[Industry Trends 2] What Jobs Will Survive in 10 Years?

Today, I will be posting the second part of the industry trends series—focusing on AI and semiconductors.

It is impossible to discuss current industry trends without mentioning AI and semiconductors, which is why I am providing this explanation.If you understand today’s topic well, it will also be of great help when discussing the industry with your children in the future.

For students who are currently in middle and high school, I will explain what jobs they might have in ten years and which companies they should consider applying to—using AI and semiconductors as a reference.

As industries evolve, the leading companies of each era also change. Companies that continue to thrive have successfully adapted by transforming their industrial focus in line with the times.

The group of companies on the far left consists of energy, distribution, and financial companies, which were once among the top 10 companies.

After a decade, the landscape changed significantly, and the companies in the far-right group represent the top 10 companies of 2025.

Big tech companies have secured dominant positions, and among them are semiconductor-related companies such as NVIDIA, Broadcom, and TSMC, which I will discuss today.

Since AI, the hottest industry today, cannot be discussed without semiconductors, it is no exaggeration to say that AI and semiconductors are now shaping the overall market.

The term "commodity" originally refers to raw materials, but it also has other meanings.

A commodity can refer to raw materials themselves, products that have undergone processing, or essential goods and services.

Previously, commodities included oil, gold, iron ore, grains, coffee, and sugar.

However, in modern times, semiconductors are found in nearly every electrical product, leading them to be regarded as the commodity of the modern era.

To reach the final discussion of today’s topic, we first need to briefly go through the eight semiconductor manufacturing processes.

The 8 Semiconductor Manufacturing Processes

I will reference Samsung's data to explain these processes. Since today’s topic revolves around which companies to apply to in 10 years and what jobs to consider, I will connect each process to the leading companies in that field.

1. Wafer Manufacturing

To produce semiconductors, they must first be made on wafers.

2. Oxidation

Once wafers are created, an oxide layer needs to be formed.

• The two best companies in wafer production: SUMCO and Shin-Etsu.

• Companies specializing in oxide layer formation: Applied Materials and Tokyo Electron (TEL, Japan).

3. Photolithography

In this step, masks are placed over wafers, and circuits are drawn onto semiconductors using photolithography.

• The dominant player in this field: ASML.

• Another significant company: Nikon.

4. Etching Process

This is the process of carving out the circuits that were drawn in the previous step.

• Leading companies: Lam Research and Tokyo Electron (TEL, Japan).

5. Deposition and Ion Implantation

A thin film is formed over the etched circuit.

• Companies excelling in this process: Applied Materials, Lam Research, and Motorios.

6. Metal Wiring Process

To ensure the periodic flow of electricity, metal wires are connected to the etched sections.

• Companies specializing in this: Applied Materials and Tokyo Electron (TEL, Japan).

Among the eight processes, the first six are known as "front-end processes", and while some American companies are involved, Japanese companies hold significant influence in this area.

7. EDS (Electrical Die Sorting) Process

This process involves testing whether the semiconductor is functioning correctly and whether electricity is flowing properly.

• Notable companies: Thermo Fisher Scientific and Oxford Instruments.

8. Packaging Process

The final step is packaging the semiconductor. However, this does not merely involve placing it in a box; it includes various forms of processing and reassembly to suit specific applications.

• A highly familiar company appears here: TSMC.

TSMC alone accounts for approximately 5.7% of Taiwan’s GDP, highlighting its immense production capacity.

There are companies that specialize in designing semiconductor chips, such as NVIDIA, which focuses on GPU development.

Other companies, like AMD, package both CPU and GPU together, while Qualcomm manufactures chips for mobile devices.

To understand the relationship between AI and semiconductors today, we need to take a macroscopic look at how semiconductors have evolved.

From the 1990s until recently, the competitive advantage in the semiconductor industry was about making chips smaller.The image above illustrates how semiconductor chips have been fabricated on wafers, progressively reducing in size over time.

Why Did Chips Need to Become Smaller?

If the circuitry on a chip is large, the distance for data transfer between components increases.A longer distance means that electrons must travel further, leading to higher resistance and ultimately lower energy efficiency.

By reducing the size of semiconductor chips:

• The distance for electrical signals to travel becomes shorter, improving energy efficiency and performance.

• Less power is wasted, leading to cost savings.

• Smaller chips can be integrated into mobile devices more efficiently.

Thus, from the 1990s to recent years, there was a strong focus on shrinking chip sizes.

Since chip size is primarily determined during front-end processing, this stage was historically considered the most critical part of semiconductor manufacturing.

As chip sizes decreased, they became suitable for mobile devices, leading to the iPhone moment in 2007.

On January 9, 2007, Steve Jobs introduced the first iPhone—a moment that changed the industry forever.

Since semiconductors had become small enough to fit into mobile devices, a huge market was created.

From this point onward, the demand for smaller and more efficient chips surged.

However, as semiconductor chips continued shrinking, a problem arose—quantum tunneling effects, where electrons leak through thin barriers.

As a result, further miniaturization became increasingly difficult, shifting the industry’s focus toward quantum computing.

On December 30, 2022, ChatGPT was launched, marking the beginning of the AI era.

This event is now referred to as the ChatGPT Moment.

With this shift, back-end processing (which was previously less emphasized) became increasingly critical.

Let’s explore why.

To train AI models, a massive number of GPUs are required.However, using small chips for AI training proved inefficient.

Instead of making smaller chips, the focus shifted to combining multiple chips to form larger processing units for AI computation.

Thus, data centers became a crucial infrastructure.

The concept of chiplet technology emerged as a way to package multiple chips together.

Now, semiconductor companies not only design chips but also engineer and manufacture chiplets—combinations of multiple chips optimized for AI workloads.While design companies focus on chiplet development, manufacturing is typically outsourced to TSMC.

So, how do data centers function in this ecosystem?

1. Chiplets are combined to form a rack (a modular unit in data centers).

2. Racks are stacked together, creating large-scale data centers.

3. These data centers power AI applications, cloud computing, and enterprise solutions.

Let’s examine the key technologies required for efficient data center operations.

1. Chip-to-Chip Communication

   • Since multiple chips exist within a data center, errors in data transmission can lead to serious computational failures.

   • Chips must be precisely designed to communicate seamlessly.

2. Chiplet-to-Chiplet Communication

   • Even within chiplets, smooth communication between components is essential.

3. Rack-to-Rack Communication

   • Data centers consist of multiple racks stacked together.

   • Effective communication between racks ensures efficient data processing and storage.

This interconnectivity is commonly referred to as networking, and the leading company in this field is BROADCOM.

Because of its dominance in networking technology, BROADCOM is often referred to as “the next NVIDIA” or “a rising blue-chip in the semiconductor industry.”

A core networking technology is SerDes (Serializer/Deserializer), which optimizes chip-to-chip communication.BROADCOM specializes in this technology, helping design efficient networking infrastructures for major enterprises.

Most people think of data centers as large physical buildings filled with servers.

However, data centers are not just hardware—they also provide cloud-based software tools for businesses.

Companies that develop AI-powered software using cloud computing infrastructure include:

• OpenAI (ChatGPT)

• Salesforce

• Snowflake

• Palantir

• Applovin

These companies leverage cloud infrastructure to build software that provides AI-driven services to businesses.

For example:

• Palantir specializes in data visualization, processing, and automation, often working with governments.

• Salesforce integrates AI into its CRM software to enhance customer management.

• Applovin provides software development kits (SDKs) to help mobile developers optimize ads and improve user engagement.

These cloud-based software companies represent a new wave of AI-driven businesses that do not manufacture semiconductors but rely on cloud computing and AI integration to generate revenue.

In addition to software, data centers require robust physical infrastructure, including:

• Temperature control

• Humidity regulation

• Efficient energy supply

A leading company in data center design is VERTIV, which specializes in optimizing power grids, cooling systems, and structural designs for data centers.

With the rise of AI, data centers will continue to expand, making networking companies like BROADCOM and MARVELL even more valuable.

For students considering their future careers, it is important to take a holistic view of the industry.

1. Semiconductor Manufacturing & Design

• Design-focused roles (e.g., NVIDIA, AMD, Qualcomm)

• Front-end processing (e.g., SUMCO, ASML, Lam Research)

• Back-end processing (e.g., TSMC, Broadcom)

2. AI & Cloud Computing Startups

• It is nearly impossible for a startup to build a front-end or back-end semiconductor company.

• Instead, launching a software or AI startup using cloud computing infrastructure would be more realistic.

3. Big Tech Careers

• Students aiming for Big Tech companies can target Apple, Amazon, Google, Tesla, Meta, and OpenAI.

• These companies are expanding rapidly in AI and data centers.

Conclusion

The ChatGPT Moment has accelerated AI adoption, increasing the importance of back-end semiconductor processing and data centers.

Many of the most influential CEOs today come from backgrounds in engineering, natural sciences, and computer science—highlighting the importance of technical expertise in shaping the future of technology.

Today, we explored the second part of industry trends, focusing on AI and semiconductors. Stay tuned for the next industry trend update.

Thank you!