth largest company on Earth is devoting billions of dollars toward this vision is not, y'know, great. But it has a silver lining. One thing I've learned from talking to AI researchers over the years is that most of them are driven by a conviction that this thing they're building is really, really socially important. Sometimes that comes with a safety tinge ('this thing could kill us, and we need to make it so it doesn't'), sometimes with an accelerationist tinge ('this thing could liberate mankind from economic scarcity'), but either way it's usually stated with real conviction. If they only wanted money they'd go work for a hedge fund. But they also want to build something they're proud of. That character trait will, I think, cause the 'throw money at smart people until they all join' strategy that Zuckerberg is attempting to fail. If superintelligence is built, it will be built by a team that is productive due to a passionate, shared, optimistic vision for what a world with superintelligence will look like. It will be made because its makers want to change the world, not sell sugar water. A team of researchers joining primarily for the money, under a leader whose boldest vision is 'what if we sold more ads on sunglasses,' is not going to make it.
Yahoo
29 minutes ago
- Yahoo
MP Materials surges after smaller loss, Pentagon, Apple deals boost outlook
(Reuters) -MP Materials shares jumped more than 10% in premarket trading on Friday, after the U.S. rare earths miner posted higher output, a smaller-than-expected quarterly loss and forecast further production gains. The Las Vegas-based company operates the only U.S. rare earths mine and has seen its stock more than quadruple this year as Washington moves to secure domestic supplies and cut dependence on China. Production of neodymium and praseodymium (NdPr), the most in-demand rare earths, jumped nearly 120% in the second quarter to a record 597 metric tons. In the current quarter, NdPr oxide production is expected to rise 10% to 20% sequentially, Founder and Chief Operating Officer Michael Rosenthal said on a post-earnings call. Analysts at TD Cowen said strong upstream volumes and magnetics output drove a 40% EBITDA beat, with commercial magnet production on track by the year-end. They now expect the company to hit a 60,000-tonne annual run-rate of rare earth oxide output a year ahead of schedule. The results come weeks after MP signed a multibillion-dollar deal with the U.S. Department of Defense that guarantees a $110 per kg price floor for NdPr, nearly double the Chinese market levels, and could make the Pentagon its biggest shareholder. That was followed by a $500 million rare earth magnet supply agreement with Apple, whose prepayments will cover most of the cost to expand MP's Independence mine in Texas. Stage III magnetics operations are already EBITDA-positive and, alongside government and Apple funding, leave the company well-capitalized to accelerate growth, brokerage DA Davidson said. MP posted an adjusted loss of 13 cents per share for second quarter, while analysts estimated a 19-cent loss.
