Raj Yavatkar

IBM + Intel Corporation => A New Enterprise AI Platform // AI Episode 10 Attn: Arvind Krishna + Pat Gelsinger A couple of days back, Qualcomm reached out to Intel for a prospective merger…The synergies seem low and the odds are not that great. But, but, but… What about IBM acquiring Intel ? Synergies: IBM’s strong innovation culture Deep technology foundation Integrated enterprise solutions and services Hybrid cloud data centers and AI platforms Intel’s X86 footprint Evolving architecture for desktop AI Weakened but still viable chip innovation Foundational brilliance The technology firmament is shifting as heavenly waters artificially surge towards creation of new islands for enterprise choice and utility… Microsoft + OpenAI Oracle +/- AWS Nvidia +/- everybody… This is the time for IBM to spring for Intel… Personally, I am rooting for Intel to find its mojo back and return to its former glory… IBM with a capitalization of $200B, can easily digest Intel which is struggling at $94B. The combination could create a $1T company where enterprise solutions for AI run on Intel’s AI-advantaged chips on desktops, supported by AI technology and infrastructure in IBM’s data centers A solution that would have corporate CEOs stop sweating and start lining up to buy a new premium offering that would leverage their captive data for AI-based advantage… Hello, IBM and Intel Corporation it is time to dance together…

5

https://lnkd.in/dpXAHnej Intel has built an amazing business and ecosystem with its x86 design and fabrication. It is seems that the same moat which defended its core has prevented it from going outside with new things as seen with :mobile, graphic, software , AI (as recently posted it passed being OpenAI chips supplier) and foundery. Drastic move is needed and coming and it will pop existential question: does intel keep its design or foundery, seems that one of them must go in order to prevent the sink of both, I am not sure which since both are facing hard challenges at the same time.

7

1 Comment

In case you still doubt what #AI can do 🤔 Just listen to Jensen Huang's Mumbai AI Summit. From #Enterprises to #Industries, AI is spreading its footprint everywhere. What matters is how quickly you adopt it. 👉 Watch AI's real potentials at ⏱ 42:30 - 45:30. AI is certainly not going anywhere; embrace it ASAP. https://lnkd.in/g-P_7fgy #aiadoption #lifewithoutai #lifewithai

5

Today, the Board has made the decision to discontinue our semiconductor business due to financial challenges and the inability to agree on a restructuring program. While disheartened by this development, I appreciate the support received, especially from Shivam Yadav and Nirmal Sharma , PhD. This setback won't deter me, and I am committed to developing a new plan that aligns with the Board's vision. #BusinessDecision #MovingForward

11

"A study published in Opto-Electronic Science discusses high-intensity spatial-mode steerable frequency up-converter toward on-chip integration. Integrated photonic devices consisting of micro-lasers, amplifiers, optical waveguides, frequency converters, and modulators on a single chip, enabling control over photon's spatial modes, frequencies, angular momenta, and phases, are essential for preparing high-dimensional quantum entangled states, high-capacity photon information processing, all-optical communication, and miniaturization of photonic computing. However, current nonlinear waveguide devices, integrating spatial modes and photon frequency conversions, heavily rely on external optical path control and spatial light modulators, failing to meet the crucial requirement of on-chip integration for photonic devices." #photonics

1

Creation of backbone based on research, hiring of an expat with intrigue and focus, international collaborations and partnerships, mergers and acquisitions, consolidation and leading a transformation from a paddy economy to the silicon powerhouse. An endeavour of four decades with changing complexion of geopolitics, technology, business and disruptions. As they say, "Change is the only Constant". Sharing my piece chronicling TSMC in the Economic Times today. India Semiconductor Mission India Electronics and Semiconductor Association Anoop Verma Harsh V. Pant Sana Hashmi, PhD Mahaveer Singhvi Sameer Patil, PhD https://lnkd.in/gQU_r7ks

64

The high-end graphics cards from Nvidia, such as the H100, have catalyzed the emergence of dedicated Rent-A-GPU farms, expanding beyond traditional hyperscale providers like AWS, GCP, and Azure. Initially, supply chain limitations coupled with high demand enabled these specialized farms to purchase these cards—each costing nearly $40,000—in bulk and offer them for rent. By late 2023, rates were around $8 per hour under commitments ranging from one to three years. Today, the landscape has shifted significantly. Some of the largest Rent-A-GPU providers have adapted to changing market dynamics by offering GPU spot markets with commitments as short as 48 hours, with rates as low as $1.8 per hour or even less. This suggests that the supply chain for high-end GPUs has finally caught up with demand, potentially heralding a substantial decrease in costs not only in the rental market but also in capital expenditures for owning these GPUs. For the first time, there is downward pressure on the prices of high-end GPUs. This development is particularly promising for real-world applications of Generation AI. The significant reduction in inference costs is crucial for the viability and value creation of GenAI within enterprises. It represents the first of many indicators that we are approaching the rapid realization of game-changing use cases, transforming how businesses leverage advanced AI capabilities. At KamiwazaAI we believe the start of an 5th Industrial Revolution Enterprise #5IR is around #1trillionInferences a day or in the range of ~$3b dollars a day utilizing chat GPT4 at .003 per Inference (100 tokens). A 1000x decrease in the cost of Inference is only 18 months away or so and its time NOW to begin implementing the technology to keep up!

21

5 Comments

No Vendor lockins, No Proprietary Stuff, Freedom in using the Cloud, Data Sovereignity is the way to go! And of course lower TCO and better quality of service. Computing should be a utility, simple and easy to use, not just for techies. This is what we are set out to do at People+ai #OpenCloudCompute #OCC

19

2 Comments

I'm always happy and proud to hear our early investors at REFASHIOND Ventures: The Industrial Transformation Fund talk about the work they are doing. Each of them is having real impact in building supply chains for tomorrow. In this podcast Zac talks about the underlying infrastructure for what what I describe as Information and Data Supply Chains. This is worth listening to if you want to get into some of the issues that will determine the future of Physical Supply Chains. #TWSCF #TNYSCM | The past ran on supply chains. The present runs on supply chains. The future will run on supply chains. The world is a supply chain.®

4

Challenges of AI Networking for hyperscalers and enterprises Sept 27, 2024 I had the opportunity to meet with Manohar Paluri, VP of AI at Meta, this evening at the AI Tech Leaders Club, just under 24 hours before the AI Infra Summit on September 27, 2024. It’s been confirmed that the large training cluster at 500K GPUs is the next frontier for the hyperscaler industry. Meanwhile, on the enterprise side, standalone cluster sizes typically range from 20K to much smaller setups for retraining and inference. From my experience, while the challenges in these two sectors are intertwined, they can still be stated separately. 1) Checkpointing Challenges: Large GPU clusters require numerous checkpoints to maintain system reliability. The industry has recently implemented networking and checkpointing solutions for clusters with around 20K GPUs. Imagine the challenges we’ll face at the scale of 500K GPUs! 2) GPU Utilization: Typical GPU utilization hovers around 20%. While hyperscalers may achieve better numbers, overall utilization remains low. For this, a startup came up a clever solution to power down the idle GPUs dynamically to save up to 60% of power consumption from a GPU cluster. This approach could either reduce power usage from idle GPUs or, in conjunction with increased GPU utilization through further resources sharing—like memory, network, and model partitioning/parallelization—enhance overall efficiency. 3) Deployment Flexibility: The deployment of data center networks (hyperscaler cloud) or private networks (on-premises) for enterprises often lacks the necessary flexibility. For more insight, I recommend reading the justification for Google’s Optical Circuit Switching (OCS) network. Following conventional practices without innovative solutions can lead to over-investment in AI and cloud infrastructure. 4) ROI Optimization: Classic cloud infrastructure is often not optimized for ROI in AI networking, which leads to operational challenges for enterprises. To understand this, a little involved discussion is needed. Once understood, it becomes part of your instinct. Consider the competitive disadvantage faced by businesses that optimize ROI in their AI networks. Free Takeaway from me: Conversations about the challenges of AI infrastructure and networking are very welcome at the AI Infra Summit, and participants will receive the latest AI networking problem slide! 😊

12

5 Comments

Optical I/O Shatters Data Bottlenecks 🔨 Ayar Labs resolves the issue of circuit power and offers high bandwidth, low power, and low latency connectivity between compute elements. #data #bandwith #AI #photonics

1

Please check out my two talks below.

25

1 Comment

My latest article on leveraging GenAI & LLMs for Semicon Chip designing. #smartsoc #genai #llm #soc #semicon

62

2 Comments

Thanks People+ai! Check out the video below to know what I am up to these days! I am building hardware to make AI accessible to all!

59

5 Comments

As #AI scales across industries and use cases, understanding how to measure AI performance becomes key in developing solutions with sufficient head room. Like we say, "not all TOPs are created equal". At Qualcomm, we focus on running AI inferencing workloads on our super-efficient and powerful Neural Processing Units (NPUs). Highly recommend reading our latest article on how to measure AI workloads, performance and the right place to run them.  https://lnkd.in/gzV33PCa

75

On stage at #Huddle2024 is Jyothis Indirabhai of Netrasemi - the #EdgeAI chip maker. His session is about the evolution of Edge AI and how On-Device processing in #IoT industry is being exponentially improved by increasing effeciancy and reducing cost. Netrasemi is India’s first Edge AI chip maker, and is a promising #startup from #Trivandrum, #Kerala. Good to see such cutting edge technology solution creators in #Kerala.

114

3 Comments

Mindgrove Technologies has achieved a new milestone for Indian #semiconductors with the release of the first #microprocessor owned, designed, and marketed entirely from #India. The #SecureIoT #microcontroller integrates several complex units designed to ensure #high #performance, #robust #security, and #versatile #functionality for various IoT applications. This 64-bit #RISCV single-core processor is capable of speeds up to 700 MHz while maintaining a peak power consumption of under 200 mW. The Secure IoT #SoC also allows Indian #OEM to cut costs by about 30% compared to international counterparts while maintaining, or even enhancing, the functionality and performance of their products. This cost efficiency opens up new opportunities for local businesses and #startups to #innovate within the #IoT space without the financial burden of expensive chip imports. A significant achievement for the Indian semiconductor industry, RISC-V's flexibility and accessibility have enabled Mindgrove Technologies to tailor the #SoC to specific market needs and security requirements without the constraints typically imposed by #ISA #royalties and restrictions. Link: https://lnkd.in/g8CkjHRg

18