What’s the current hullabaloo around the future of AI chips?

There is lot of hullabaloo around AI, and now, AI chips!

As per Malcolm Penn, CEO, Future Horizons,  AI will transform many industries and the way people work. But, the road to this end goal is currently shrouded in a fog of euphoric hype and hysteria.

AI is also not a product per se, like an iPhone or laptop! You cannot log into Amazon or Etsy and buy an AI!! Arguably, products like ChatGPT are more Artificial Plagiarism than Intelligence!

Just look at the development of AI! If it is that easy to develop and deploy, it just cannot be a wonder, at least, not for me! At best, it is a tool!

Let me also remind everyone that AI was first developed and demonstrated by Dr. Wally Rhines, former CEO, Mentor Graphics, now Siemens EDA.

Dr. Wally Rhines is credited to be the founder of artificial intelligence (AI) back in 1986. According to him, AI is not a new technology. Here is the cover of High Technology magazine in July 1986. Dr. Wally Rhines is the person on the left, and George Heilmeier, former head of DARPA, is on the right. “We tried hard in the 1980s, but the infrastructure had not developed to a level where AI would provide profitable opportunities,” he added.

Sam Altman, CEO of AI research lab OpenAI, made waves with a mind-blowing proposal: to raise $7 trillion to build a global AI infrastructure!

As per Tal Navarro, a serial entrepreneur, and owner of AIVibes.ai, it is perhaps, well known that AI is the future of business. But, it's not here to automate everything! AI is here to help us be more human. AI is not about replacing existing jobs! AI is about empowering the employees to focus on what they do best -- being human. That’s how AI can actually help businesses grow!

By making use of AI, we can:

* Personalize online experiences

* Respond to customer needs faster

* Improve decision-making

* Simplify complex business processes.

Benefits of AI chip design
By bringing together powerful combination of greater intelligence and speed in tackling otherwise repetitive tasks, AI-driven chip design can generate better silicon outcomes, and substantially enhanced engineering productivity.

There are variety of benefits of AI chip design, including:

* Enhanced PPA. AI can enhance PPA by taking on exploration of these large design spaces to identify areas for optimization.

* Enhanced productivity. By handling iterative tasks, AI frees engineers to focus on chip design differentiation and quality while meeting time-to-market targets.

* Support for reuse. AI drives even greater efficiencies into chip development processes.

* Faster design migration. With the support of AI, chip design teams can more quickly migrate their designs from one process node to another.

Energy impact of AI apps also looms large. AI design tools can reduce carbon footprint by optimizing AI processor chips (and workflows to design, verify, and test chips) for better energy efficiency.

How can AI chips work in future?
So, how can AI chips work in future? For starters, Nvidia’s AI chip dominance is being targeted by Google, Intel, Arm, Qualcomm, Samsung, Fujitsu, and other tech companies, as The Unified Acceleration Foundation (UXL). They are driving an open standard accelerator software ecosystem.

Brett Winton, ARK Invest Chief Futurist, said that by 2030, AI software companies will be needing trillions of dollars in AI chips. Apple will be licensing Google's Gemini model. GenAI will be coming soon to the Apple iPhone. GenAI Gemini from Google will be part of iOS18, beginning later this year. It will give Google about 2 billion devices to feature their LLMs. It will also give Apple new features that their users have been asking for. Apple has also talked about a deal with OpenAI.

Next, Anthropic is of extreme interest. It has AI research and products that put safety at the frontier. Its Claude 3 is now available. It is a friendly assistant, and fast, capable, and truly conversational. 

Tesla has perhaps, the most compelling AI story in the market. Advances in AI are going to make autonomous driving a reality. It is also working on robo-taxi and software to develop into a commercial product.

Entertainment can also be transformed using highly-compelling AI models in the future. Imagine generative content, and how compelling that would be for the end users. Meta also has the open sourcing of its AI operating system.

EDA firms advances in AI
Let us look at the work being done by the leading EDA firms in AI.

As per Cadence Design Systems, there are key benefits of AI in chip design. One, ability to optimize PPA of chips. Two, ability to automate certain chip design tasks. Three, bridging the gap in chip design talent. Reinforcement learning (RL) is an AI technology used for chip design. RL utilizes multiple chip floor designs to achieve the best PPA configurations.

Cadence Allegro X AI is a revolutionary system design technology that extends the Allegro X platform with the power of artificial intelligence to enable layout automation for small to medium-sized PCBs.

Cadence Virtuoso Studio leverages 30 years of industry knowledge and leadership in custom/analog design to provide broader support for systems, including RF, mixed-signal, photonics, and advanced heterogeneous designs. Innovative AI techniques, cloud enablement, infrastructure improvements, and integration across Cadence products complement these design flows.

Cadence Verisium AI-Driven Platform represents a generational shift from single-run, single-engine algorithms to algorithms that leverage big data and AI across multiple runs of multiple engines throughout an entire SoC verification campaign. Cadence Cerebrus Intelligent Chip Explorer is a revolutionary, machine learning-driven, automated approach to chip design flow optimization. Cadence Optimality Intelligent System Explorer is a multiphysics optimization software that enables the analysis and optimization realization of electronic systems.

Synopsys.ai Copilot, became the industry's first GenAI capability for chip design. Synopsys is bringing GenAI to semiconductor and electronics design.

Synopsys DSO.ai autonomously searches for optimization targets in a chip design’s very large solution spaces. VSO.ai autonomously achieves faster verification coverage closure and regression analysis for faster functional testing closure, higher coverage, and predictive bug detection. TSO.ai automatically searches for optimal solutions in large test search spaces to minimize pattern count and automatic test pattern generation (ATPG) turnaround time.

As part of Siemens Xcelerator portfolio, Siemens EDA's AI and ML solution is uniquely positioned to help companies leverage AI- and ML-powered EDA tools to deliver differentiated AI- and ML-driven innovations to market faster. Its Tessent products deploy three kinds of AI to improve automation software.

Don’t count out China!
Now, let us also look up China. Zeng Yi from Shenzhen-based China Electronics Corp., said his firm has a long way to go to catch up with the USA. There seems to be growing anxiety about a widening gap in AI across China. Leading Chinese tech firms, Huawei Technologies and ZTE, have invested heavily in R&D of AI chips.

Next, Prof. Zhou Jun and his team from the University of Electronic Science and Technology of China (UESTC), unveiled two ultra-low-power AI chips with record-breaking performance via algorithm and architectural optimization at IEEE International Solid-State Circuits Conference (ISSCC) 2024. The first of the two AI chips were to be embedded into smart devices to enable offline voice control.

Birth of SEIDA
All of this led to the creation of SEIDA. As per reports, some former executives from SiemensEDA established an electronic design automation (EDA) company in Hangzhou, China, called SEIDA. It is said to be headed by Liguo “Recoo” Zhang, formerly of SiemensEDA. SEIDA aims to “enable chip success.” SEIDA also plans to launch optical proximity correction (OPC) software sometime this year.

Malcolm Penn noted: “Indeed! The more the west pushes back, the more innovative China becomes SEIDA is a perfect example of why a containment strategy never works. It’s like trying to slay a hydra! You cut off one head, and two more emerge!”

Jaswinder Ahuja, Corporate VP and MD, Cadence Design India, said on the importance of optical proximity correction (OPC) for EDA: “OPC is extremely important for the manufacture-ability, and becomes even more important as we progress to smaller geometries.” 

Let's see how that goes!

US far ahead?
In a report, titled AI chips: what they are, and why they matter, from Center for Security and Emerging Technology, USA, the authors, Saif M. Khan and Alexander Mann, noted that US companies dominate AI chip design, with Chinese companies far behind in AI chip designs. China is reliant on US EDA software to design AI chips, and need US and allied SME and fabs to fabricate AI chips based on these designs. 

The value of state-of-the-art AI chips, combined with the concentration of their supply chains in the USA and allied countries, presents a point of leverage for the United States and its allies to ensure beneficial development and adoption of AI technologies.

Perhaps, an AI chip that is 1,000x as efficient as a CPU for a given node provides an improvement equivalent to 26 years of CPU improvements! Also, leading node AI chips are increasingly necessary for cost-effective, fast training and inference of AI algorithms. The efficiency of state-of-the-art AI chips also translates into cost-effectiveness.

Market status
Nvidia and AMD have duopoly over the world GPU design market. China’s GPU company, Jingjia Microelectronics, fields slower GPUs. Xilinx and Intel dominate the global FPGA market. China’s leading FPGA companies -- Efinix, Gowin Semiconductor, and Shenzhen Pango Microsystem have developed trailing node FPGAs.

AI ASIC market is distributed with lower barriers to entry, as ASICs and inference chips are easier to design. Google, Tesla, and Amazon are designing AI ASICs specialized for own AI applications. Next, Google’s TPU is a leading commercial AI ASIC. Intel is developing powerful commercial AI ASICs

Chinese firms in AI ASIC include Baidu, Alibaba, Tencent, HiSilicon (owned by Huawei), Cambricon Technologies, Intellifusion, and Horizon Robotics. Chinese players are said to be limited to inference, although Huawei announced development of an AI training ASIC.

Experts also disagree on the need for leading nodes for AI chips. Everybody who wants to do AI, needs the performance, power, and form factor of 7nm and below. Hence, the birth of SEIDA, in China, for developing EDA tools.

A semiconductor researcher at Hong Kong Applied Science and Technology Institute said: “For AI chips, manufacturing costs will be much lower if you use 28nm technology and not 10 or 14nm tech. You need to spend a lot of effort from scratch [to design at leading nodes]—mathematical models, physical layers, computational language -- all of these will need investment!”

Also, it was reported that only few fabs are capable of manufacturing near-state-of-the-art AI chips. Only approximately 8.5% of global fab capacity could be used to fabricate near-state-of-the-art AI chips.

There is much more coming in AI and AI chips development! Watch this space!