• Insight Weekly
  • Posts
  • Spotlight : AI chips stoking the next Industrial Revolution

Spotlight : AI chips stoking the next Industrial Revolution

INSIGHT WEEKLY : March 10, 2024

An easy to read economic and financial summary. If the images do not load, click to download external images in your email to see the newsletter in full, or click the link above to read online.

šŸŒ Major indexes

Another four records set this week :

Franceā€™s CAC 40 set a new record this week. Along with other markets, optimism that inflation is set to come down has brought buyers back into the market.

Indiaā€™s Nifty 50 also set another record continuing the trend of confidence in the Indian economy.

See earlier spotlight on India - an emerging economic power

Japanā€™s Nikkei 225 also set a new record in the week with continued optimism before pulling back a little at the end of the week.

See earlier spotlight on Japan 

S&P 500 also set a new record during the week despite a pull back in tech stocks during the week

Spotlight : AI chips stoking the next Industrial Revolution

Potential for AI is huge and some pundits have dubbed the AI revolution as the next industrial revolution. The chips (semi conductors) are driving the revolution.

Regular chips are multi purpose and handle a variety of tasks. These are typical referred to as CPUā€™s (central processing units) and are commonly found in smart phones, computers, cars as well as in many appliances. These chips process tasks in turn (sequentially).

AI chips are designed specifically to process AI and machine learning. These chips are GPUā€™s (Graphic Processing Unit), TPUā€™s (Tensor Processing Unit) and ASICā€™s (Application Specific Integrated Circuits). AI chips handle tasks in parallel and so can handle multiple operations simultaneously. The most proficient multi tasker !

What do AI chips do ? These chips are designed for tasks that involve learning from data (machine learning) and performing calculations very quickly. For example, when you use natural language to interact with AI applications such as ChatGPT, these interactions require chips to have the ability to process information rapidly and provide quick responses. Furthermore, these chips are designed to efficiently manage and analyze huge volumes of data and images. Their construction supports the ability to handle multiple tasks at once (parallel processing) while also using energy in a very efficient manner.

Chip makers are in two categories - designers and manufacturers. The barriers to entry into the market are huge, and new entrants are unlikely to have a significant impact. A few companies dominate this market.

āœļøAI Chip designers :

Company

Headquarters

Metric (Approx. Market Share)

Focus Area

NVIDIA

Santa Clara, California, USA

40-45%

GPUs for training and inference

AMD

Santa Clara, California, USA

20-25%

GPUs and specialty AI chips (e.g., EPYC CPUs with AI features)

Intel

Santa Clara, California, USA

15-20%

Xeon CPUs with AI capabilities, neuromorphic chips (e.g., Loihi)

Suzhou Syntilante Technology

Jiangsu, China

5-10%

Cambrian series AI chips designed for training and inference

Habana Labs (acquired by Intel in 2019)

Mountain View, California, USA

5-10%

Goya training accelerators

This table highlights some of the leading players, but other companies are also developing AI chips, including:

  • Qualcomm (focus on mobile AI)

  • Samsung (focus on mobile AI and Exynos AI chips)

  • Graphcore (focus on Intelligence Processing Units - IPUs)

  • Cadence Design Systems (focus on AI design tools and IP)

  • ARM is a UK based designer who design part of the chip, mainly the processing cores

šŸ­ Biggest AI Chip manufacturers :

Company

HQ

Focus Area

Manufacturing Capability

Foundry Model

TSMC

Hsinchu, Taiwan

Logic chips, advanced process nodes (e.g., 5nm, 3nm)

Contract manufacture

Pure-play foundry

Samsung

Suwon, South Korea

Logic chips, memory chips (DRAM, NAND), advanced process nodes

Contract manufacture & In-house production

Integrated Device Manufacture (IDM)

Intel

Santa Clara, California, USA

Logic chips, memory chips, processors (CPUs)

Primarily in-house production with some foundry usage

IDM

SK Hynix

Icheon, South Korea

Memory chips (DRAM, NAND)

In-house production

IDM

Micron Technology

Boise, Idaho, USA

Memory chips (DRAM, NAND)

In-house production

IDM

United Microelectronics Corporation (UMC)

Hsinchu, Taiwan

Logic chips, mature process nodes

Contract manufacture

Pure-play foundry

Explanation of Columns:

  • Focus Area: The type of chips the company primarily manufactures (logic chips for processors, memory chips like DRAM and NAND, etc.)

  • Manufacturing Capability: Refers to the company's ability to produce chips using advanced manufacturing processes (measured in nanometers, nm). Smaller process nodes generally allow for cramming more transistors onto a chip, leading to increased performance and efficiency.

  • Foundry Model: Distinguishes between two main models:

    • Pure-play foundry: These companies manufacture chips designed by other companies (fabless chip designers). TSMC and UMC are leading examples.

    • Integrated Device Manufacturer (IDM): These companies both design and manufacture their own chips. Examples include Intel, Samsung, SK Hynix, and Micron. However, some IDMs may also utilize foundries for specific chip production needs.

How are chips manufactured ? By chip lithography in which a design for a circuit is transferred to a light sensitive material on the surface of a silicon wafer.

ASML is a Dutch company that does most of the lithography for chips and has over 80% market share overall and a monopoly on machines for advanced chip manufacturing. ASML sells these large complex machines to manufacturers.

Here's why ASML is so important in chip manufacturing:

  • Specialization in Photolithography: ASML designs and manufactures photolithography machines used in the production of nearly all integrated circuits. These machines are crucial for etching intricate patterns onto silicon wafers, forming the foundation of modern computer chips.

  • Cutting-Edge Technology: They are the sole supplier of extreme ultraviolet (EUV) lithography machines needed for manufacturing the most advanced chips. These EUV machines allow for the creation of even smaller and more powerful transistors on a chip.

šŸ“ˆ How have AI company stocks done so far this year ? To pick just a few : Nvidia (+80%), AMD (+49%), Intel (-8%) , ASML (+39%), TSMC (+32%)

šŸ›ļø US CHIPS and Science Act was passed in 2022 to reduce reliance on non-US manufacturing, grow a strong domestic industry, create jobs and protect national security. Restrictions are placed on funding recipients from expanding semi conductor manufacturing in China or any other country deemed to be a national security threat. Chips are a strategic geopolitical topic, considering the vast potential of AI.

šŸ’°What is the proposed $7 trillion fund raising ? Not a typo, it really is $7 trillion and not $7 billion !

Sam Altman, the Open AI chief, is hoping to raise $7 trillion to reshape the landscape of AI and chips. This mind boggling amount is being proposed to restructure the entire semiconductor industry to prepare for the expected huge demand for AI chips to power the next industrial revolution. It is time to think big, and he certainly is thinking big. There are over a trillion chips manufactured each year for use in our phones, cars, computers etc. In that context, $7 trillion could be a reasonable number !

šŸ‡¬šŸ‡§ UK

UK National Budget was delivered this week by the Chancellor of the Exchequer.

The widely signalled ā€œnon domsā€ tax status removal was announced in the budget. From 2025, overseas citizens living in the UK will pay tax on worldwide income and not just UK income after four years of living in the UK. This may put the UK at a competitive disadvantage in the search for foreign investment as some other countries have established non dom tax status.

The second widely signalled change was a cut in taxation but delivered by a rate reduction 2% in National Insurance payments rather than on income taxation. National Insurance deduction may be phased out in the future and simplified by a method of collecting tax revenue through income tax. The government is promoting this as ending the ā€œdouble taxationā€ of work but if this is going to be offset by an increased tax on savings and pensions, the amount of tax paid is not going to be less. By keeping the tax bands unchanged, the ā€œfiscal dragā€ is pulling in four million people into paying tax at the lower rate, and pushing three million into the higher tax bracket.

FTSE100 is -0.8% since the start of the year.

šŸ‡ŗšŸ‡ø US

S&P500 at 5,137 is up over 8% this year, and down 0.27% this week. There are some voices warning of a market correction, broadly defined as a reduction of between 10% and 20% but this does not seem to get much traction among investors. It is the tech sector that is driving market indexes higher and there is still confidence that this sector has more upside.

The US economy added 275,000 jobs, which was more than expected but the total unemployment rate moved up from 3.7% to 3.9%.

Fed chair Powell indicated in his appearance before Congress that policymakers are ā€œnot farā€ from having confidence that inflationā€™s downtrend will be sustained. So interest rate cuts are expected to follow in due course. Probably around end of Q2 ?

Treasury yields showed another slight decrease at the long end (10 year).

If you are liking this newsletter, please send this link to friends, family and colleagues and post on social media.

šŸ‡ÆšŸ‡µ Japan

Nikkei 225 was down 0.56% in the week. It is up over 19% so far this year. There is increasing confidence that the Bank of Japan will begin to increase rates from the current negative levels. Prices seem to be rising in line with wages, allowing the BOJ to start monetary tightening. Household spending declined and remains weak.

See previous spotlight on Japan.

šŸ…5ļøāƒ£ Billionaire Leaderboard

Change in week :

  1. Bernard Arnault and family (LVMH) $233bn ā¬†ļø $2bn

  2. Elon Musk (Tesla, SpaceX)$195bn  ā¬‡ļø $15bn (Tesla is down 29% so far this year)

  3. Jeff Bezos (Amazon) $194bn ā¬‡ļø $5bn

  4. Mark Zuckerberg  (Facebook/Meta) $171bn ā¬†ļø $1bn

  5. Larry Ellison (Oracle)  $139bn ā¬‡ļø $3bn

and watching

  1. Jensen Huang (Nvidia) $77bn ā¬†ļø $5bn. Nvidia is up 4% this week, and over 270% in the last 12 months. Will AI take him to the top ?

SPOTLIGHTS

Earlier :

This week : AI chips

Coming soon : More on AI, Cryptos, National debt, future growth sectors, Gold and more !

If you liked this newsletter, please send this link to friends, family and colleagues and post on social media. https://insight-weekly.beehiiv.com/subscribe

Stay tuned for more insights and updates in a 5 minute round up each week.