Mozilla and Firefox have undergone a slew of changes of late, with the non-profit laying off its entire advocacy team and subsequently working with a famous design house on a branding overhaul. At the same time, Google is facing a monumental US antitrust lawsuit that could see the search giant prohibited from paying to be the default search engine on a variety of platforms. Speculation until now has predicted that if Google loses the antitrust case, it would deprive Mozilla and Firefox of a massive revenue stream, effectively ending the Firefox browser as we know it.

Perhaps in preparation for this imminent change, Mozilla has just announced a new search partnership with Ecosia, the search engine that prioritizes privacy and environmental concerns. Mozilla didn't explain what the partnership entails, but the official announcement seems to suggest that Ecosia won't become the default search engine for Firefox, although the blog post does encourage switching to Ecosia as the default search engine. Ecosia advertises that it is transparent when it comes to its earnings, and it donates all of its profits to environmental conservation efforts. It also embodies many of the privacy protections that Firefox has historically been known for. A more cynical reading of the situation might suggest that this is simply part of Mozilla trying to save face after its recent layoffs and a round of bad news.

Imagine being able to step into any world in an instant—from the bustling streets of New York City to the snowy mountain tops of the Alps. These worlds are no longer merely places to observe from afar. They can now be explored and interacted with simply through a gaze, a gesture or your voice. What once seemed like science fiction has become a reality as eXtended reality (XR) transforms how we engage with the world around us.

XR is an umbrella term for technologies that use digital elements to extend or alter reality by merging the physical and digital worlds together. This includes virtual reality (VR), augmented reality (AR), mixed reality (MR) as well as other similar technologies yet to be developed. XR offers infinite possibilities, creating a dynamic spatial canvas where users' sight, sound and motion combine to interact with the outside world. A world that will unlock unprecedented experiences across core areas of life, from working and learning to entertainment, gaming and even health and wellness.

Google Quantum AI lab has announced a huge advancement in quantum computing with its new Willow processor. The chip demonstrated remarkable error correction capabilities and computational power beyond traditional supercomputers, including ExaFLOP machines such as Frontier and El Capitan. In research published in Nature, Google's team showed that Willow can exponentially reduce error rates as more qubits are added to the system—a feat known as operating "below threshold" that has been a major challenge in quantum computing since 1995. Using arrays of quantum bits in increasingly larger grids, the team successfully cut error rates in half with each expansion. The chip's performance is particularly recorded in random circuit sampling (RCS), where Willow completed calculations in under five minutes. Something like that would take today's fastest supercomputer approximately ten septillion years to solve—a timespan far greater than the universe's age.

Manufactured at Google's specialized facility in Santa Barbara, the 105-qubit Willow chip also achieved impressive coherence times, with qubits maintaining their quantum states for up to 100 microseconds—five times longer than previous generations. Dr. Hartmut Neven, who founded Google Quantum AI in 2012, emphasized that the breakthrough brings quantum computing closer to scaling into more complex systems for data processing. Potential applications include discovering new medicines, designing more efficient batteries for electric vehicles, and accelerating fusion energy research. The next challenge for Google's quantum team is demonstrating a "useful, beyond-classical" computation that addresses practical applications. While Willow has shown superior performance in benchmark tests, researchers are now focused on developing algorithms that can tackle commercially relevant problems that are impossible for traditional computers to solve.

For better or worse, generative AI has been a disruptive force in many industries, although its reception in video games has been lukewarm at best, with attempts at integrating AI-powered NPCs into games failing to impress most gamers. Now, Google's DeepMind AI has a new model called Genie 2, which can supposedly be used to generate "action-controllable, playable, 3D environments for training and evaluating embodied agents." All the environments generated by Genie 2 can supposedly be interacted with, whether by a human piloting a character with a mouse and keyboard or an AI-controlled NPC, although it's unclear what the behind-the-scenes code and optimizations look like, both aspects of which will be key to any real-world applications of the tech. Google says worlds created by Genie 2 can simulate consequences of actions in addition to the world itself, all in real-time. This means that when a player interacts with a world generated by Genie 2, the AI will respond with what its model suggests is the result of that action (like stepping on a leaf resulting in the destruction of said leaf). This extends to things like lighting, reflections, and physics, with Google showing off some impressively accurate water, volumetric effects, and accurate gravity.

In a demo video, Google showed a number of different AI-generated worlds, each with their own interactive characters, from a spaceship interior being explored by an astronaut to a robot taking a stroll in a futuristic cyberpunk urban environment, and even a sailboat sailing over water and a cowboy riding through some grassy plains on horseback. What's perhaps most interesting about Genie 2's generated environments is that Genie has apparently given each world a different perspective and camera control scheme. Some of the examples shown are first-person, while others are third-person with the camera either locked to the character or free-floating around the character. Of course, being generative AI, there is some weirdness, and Google clearly chose its demo clips carefully to avoid graphical anomalies from taking center stage. What's more, at least a few clips seem to very strongly resemble worlds from popular video games, Assassin's Creed, Red Dead Redemption, Sony's Horizon franchise, and what appears to be a mix of various sci-fi games, including Warframe, Destiny, Mass Effect, and Subnautica. This isn't surprising, since the worlds Google used to showcase the AI are all generated with an image and text prompt as inputs, and, given what Google says it used as training data used, it seems likely that gaming clips from those games made it into the AI model's training data.

Your favorite shows, movies, series, live TV, and games are in your space on the screen of your AGON by AOC Q27G4XY SMART GAMING MONITOR. The AGON by AOC Q27G4XY 27" SMART GAMING MONITOR is a pioneer in this field, bringing the convenience of Google TV to your monitor screen. This 27" QHD 2560x1440 WLED monitor, featuring a 180 Hz refresh rate, 1 ms GtG response time, AMD FreeSync, NVIDIA G-SYNC, and HDR10, combined with Google TV, delivers an immersive gaming and streaming experience—ideal for any dorm room!

Google TV
When you're done gaming, there's no need to switch devices. The AGON by AOC Q27G4XY Smart Gaming Monitor, powered by Google TV's built-in operating system, brings all your entertainment to one place. Instantly browse movies, shows, live TV, sports, and more across your favorite apps. Google TV brings together everything you stream on one screen, so you can find what to watch without jumping from app to app. Not sure what to watch? Just ask Google Assistant and search across apps with your voice to find what you're looking for. Finding your next watch is now as easy as ever.

The CHERI Alliance CIC (Community Interest Company) today announced its official launch and the expansion of its membership, welcoming Chevin Technology (UK), Critical Technologies (USA), the Defence Science and Technology Laboratory (DSTL, UK), Google (USA), Light Momentum Technology Corporation (Taiwan), National Cyber Security Centre (NCSC, a part of GCHQ, UK), Parvat Infotech (India), SRI International (USA), TechWorks (UK), Trusted Computer Center of Excellence (USA), the University of Birmingham (UK), and the University of Glasgow (UK) as founding members.

Founded to unite hardware security leaders and system developers, the CHERI Alliance aims to establish CHERI (Capability Hardware Enhanced RISC Instructions) as the new standard for memory safety and scalable software compartmentalization.

During Alphabet's Q3 earnings call, CEO Sundar Pichai announced that AI now generates more than a quarter of the company's new code, marking a significant milestone for AI advancement and for the tech giant. This development comes alongside impressive financial results, with the company reporting $88.2 billion in revenue, representing a 15% year-over-year increase. Implementing AI in code generation has raised concerns, though Google maintains rigorous safety protocols. Every AI-generated code segment undergoes thorough review by human (natural intelligence) engineers before deployment, ensuring quality and security standards are met. This hybrid approach helps Google balance productivity with reliability. The tech giant's commitment to AI development extends beyond code generation.

Recent achievements include the revolutionary AI Overviews feature, which has undergone significant optimization. Through optimizing hardware solutions and technical improvements, Google has managed to reduce query costs by over 90% while simultaneously doubling the capacity of their custom Gemini models. Google's AI push has also garnered prestigious recognition, with DeepMind researchers Demis Hassabis and John Jumper receiving the Nobel Prize in Chemistry for their groundbreaking AlphaFold project. Former Google researcher Geoff Hinton also achieved Nobel recognition in Physics. The impact of Google's AI integration is evident across its product ecosystem, with Gemini models now powering seven platforms that each serve over two billion monthly users. Google Maps recently joined this elite group, while the company has expanded its AI capabilities to external developers through partnerships with platforms like GitHub Copilot to help developers write code with AI assistance.

ASUS today announced ZenScreen Smart MS27UC, the brand's first Google TV smart monitor. Google TV brings together movies, shows, and more from a variety of streaming services and offers users a personalized entertainment experience. ZenScreen Smart MS27UC boasts a 27-inch 4K (3840 x 2160) IPS panel with wide 178° viewing angles, a 99% sRGB gamut, VESA DisplayHDR 400 certification, and dual front-facing 5 W Harman Kardon-certified speakers. Plus, a source control feature enables viewers to easily switch between Productivity or Play modes.

With ZenScreen Smart MS27UC, there's no need for a dedicated PC. Users can conduct online meetings and browse content via the web browser, and logging in to a Google Account allows them to create and sync work and documents on Google Workspace. The monitor also offers seamless connections to multiple devices, including PCs and smartphones through Miracast or Google Cast, with the multi-control feature enabling seamless navigation via a single Bluetooth mouse and keyboard.

Ultra Accelerator Link (UALink ) Consortium, led by Board Members from AMD, Amazon Web Services (AWS), Astera Labs, Cisco, Google, Hewlett Packard Enterprise (HPE), Intel, Meta and Microsoft, have announced the incorporation of the Consortium and are extending an invitation for membership to the community. The UALink Promoter Group was founded in May 2024 to define a high-speed, low-latency interconnect for scale-up communications between accelerators and switches in AI pods & clusters. "The UALink standard defines high-speed and low latency communication for scale-up AI systems in data centers"

Google is developing an advanced AI system capable of autonomously operating web browsers and controlling computer functions, according to recent reports. Codenamed "Project Jarvis," this technology is expected to debut alongside Google's upcoming Gemini large language model this December. The AI-powered system aims to transform everyday online tasks by handling everything from complex research projects to routine shopping activities. Unlike traditional AI assistants that operate within confined parameters, Project Jarvis would have direct control over browser functions, enabling it to navigate websites, fill forms, and execute commands independently. The development follows a broader industry trend toward AI systems with expanded capabilities, with Anthropic recently demonstrating similar technology allowing AI models to execute tasks directly on users' computers.

For businesses, such advances could simplify operations by automating repetitive tasks and increasing productivity, while individual users might benefit from delegating time-consuming activities like price comparisons or travel bookings to an AI assistant. However, this level of system access raises significant security and privacy concerns. Granting AI systems direct control over personal computers creates new security challenges requiring robust safeguards. Unrestricted access to users' systems could create new vulnerabilities, potentially enabling sophisticated cyberattacks. This heightened risk necessitates extensive security testing and robust safeguards before the technology can be deployed with confidence.

Details of what is claimed to be Google's upcoming Tensor G5 and G6 SoCs have popped up over on Notebookcheck.net and the site claims to have found the specs on a public platform, without going into any further details. Those that were betting on the Tensor G5—codenamed Laguna—delivering vastly improved performance over the Tensor G4, are likely to be disappointed, at least on the CPU side of things. As previous rumours have suggested, the chip is expected to be manufactured by TSMC, using its N3E process node, but the Tensor G5 will retain the single Arm Cortex-X4 core, although it will see a slight upgrade to five Cortex-A725 cores vs. the three Cortex-A720 cores of the Tensor G4. The G5 loses two Cortex-A520 cores in favour of the extra Cortex-A725 cores. The Cortex-X4 will also remain clocked at the same peak 3.1 GHz as that of the Tensor G4.

Interestingly it looks like Google will drop the Arm Mali GPU in favour of an Imagination Technologies DXT GPU, although the specs listed by Notebookcheck doesn't add up with any of the specs listed by Imagination Technologies. The G5 will continue to support 4x 16-bit LPDDR5 or LPDDR5X memory chips, but Google has added support for UFS 4.0 memory, something that's been a point of complaint for the Tensor G4. Other new additions is support for 10 Gbps USB 3.2 Gen 2 and PCI Express 4.0. Some improvements to the camera logic has also been made, with support for up to 200 Megapixel sensors or 108 Megapixels with zero shutter lag, but if Google will use such a camera or not is anyone's guess at this point in time.

Opera, the Norwegian browser company, is releasing Opera One R2, the new major version of its AI-powered browser. Building upon the modular design and integrated AI capabilities of Opera One, which was released in 2023, the new Opera One R2 introduces a host of new features and enhancements, making it the most powerful and best-looking Opera browser to date.

Opera One R2 features Opera's latest AI innovations. It also introduces dynamic themes that allow users to personalize their browsing experience with a static or animated interface, sound effects, and music.

Last week, we got a preview of Microsoft's Azure production-ready NVIDIA "Blackwell" GB200 system, showing that only a third of the rack that goes in the data center is actually holding the compute elements, with the other two-thirds holding the cooling compartment to cool down the immense heat output from tens of GB200 GPUs. Today, Google is showing off a part of its own infrastructure ahead of the Google Cloud App Dev & Infrastructure Summit, taking place on October 30, digitally as an event. Shown below are two racks standing side by side, connecting NVIDIA "Blackwell" GB200 NVL cards with the rest of the Google infrastructure. Unlike Microsoft Azure, Google Cloud uses a different data center design in its facilities.

There is one rack with power distribution units, networking switches, and cooling distribution units, all connected to the compute rack, which houses power supplies, GPUs, and CPU servers. Networking equipment is present, and it connects to Google's "global" data center network, which is Google's own data center fabric. We are not sure what is the fabric connection of choice between these racks; as for optimal performance, NVIDIA recommends InfiniBand (Mellanox acquisition). However, given that Google's infrastructure is set up differently, there may be Ethernet switches present. Interestingly, Google's design of GB200 racks differs from Azure's, as it uses additional rack space to distribute the coolant to its local heat exchangers, i.e., coolers. We are curious to see if Google releases more information on infrastructure, as it has been known as the infrastructure king because of its ability to scale and keep everything organized.

Intel Corp. (INTC) and AMD (NASDAQ: AMD) today announced the creation of an x86 ecosystem advisory group bringing together technology leaders to shape the future of the world's most widely used computing architecture. x86 is uniquely positioned to meet customers' emerging needs by delivering superior performance and seamless interoperability across hardware and software platforms. The group will focus on identifying new ways to expand the x86 ecosystem by enabling compatibility across platforms, simplifying software development, and providing developers with a platform to identify architectural needs and features to create innovative and scalable solutions for the future.

For over four decades, x86 has served as the bedrock of modern computing, establishing itself as the preferred architecture in data centers and PCs worldwide. In today's evolving landscape - characterized by dynamic AI workloads, custom chiplets, and advancements in 3D packaging and system architectures - the importance of a robust and expanding x86 ecosystem is more crucial than ever.

BenQ today premiered the GP520, a groundbreaking 4K all-in-one home entertainment projector designed specifically for modern open-plan living spaces. Perfect for young families who value quality time together, the GP520 delivers the onboard cinema experience of Netflix and Google TV with premium 2600 ANSI-lumen HDR picture quality and ambient adaptive technology, coupled with smart auto setup for stunning viewing quality, regardless of the wall color, lighting conditions, or room layout.

"Families often struggle to achieve high-quality projection in open living spaces," said Peter Huang, President of BenQ Corporation. "The GP520 addresses three critical issues: color distortion from ambient light, optimization to different wall colors, and auto-setup without constant adjustment when the projector is moved into different corners. With the GP520, families can enjoy flawless 4K visuals no matter where or how it is set up."

ViewSonic Corp., a global leader in visual solutions, unveils two new projectors - the X1-4K Pro and X2-4K Pro. Designed to elevate the console gaming experience, these projectors deliver images with QHD (1440p) resolution at a lightning-fast 120 Hz refresh rate, ensuring ultra-smooth gameplay and detailed visuals on a screen larger than 100 inches. Complementing the native 4K UHD visual performance, built-in Harman Kardon speakers provide cinema-quality sound, while the integrated Google TV provides seamless access to a broad spectrum of streaming options, creating a truly immersive and all-encompassing home entertainment experience.

"Compared to traditional gaming displays, projectors offer an unmatched immersive experience and greater setup flexibility, so it's no surprise that they're becoming increasingly popular in console gaming," said Dean Tsai, General Manager of Projector & LED Display Business Unit at ViewSonic. "Reflecting this trend, last year we introduced the world's first 'Designed for Xbox' gaming projectors, the X1-4K and X2-4K, which set new industry standards and have consistently received positive feedback from the market. We are now excited to launch their upgrades, the X1-4K Pro and X2-4K Pro. These models not only enhance visuals but also feature a built-in Google TV system, providing a comprehensive solution that caters to a wide range of home entertainment needs."

Following Epic Games's recent victory over Apple, which effectively forced the iPhone maker to allow third-party apps on iOS in the EU, the game developer and distributor has won a similar case against Google, except in the US, this time. The ruling places several restrictions on what Google can and can't do when it comes to the Android operating system and how it positions its own apps and services on Android phones, including prohibiting Google from paying device manufacturers to include its Play software on their devices and forcing developers to use Google's payment platform.

The ruling (PDF) and its consequent restrictions will start on November 1, 2024, will last three years, and Google isn't particularly happy about the mandated changes, having already confirmed that it will appeal the decision, citing security concerns and arguing that the order would negatively affect developers: "these changes would put consumers' privacy and security at risk, make it harder for developers to promote their apps, and reduce competition on devices."

Formovie Technology, a leading player in the world of display technology, has announced the Formovie Cinema Edge.This ultra-short throw (UST) powerhouse features a deep and sharp 4K resolution and combines Google TV with Netflix for seamless streaming. Users can enjoy rich visuals and immersive Dolby Audio for an incredible home theater experience—all at an affordable price.

Formovie Cinema Edge comes with native Netflix support and is available at an affordable price point of $2,199 / €2,199 pre-order price at Formovie's website. The projector displays images up to 150 inches, features 110% Rec. 709 color reproduction, and boasts a contrast ratio of 3000:1. This allows for easy and seamless streaming of content without the need for additional devices. Moreover, its Google TV certification and integrated Google Assistant add convenience for everyday use.

Today, Google has announced that it will discontinue all of its Chromecast products, as the company is transitioning to the Google TV Streamer, which offers a very different form factor to its Chromecast products. Over the past 11 years, Google has sold some 100 million Chromecasts over a handful of different generations of products with quite different feature sets. The most recent addition launched in 2022 and Google has promised to continue to support the latest generations of Chromecast devices for now, although the company didn't say for how long.

The Google TV Streamer moves away from the HDMI dongle design, which some might see as a mistake by the company, but those wanting a set-top-box type device will on the other hand be pleased. The white (porcelain) or grey (hazel) slate isn't the most attractive design maybe, somewhat due to the choice of colours by Google, but it's what's on the inside that counts, right? Although Google didn't reveal the CPU used, the company claims the Google TV Streamer is 22 percent faster than the Chromecast with Google TV, comes with 4 GB of RAM and more importantly, it has 32 GB of built-in storage, compared to 8 GB for its latest generation of Chromecast devices.

Apple has disclosed that its newly announced Apple Intelligence features were developed using Google's Tensor Processing Units (TPUs) rather than NVIDIA's widely adopted hardware accelerators like H100. This unexpected choice was detailed in an official Apple research paper, shedding light on the company's approach to AI development. The paper outlines how systems equipped with Google's TPUv4 and TPUv5 chips played a crucial role in creating Apple Foundation Models (AFMs). These models, including AFM-server and AFM-on-device, are designed to power both online and offline Apple Intelligence features introduced at WWDC 2024. For the training of the 6.4 billion parameter AFM-server, Apple's largest language model, the company utilized an impressive array of 8,192 TPUv4 chips, provisioned as 8×1024 chip slices. The training process involved a three-stage approach, processing a total of 7.4 trillion tokens. Meanwhile, the more compact 3 billion parameter AFM-on-device model, optimized for on-device processing, was trained using 2,048 TPUv5p chips.

Apple's training data came from various sources, including the Applebot web crawler and licensed high-quality datasets. The company also incorporated carefully selected code, math, and public datasets to enhance the models' capabilities. Benchmark results shared in the paper suggest that both AFM-server and AFM-on-device excel in areas such as Instruction Following, Tool Use, and Writing, positioning Apple as a strong contender in the AI race despite its relatively late entry. However, Apple's penetration tactic into the AI market is much more complex than any other AI competitor. Given Apple's massive user base and millions of devices compatible with Apple Intelligence, the AFM has the potential to change user interaction with devices for good, especially for everyday tasks. Hence, refining AI models for these tasks is critical before massive deployment. Another unexpected feature is transparency from Apple, a company typically known for its secrecy. The AI boom is changing some of Apple's ways, and revealing these inner workings is always interesting.

The market for artificial intelligence (AI) platforms software grew at a rapid pace in 2023 and is projected to maintain its remarkable momentum, driven by the increasing adoption of AI across many industries. A new International Data Corporation (IDC) forecast shows that worldwide revenue for AI platforms software will grow to $153.0 billion in 2028 with a compound annual growth rate (CAGR) of 40.6% over the 2023-2028 forecast period.

"The AI platforms market shows no signs of slowing down. Rapid innovations in generative AI is changing how companies think about their products, how they develop and deploy AI applications, and how they leverage technology themselves for reinventing their business models and competitive positioning," said Ritu Jyoti, group vice president and general manager of IDC's Artificial Intelligence, Automation, Data and Analytics research. "IDC expects this upward trajectory will continue to accelerate with the emergence of unified platforms for predictive and generative AI that supports interoperating APIs, ecosystem extensibility, and responsible AI adoption at scale."

According to The Information, OpenAI is reportedly in talks with Broadcom about developing a custom AI accelerator to power OpenAI's growing demand for high-performance solutions. Broadcom is a fabless chip designer known for a wide range of silicon solutions spanning from networking, PCIe, SSD controllers, and PHYs all the way up to custom ASICs. The latter part is what OpenAI wants to focus on, but all the aforementioned IP developed by Broadcom is of use in a data center. Suppose OpenAI decides to use Broadcom's solutions. In that case, the fabless silicon designer offers a complete vertical stack of products for inter-system communication using various protocols such as PCIe, system-to-system communication using Ethernet networking with Broadcom Tomahawk 6 and future revisions, alongside storage solutions and many other complimentary elements of a data center.

As a company skilled in making various IPs, it also makes ASIC solutions for other companies and has assisted Google in the making of its Tensor Processing Unit (TPU), which is now in its sixth generation. Google TPUs are massively successful as Google deploys millions of them and provides AI solutions to billions of users across the globe. Now, OpenAI wants to be part of the AI chip game, and Broadcom could come to the rescue with its already-established AI success and various other data center componentry to help make a custom AI accelerator to power OpenAI's infrastructure needed for the next generation of AI models. With each new AI model released by OpenAI, compute demand spikes by several orders of magnitude, and having an AI accelerator that exactly matches their need will help the company move faster and run even bigger AI models.

Samsung Electronics today announced its all-new Galaxy Z Fold6 and Galaxy Z Flip6, along with Galaxy Buds3 and Galaxy Buds3 Pro at Galaxy Unpacked in Paris.

Earlier this year, Samsung ushered in the era of mobile AI through the power of Galaxy AI. With the introduction of the new Galaxy Z series, Samsung is opening the next chapter of Galaxy AI by leveraging its most versatile and flexible form factor perfectly designed to enable a range of unique mobile experiences. Whether using Galaxy Z Fold's large screen, Galaxy Z Flip's FlexWindow or making the most of the iconic FlexMode, Galaxy Z Fold6 and Flip6 will provide more opportunities to maximize AI capabilities. Built on the foundation of Samsung's history of form factor innovation, Galaxy AI uses powerful, intelligent, and durable foldable experience to accelerate a new era of communication, productivity, and creativity.

Opera GX, the browser for gamers, is bringing a significant update to the browser's built-in AI, Aria. This update provides users with the latest AI features that Opera has been releasing as part of their experimental AI Feature Drops program in the Developer stream of the Opera One browser. The features that are arriving to Opera GX increase Aria's capabilities by introducing image generation and understanding, voice output, a chat summary option, and links to sources.

Image Generation and Voice Output
Images are crucial to today's web, so this Opera GX update places a strong focus on the visual. With this update, Aria is gaining the ability to turn text prompts and descriptions into unique images using the image generation model Imagen2 by Google. Aria identifies the user's intention to generate an image based on conversational prompts. They can also use the "regenerate" option to have Aria come up with a new image. Aria allows them to generate 30 images per day per user.

A new startup emerged out of stealth mode today to power the next generation of generative AI. Etched is a company that makes an application-specific integrated circuit (ASIC) to process "Transformers." The transformer is an architecture for designing deep learning models developed by Google and is now the powerhouse behind models like OpenAI's GPT-4o in ChatGPT, Anthropic Claude, Google Gemini, and Meta's Llama family. Etched wanted to create an ASIC for processing only the transformer models, making a chip called Sohu. The claim is Sohu outperforms NVIDIA's latest and greatest by an entire order of magnitude. Where a server configuration with eight NVIDIA H100 GPU clusters pushes Llama-3 70B models at 25,000 tokens per second, and the latest eight B200 "Blackwell" GPU cluster pushes 43,000 tokens/s, the eight Sohu clusters manage to output 500,000 tokens per second.

Why is this important? Not only does the ASIC outperform Hopper by 20x and Blackwell by 10x, but it also serves so many tokens per second that it enables an entirely new fleet of AI applications requiring real-time output. The Sohu architecture is so efficient that 90% of the FLOPS can be used, while traditional GPUs boast a 30-40% FLOP utilization rate. This translates into inefficiency and waste of power, which Etched hopes to solve by building an accelerator dedicated to power transformers (the "T" in GPT) at massive scales. Given that the frontier model development costs more than one billion US dollars, and hardware costs are measured in tens of billions of US Dollars, having an accelerator dedicated to powering a specific application can help advance AI faster. AI researchers often say that "scale is all you need" (resembling the legendary "attention is all you need" paper), and Etched wants to build on that.