Artificial Intelligence - Page 10

NVIDIA is pushing to make its best AI semiconductors in the USA, but still requires Taiwanese companies for advanced packaging, and more. Now, NVIDIA CEO Jensen Huang has personally thanked Taiwanese partners, because without them, the big "Made in USA" push would be a much harder pursuit.

Speaking with Fox News, NVIDIA CEO Jensen Huang said slowly but surely, the US is becoming a technology manufacturing hub for the world, with Jensen saying that this achievement of "Made in USA" isn't a job that NVIDIA can do on its own on American soil.

Jensen said that within 9 months, with the help of its partnership with TSMC, which has been an "incredible partner" for the United States, Foxconn, Wistron, Amkor, and SPIL, NVIDIA has been able to manufacture the most advanced AI chip in the world -- Blackwell -- "completely in the United States and Arizona".

The Content Overseas Distribution Association (CODA) is a Japanese anti-piracy and copyright-related organization that represents a large number of game development studios and publishers, including Square Enix, Bandai Namco, and FromSoftware - creators of some of the most influential and recognized gaming franchises. Final Fantasy, Dragon Quest, Tekken, Pac-Man, Dark Souls, Elden Ring, and more.

And on October 27, CODA submitted a written request to OpenAI regarding its Sora 2 AI video generator. "CODA has confirmed that a large portion of content produced by Sora 2 closely resembles Japanese content or images," the announcement says. "CODA has determined that this is the result of using Japanese content as machine learning data."

The organization is requesting that OpenAI respond "sincerely such that both the healthy development of AI technology and the protection of rightsholders and creators' rights are ensured."

Apple is yet to release the next generation of Siri that it marketed with the previous generation of iPhone, and now we are hearing the Cupertino company will be leaning on Google's Gemini AI model to improve the voice assistant.

Apple's next generation of Siri is meant to help the company catch up to the likes of OpenAI and other AI companies that have created products such as ChatGPT, which far exceed the capabilities of Siri. Apple has been lagging behind in the integration of AI into its products and services, and in order to catch up to its competitors, the company is looking to integrate already established models such as Google's Gemini, Anthropic's Claude, and OpenAI's ChatGPT.

Notably, ChatGPT is already available on Apple's AI-compatible devices, but the OpenAI product isn't fully integrated into the iPhone, meaning the service isn't accessing personal user data such as contacts, messages, etc. For Apple to fulfill what it used as a selling point for the previous generation of iPhone, the selected AI will give Siri the power to access personal user data, broadening the capabilities of the voice assistant and enabling users to request much more refined information from it.

Amazon and OpenAI have just announced a new partnership, where Amazon Web Services (AWS) will be one of the primary compute providers to the AI startup, hosting NVIDIA's GB200 and GB300 AI servers.

The multi-year partnership between AWS and OpenAI will see OpenAI getting access to a pool of Amazon's NVIDIA GB200 and GB300 AI servers, in a deal worth around $38 billion, and will span 7 years. OpenAI gets access to NVIDIA GB200 and GB300 AI servers which is important, as all planned capacity is expected to be deployed for use by the end of 2026.

This means that OpenAI will have even more access to the best NVIDIA AI GPU hardware by next year, helping it continue to expand its AI operations and services. OpenAI has been signing new deals left, right, and center, including with US tech giants NVIDIA, AMD, Microsoft, Broadcom, and Oracle, and now adding Amazon to that growing list.

NVIDIA will be sending over 60,000 of its Blackwell AI GPUs to the United Arab Emirates (UAE) in a $15 billion deal with Microsoft, with approval for the AI chips coming from the Trump administration.

The UAE has been spending billions of dollars to make itself a global AI hub, using its close relations with the US government to get its hands on US semiconductor technologies, including the most advanced AI GPUs. Microsoft Vice Chair and President Brad Smith, said in an interview: "The biggest share of (the investment), by far, both looking back and looking forward, is the expansion of AI data centres across the UAE".

On the sidelines of the ADIPEC energy conference in Abu Dhabi, Smith added: "From our perspective, it's an investment that is critical to meet the demand here for the use of AI".

President Trump has said that NVIDIA's most advanced Blackwell AI GPUs won't be allowed to be given to any other country that is outside of the United States, keeping China's hands off bleeding-edge AI silicon.

In a new interview between President Trump and CBS on 60 Minutes, as well as comments to reporters on-board Air Force One, President Trump said that only US customers should have access to the best Blackwell chips from NVIDIA. President Trump told CNBC: "The most advanced, we will not let anybody have them other than the United States. We don't give (the Blackwell) chip to other people".

We might see tighter US restrictions around bleeding-edge US semiconductors, more so than previous reports have suggested, with China and most of the world banned from accessing NVIDIA's best AI silicon. However, just last Friday, NVIDIA said it would be supplying over 260,000+ Blackwell AI GPUs to South Korea, and to some of the country's biggest businesses, including SK hynix and Samsung, both of which supply HBM memory to NVIDIA.

SK hynix has unveiled its next-generation technology roadmap, including HBM5, HBM5E, GDDR7-next, DDR6, and 400+ layer 4D NAND and more to come for 2029-2031.

During the SK AI Summit 2025 event held in Seoul, South Korea, on November 3, the company unveiled its next-gen DRAM memory and NAND flash product roadmap. This included discussion on next-gen technologies including custom HBM memory, HBF, and a slew of AI-optimized products of the future.

SK hynix's new DRAM and NAND roadmaps cover 2026 to 2028, and then 2029 to 2031, where SK hynix will have next-gen HBM4 16-Hi stacks, as well as even faster HBM4E in 8/12/16-Hi stacks between 2026-2028. SK hynix will also have custom HBM4E memory, LPDDR6 memory, and new AI-D series memory with LPDDR5X SOCAMM 2, MRDIMM Gen2, LPDDR5R, and 2nd Gen CXL LPDDR6 memory for conventional DRAM.

SK hynix CEO Kwak Noh-Jung has just announced the company's new vision of its "Full Stack AI Memory Creator" at the SK AI Summit 2025 event, hosted in Seoul, South Korea, on November 3.

SK hynix is in a fierce memory battle with fellow South Korean memory maker Samsung, in providing the most -- and the best -- HBM memory for AI chips to companies like NVIDIA and AMD. SK hynix dominated the HBM3 and HBM3E supply to NVIDIA for its Hopper and Blackwell GPUs, but Samsung is catching up quite fast when it comes to next-gen HBM4, so SK hynix is jumping out ahead highlighting its new Full Stack AI Memory Creator vision.

SK hynix CEO Kwak Noh-Jung said: "SK hynix has been playing the role of a Full Stack Memory Provider by supplying products aligned with customer needs and time. Moving forward, the company aims to exceed customers' expectation by actively collaborating within the ecosystem and by solving the customers' hurdles together. We will become a creator who builds "Full Stack AI Memory" as a co-architect, partner, and eco-contributor".

YouTuber "PewDiePie" has been busy building a monster PC system and playing around creating his own AI service using Chinese open-source models, and modded NVIDIA GeForce RTX 4090 48GB graphics cards. Check it out:

PewDiePie has effectively created a 10-way GPU cluster as a "mini-datacenter" that uses PCIe bifurcation to spread out the PCIe lanes to handle all of the GPUs. The YouTuber was playing around by hosting models like Llama 70B, but realized that he has enough GPUs and VRAM capacity to run a much bigger 240 billion parameter AI model.

This is when he discovered Baidu's Qwen open-source AI model, creating a totally private, self-hosted AI system. This is where things take a hilarious turn... as PewDiePie assigned each one of the 8 GPUs as a different council member, with different personalities. He said that at this point, it's a democratic process, where he would consult his council, all giving a different answer, and then they vote. He wanted it to be better, so he created a new system that automatically generates new council members, different ones, so that it's constantly improving.

Samsung has finally inked its important HBM memory deal with NVIDIA, where it will supply its next-gen HBM4 memory with up to 11Gbps HBM4 memory chips to use on NVIDIA's next-gen Rubin AI GPUs in 2026.

After what has felt like forever in trying to get its HBM3 and HBM3E memory certified for use by NVIDIA, Samsung and NVIDIA have gotten closer than ever, working on HBM4 together as well as working on a new AI factory that's powered by 50,000+ of NVIDIA's AI GPUs.

Samsung and NVIDIA are working together on HBM4, with Samsung explaining on its press release that with incredibly high bandwidth and energy efficiency, Samsung's advanced HBM solutions are expected to help accelerate the development of future AI applications, and form a critical foundation for manufacturing infrastructure driven by these technologies.

Samsung and NVIDIA have just announced plans to build a new AI factory, with the new state-of-the-art AI factory combining Samsung's semiconductor technologies with NVIDIA platforms to establish the foundation of next-gen, AI-driven production.

The new AI factory will be powered by over 50,000 NVIDIA GPUs, where Samsung's new semiconductor AI factory to be the centerpiece of the company's digital transformation, integrating accelerated computing directly into full-fledged advanced chip manufacturing. Samsung and NVIDIA will be collaborating in other businesses, including next-gen HBM4, EDA tools, AI-RAN, and more.

Samsung plans to use the AI megafactory to integrate its labyrinthine manufacturing processes into "a single intelligent network, where AI continuously analyzes, predicts, and optimizes production environments in real time". Samsung will also be building digital twins to digitally visualize fab operations, where it will be using the "virtual environments to identify anomalies, perform predictive maintenance, and optimize production before changes are applied in the physical world".

Apple isn't looking at moving to TSMC's bleeding-edge A16 process node anytime soon, even faster than the company had TSMC fab its new A20 and A20 Pro chips on its new 2nm process.

In the future, Apple is poised to use TSMC's even newer A14 process node, but NVIDIA looks to be the first A16 customer for TSMC on its next-gen Feynman AI GPUs in 2027. TSMC's cutting-edge A16 process node is the first from the semiconductor giant to use a backside power supply delivery network (BSPDN) on its new 2nm process (A16).

In a new report from Korean outlet EBN, an official in charge of semiconductor design at a major North American big tech company met with reports, where they said: "currently, NVIDIA is the only customer for TSMC's 'A16'" and that "NVIDIA and TSMC are testing together". The official continued talking about the new A16 node, adding: "It will be applied around 2027. I understand that Apple's mobile application processor (AP) is not yet in talks with TSMC".

According to a local news report from WBAL-TV 11 News, an "artificial intelligence detector" called the cops on a student at a Baltimore County school for mistaking a bag of Doritos for a gun. Apparently, the student was outside his high school, eating some Doritos and chatting to friends when the Omnilert AI system notified the police that a gun-carrying individual was on school property.

'Omnilert: AI Gun Detection & Emergency Response Technology' is a security and proactive system that automatically notifies law enforcement and emergency services when it detects firearms within a "fraction of a second." According to Omnilert's official website, there's a human verification step with one of its certified operators before a call is placed to the police, so it's unclear how a bag of chips could have been mistaken for a gun.

The student in question, Taki Allen, told reporters that "eight cop cars" pulled up with officers pointing their guns at him, asking him to get on the ground, on the suspicion that he was carrying a firearm. "They made me get on my knees, put my hands behind my back, and cuffed me," Allen said. "Then, they searched me and they figured out I had nothing."

Samsung has just announced it will begin mass production of its new 1c DRAM-based next-gen HBM4 memory, as well as new 24Gb GDDR7 memory dies, and 128GB+ DDR5 products in 2026.

The company announced its recent Q3 2025 earnings, with a 15.4% increase in revenue over Q2 2025, with a new all-time high in quarterly sales from its Memory business, thanks to strong demand for HBM3E memory and server SSDs in the continuing wave of AI demand.

Samsung recently showed off its new HBM4 memory, offering up to 11Gbps bandwidth per IC, and should be featured inside of the next-generation of AI hardware from NVIDIA and AMD in the upcoming Rubin and Instinct MI400 series AI GPUs.

NVIDIA has officially shown off its next-gen Vera Rubin Superchip at GTC in Washington, the first time in the flesh, ready with next-gen chips and HBM4 memory for the next dominator in AI chips.

NVIDIA CEO Jensen Huang first showcased the company's next-generation Vera Rubin Superchip, but this is the very first time an actual sample of the motherboard, which NVIDIA calls the Superchip, has been shown. We have a Vera Arm-based CPU joined by not one but two gigantic Rubin GPUs. NVIDIA's motherboard for the Vera Rubin Superchip features LPDDR system memory that gets combined with the on-board HBM4 memory on both Rubin GPUs.

Jensen mentioned that NVIDIA's new Rubin GPUs are back in the NVIDIA labs, which means what he showed off was directly from the semiconductor fab at TSMC in Taiwan. Each Rubin GPU is surrounded with a bunch of power circuitry, with each of the Rubin GPU chips sporting 8 HBM4 sites, and two Reticle-sized GPU dies. The Vera CPU features 88 cores and 176 threads of custom Arm-based CPU cores.

AMD has just inked a deal with the US Department of Energy (DoE) on two new supercomputer projects using AMD hardware.

AMD will be building out two new supercomputers -- the Lux and Discovery supercomputers -- for the US DoE that will be used mostly for academic purposes. In a press release, AMD explained that when fully deployed, the new Lux and Discovery supercomputers will represent a combined $1 billion investment of private and public funding, enabling the DoE to build a secure, federated and standards-based infrastructure for sovereign US-based AI and science.

The new Lux AI supercomputer is co-developed between AMD, ORNL, Oracle Cloud Infrastructure (OCI) and HPE, powered by AMD Instinct MI355X AI GPUs, AMD EPYC CPUs, and AMD Pensando advanced networking technologies. The new Discovery AI supercomputer deepens the collaboration between DoE, ORNL, HPE, and AMD with next-gen AMD "Venice" GPUs, and AMD Instinct MI430X GPUs -- a new MI400 series AI accelerator engineered specifically for sovereign AI and scientific computing.

Apple has pledged more than $600 billion in US investments with some of the fruits of that labor now being seen, as the company has started shipping its new American-made AI servers into Apple's data center operations.

In a new post on X, Apple CEO Tim Cook said: "Apple's American-made advanced servers are now shipping from our new Houston facility to Apple data centers! These servers will help power Private Cloud Compute and Apple Intelligence, as part of our $600 billion US commitment".

The new American-made Apple servers will power things like Private Cloud Compute, a computational hierarchy where simple AI tasks are performed using on-device computational resources -- like on your iPhone -- while more complex tasks are offloaded onto Apple's private cloud servers using encrypted and stateless data. Meanwhile, Apple Intelligence handles everything else on your iDevices, including a retooled Siri agent.

Elon Musk has said that Tesla's next-generation AI5 processor is an incredible 40x faster -- yes, not 40% but 40 times faster -- than the current AI4 processor.

Tesla recently inked a $16.5 billion deal with Samsung Foundry to fab its next-next-gen AI6 processor, with reports suggesting Samsung is making a loss on the deal but considers it a "blessing in disguise", with more on that in the links below.

But during Tesla's recent Q3 2025 earnings call, Musk said that both Samsung and TSMC would be fabbing its AI5 chips, where he added that Samsung's new Taylor, Texas semiconductor fab is "more advanced" when it comes to semiconductor fabrication than TSMC's new Arizona fab.

Clippy is back! Well, no, the much-maligned assistant isn't, but a new Windows 11 take on a long line of desktop-based helpers has been revealed by Microsoft: it's called Mico.

Mico is part of Copilot in Windows 11 - the name is a contraction of Microsoft Copilot, as in Mi-Co - and the idea is to have a more humanlike assistant in said long line of helpers (from Rover the Dog of Microsoft Bob fame, through to Clippy, then Cortana in more modern times).

It is, if you like, a face for Copilot. As Microsoft explains: "The new Mico character is expressive, customizable, and warm. This optional visual presence listens, reacts, and even changes colors to reflect your interactions, making voice conversations feel more natural. Mico shows support through animation and expressions, creating a friendly and engaging experience."

Foxconn has already started preparing for NVIDIA's next-gen Vera Rubin AI platform, just as NVIDIA is pumping out its new Blackwell Ultra GB300 AI servers.

NVIDIA's next-gen Vera Rubin AI family of chips will be a huge release, as the entire tech stack is being upgraded, including next-gen HBM4 memory. NVIDIA will continue using the same rack configuration, but there will be huge increases in power, performance, and everything in between.

In a new report from Taiwan Economic Daily, we're hearing that one of NVIDIA's largest partners -- Foxconn -- has already started development on the next-generation Vera Rubin NVL144 MGX servers, with mass production aiming for the second half of 2026.