WHY IT’S IMPORTANT FOR THE UNITED STATES TO MAINTAIN ITS TECHNOLOGICAL EDGE
Ever hear of AIAC? I didn’t until about a year ago. In my continuing quest to learn as many irritating acronyms as possible, I figured it out. It stands for Artificial Intelligence Accelerator Chips. I figured out the Artificial Intelligence part, as I haven’t been living under a rock, but the Accelerator Chip part of the acronym got me doing more digging.
When I was in law school, which was during the period of when we had a peanut farmer for president, and people actually got along, one of my professors told me sometimes the best nuggets of information are contained in the footnotes- so read the footnotes.
IT’S A BIG MARKET
The chip market is massive and only growing. The global semiconductor market size was valued at $611.35 billion in 2023 and is projected to grow from $681.05 billion in 2024 to $2062.59 billion by 2032 (FN1).
THE GOOD, THE BAD, AND WORKING OUT THE BUGS
The integration of artificial intelligence into chips (creating AIAC) is a natural progression. There is a lot of internet chatter about the processing power AI will require. Think about it, smarter people have bigger heads…just kidding. AI, particularly AI working in the neural network and large sets of generative data, potentially can stress existing data centers and cloud environments. Adding more servers is not going to solve the problem and may even result in an intolerable latency effect grinding business operations down to a crawl. I don’t think you are as naive as me and look up into the sky and see clouds and think there is an inexhaustible supply and, therefore, think ‘why worry’. Much of the cloud as we know it is just off-site servers designed to run multiple workloads at the same time that leverage general-purpose CPU chips which is also potentially going to reach a tipping point (FN2).
AIAC CHIPS
The development of embedding AI into chips is designed to improve the performance of AI models by optimizing computation and reducing latency making them essential for high-speed data processing and real-time AI applications. The idea has been around since the early 1980’s with the advent of neural network processors but really took off in 2011, when some small company called Qualcomm (just kidding as they are somewhere north of $36 billion) introduced the first mobile AI processor, which paved the way for AI chips in mobile devices. AIAC can also potentially speed up the deployment of more complex models and the handling of larger datasets, which is essential for advancements in machine learning and deep learning technologies. The neat thing about AI is it can offer a myriad of customizable models where the designers can specify parameters and constraints and then choose the most efficient and effective designs. Heck, AIAC can probably get rid of the designers and do it by itself (just kidding). The possibilities are limitless. A chip for, say, manufacturing inventory analysis, a chip for development of new pharmaceuticals, a chip for pedestrian tracking, etc (FN3).
By focusing on generative AI accelerator chips and leveraging advancements in the semiconductor industry like new material science (FN4), the US can continue to drive innovation and maintain the competitive edge in the rapidly evolving AI landscape.
KEEP THE US AHEAD IN AIAC DEVELOPMENT 🇺🇸
Reasonable minds may differ, but I believe there is sufficient policy reasoning for the US to support the policy that AI accelerator chips be manufactured domestically.
We made good progress when Congress passed the CHIPS Act of 2022 (FN5) to strengthen domestic semiconductor manufacturing, design, and research. The intent is to enhance economic and national security as well as fortify our supply chains. The CHIPS act is needed as the US Chips manufacturing is somewhere around 12% today versus 37% in 1990. On the other hand, it is indisputable that the majority of US manufacturing is reliant on chips. I am old enough to have gone through the various cycles of “Go Global” to “Buy American” to “Go to Mars”, but if US businesses are so dependent on chips, domestically manufacturing them makes sense.
THE “C” WORD
The US has also imposed restrictions on exporting chips to that country that begins with the letter “C”… I can’t say it, as someone in that government will probably put me under surveillance. Oh heck, I don’t care. It’s China. We have a ways to go as countries and international businesses have found ways to beat the system. Shell companies and third country relaxed export controls are just some of the ways to circumvent the regulations (FN6).
There is discussion on tougher China trade sanctions (FN7).
I KNOW WHAT YOU ARE DOING! I am of the opinion that domestically produced chips reduce the risk of espionage and tampering, which are potential threats when sourcing from foreign suppliers. I am not sure if the reader has ever been to China, but it’s the one country where I turn off my phone and never use my computer. I am of the opinion that Chinese espionage is a very real threat, and having a US made chip in my phone, at the very least, sure makes me feel better.
USA! USA! Leading the world. There has never been a greater focus on developing economic competitiveness than in the past several years. Whether it’s clean energy, AI, or chips, keeping the US globally competitive strengthens our economy, creates jobs, spurs innovation and allows the US to continue to have the upper hand in international trade negotiations. With the integration of AI and chips, the phrase technological sovereignty is more important today than it has ever been. Maybe someday we can say we can get along with every country but we need to look internally first.
CONCLUSION
There are smarter people than me to figure out whether trade sanctions are working, but when you dumb it down, the CHIPS act can be a powerful carrot and trade sanctions can be a powerful stick. The devil is in the detail (FN8).
About the author
Gary Sumihiro is the founder of Sumihiro Investments, LLC a global strategic consulting firm. Learn more at www.sumihiroinvestmentsllc.com.
Footnotes:
FN1: Read More at: https://www.fortunebusinessinsights.com/semiconductor-market-102365
FN2: There is a good summary in this article https://www.foley.com/p/102j836/ai-requires-lots-of-power/
FN3: Talk to Chris Medina of nureal.ai who knows the possibilities far better than me:
About nureal.ai
Leading the democratization of artificial intelligence by offering a marketplace of affordable AI programs. Learn more at www.nureal.ai.
FN4: Unfortunately, a side effect of chip processing is the transistors in the chip cause heat to build up. Since they are all squished together, they are especially vulnerable to heat.The new AIAC chips are designed to deliver more computational power so guess what, there could be more heat. The structural housing to diminish the heat is basically thinking old school- let’s put bigger more directional fans to cool down the chips. However, there are also advances in material science to assist in heat reduction. IN the meantime, I keep a pocket fan near my computer… www.sciencedaily.com/releases/2021/06/210629191730.htm.
FN5: www.semiconductors.org/chips/
FN6: thediplomat.com/2024/06/4-ways-china-gets-around-us-ai-chip-restrictions/
FN8: I am always fascinated with where words and phrases come from. The phrase the devil is in the detail came from Le bon Dieu est dans le détail” (“the good God is in the detail”) which is generally attributed to Gustave Flaubert in the early 18th century.