Eugene Zetserov’s Post

Groq, a leading #AI chip #startup, has reached an impressive valuation of 2.8 billion, setting its sights on outpacing NVIDIA in specific fields. The innovative Language Processing Units or #LPUs present a game-changing alternative to #GPUs, offering superior speed, cost efficiency, and energy optimization. While GPUs are known for their versatility, LPUs are finely tuned for language tasks, potentially transforming AI and #LLMs. I’m eager to see where this leads and if it really reshapes the AI landscape 🤔

Congratulations Cognifiber! We want to thank our investors for sharing the vision we have for Cognifiber. Cognifiber raises $5 million to develop processors that perform AI tasks 1,000 times faster than Nvidia. The Israeli startup is developing DeepLight processors (a new class of photonic processors) that utilize multi-core optical fibers to process data at the speed of light as it travels through the fiber, aiming to increase processing speeds up to 1,000 times during AI inference while using only about 1% of the energy. Super news for Green AI Data Centres. #PhotonicComputing #AI #GreenDataCentres #ComputingInfrastructure #OpticalCommunication #Photonics

FOLDSTRUCT is glad to be joining the NVIDIA Inception program for exceling AI startups. Recent advancements in generative #AI are a breeding ground for unleashing creative ingenuity in #construction. Foldstruct leverages this new force to empower the planners and engineers of tomorrow (literally tomorrow) to create better buildings. How? We use #GPT, #diffusion and #parametric logic to automate and improve the planning process, creating green, iconic and cost efficient buildings. Stay tuned for more updates! We see Nvidia as a natural partner when talking about stretching the boundaries of computation in the field of #AEC and look forward to challenging the most complex of chips in the benefit of the built environment. #contech #poptech #innovation

At this year’s #InnoVEX #COMPUTEX I sat for a great talk, and luckily the talk is publicly available on YouTube (The link is below). Within this talk mentioned a startup called Eva, an #alternativeAI computation device company. First of all, why pay attention to alternative AI devices, when what we have today seems to work well and still vastly untapped (i.e. silicon G/NPU’s, Nvidia and its competitions)? It’s because they require lots and lots of hard to scale energy, to train and to infer, to shuffle data between memory and processors, etc. While the best-known neural net (i.e. the human brain) “can operate on a bowl of rice.” The way the analog computing works is to use physics and physical devices to do matric operations AS VOLTAGE FLOW THROUGH. Think wire crossings makes a matrix grid, with a variable/programmable resistor at each intersection. The neural model can live as non-volatile memory resistor values in this matrix grid, as it’s trained and as it’s used. The instant the electricity flow through this matrix, the output is already the answer. Physics has done all the otherwise intensive floating-point, vector calculations in an instant. The power of this approach truly comes to being when models necessarily get bigger and bigger, when math outscale the Moore’s law for “conventional” computations, physical grids can scale with the demands and for a lot less energy. What Eva’s done is to take this analog computing to practicality. Its tech allows higher voltage, thus faster speed, to program these resistors. The protonic resistors literally and physically move ions to change resistivity (and because it’s physical mobility, once in place the values are robust from usage and require no additional energy to sustain). This is kinda like E Ink, another MIT invention. Keynote by Jerry Yang (HCVC, not YouTube) @ InnoVEX "Exploring the Future of AI Forum". Mentioned Eva but also a very interesting talk on its own.  https://lnkd.in/gUy_5gQY Disclaimer: I have no affiliations with anyone mentioned above. Wrote this up just thought it was cool and while still fresh in mind. 

AI startup Cerebras unveils the WSE-3, the largest chip yet for #generativeAI 4 TRILLION TRANSISTORS ON ONE SINGLE #CHIP!!! 🤯🤯🤩🤩 The #chip, the size of a single #semiconductor #wafer, has doubled performance to handle large language models in the tens of trillions of parameters. Cerebras Systems released the WSE-2 in April 2021. Its successor, the WSE-3, is designed for training AI models, meaning refining their neural weights, or parameters, to optimize their functionality before they are put into production. "It's twice the performance, same power draw, same price, so this would be a true Moore's Law step, and we haven't seen that in a long time in our industry," Cerebras co-founder and CEO Andrew Feldman said in a press briefing for the #chip, referring to the decades-old rule that chip circuitry doubles roughly every 18 months. The WSE-3 doubles the rate of instructions carried out, from 62.5 petaFLOPs to 125 petaFLOPS. One petaFLOP refers to 1,000,000,000,000,000 (1 quadrillion) floating-point operations per second. The size of almost an entire 12-inch wafer, like its predecessor, the WSE-3 has shrunk its transistors from 7 nanometers -- seven billionths of a meter -- to 5 nanometers, boosting the transistor count from 2.6 trillion transistors in WSE-2 to 4 trillion. TSMC, the world's largest contract chipmaker, is manufacturing the WSE-3. A big thank you to Tiernan Ray and ZDNET for the full article with more background and insights via the link below 💡🙏👇 https://lnkd.in/gn4489bS #semiconductormanufacturing #technology #semiconductor #cloud #hpc #computer #chip #datacenter #ai #artificialintelligence #asml #euv #taiwan

Generative AI is the definitive technology of the decade. However, on the hardware front, NVIDIA has made the most of the AI obsession. Their graphical processing units (GPUs) are perfectly suited for efficient AI processing, propelling them to become the most valuable tech company due to surging demand. Now, rivals rivals want in on the action. Nvidia faces a two-pronged challenge: established competitors, some of whom are also major customers, and ambitious startups like Groq. Despite continued purchases of Nvidia's silicon, these companies are actively developing their own AI-specific chips. Read more: https://lnkd.in/dyrg9eBM #Nvidia #AI #ArtificialIntelligence #AIchip #Technology #TechNews

Eliyan raises $60M for chiplet interconnects that speed up AI chips Eliyan has raised $60 million in funding for its chiplet interconnect technology that speeds up the processing for AI chips. Samsung Catalyst Fund and Tiger Global Management both led the round to help the team address the challenges of development of generative AI chips. Driven by demand for AI chips, industry forecasters are calling for robust growth in the high-bandwidth memory (HBM) sector, as much as 331% increase this year, followed by 124% in 2025, according to market researcher Arete Research #aichips #chiplets #interconnectivity

🌟February Start-up of the Month🌟 Introducing Lumai, our standout start-up for February! 🚀Lumai is enabling AI deployment at scale with the fastest, most energy-efficient AI processor for datacentres. The company is solving the challenge of how to meet the ever-increasing AI performance demand while minimising energy consumption. Lumai’s unique 3D optical technology is taking AI processing way beyond the limitations of both silicon-based GPUs and photonic solutions. With its outstanding power efficiency, Lumai significantly reduces the operating costs of running AI processors. 🎓 The company is an Oxford University spinout, and the technology is based on many years of ground-breaking research. The Lumai team brings together outstanding technical knowledge alongside commercial and scaleup expertise. Lumai in the news: 🎥 Featured in Silicon Semiconductor's recent video 🔍 Recognized as one of the AI start-ups to watch in 2024 by Tech.eu 👉 Watch and read about these via links in the comments 👇 For more information, visit https://lumai.co.uk/ For more tech insights, head to our blog https://humand.co.uk/blog/ #Lumai #Technology #StartupOfTheMonth #AI #Startups

Great read on how chipmaker Nvidia has become dominant in using artificial intelligence to design drugs, from which it’s built a $1 billion-plus revenue stream. In addition to selling chips, they also sell software to the industry and have a venture capital arm which invests in many AI startups in biotech, enabling them to get further entrenched. They’ve expanded from GPUs designed for video games to simulating how small molecules and proteins moved and interacted which then led to further advances. Kimberly Powell, Nvidia’s vice president of healthcare has a great quote, “Once you have models, you can start learning more about their interactions.” This completely resonates as someone who’s used models to generate previously hidden insights. https://lnkd.in/eFMEWh4A #technology #biotech #insights

#Semiconductor technologies evolving to meet #generativeAI demand Power and processor-hungry #AI set to drive next stage of #chip development – Generative #artificialintelligence (AI) is redefining the value of work, and its exponential need for performance is driving demand for power consumption and processing capacity in ways that it is not yet clear how this will be met. Tools based on generative AI have moved beyond niche technical applications to consumer use, changing the way images, video, music, and other content are produced. The impact is expected to reach almost every professional role, including #lawyers, #architects, #doctors and more. The AI #server market is expected to expand at a compound annual growth rate of 36% from 2023 to 2029, outpacing overall #datacenter growth. Thanks again to Yole Group for the full article with more background , charts and insights via the link below 💡🙏👇 https://lnkd.in/evN2XkmD #semiconductormanufacturing #semiconductorindustry #technology #it #tech #chip #chips #technology #tech #innovation #ic #icdesign #dram