Beginner in Game Development, Which Engine Should I Use? Comparison of Unity vs Unreal vs Godot (Focusing on Free Engines) Choosing a Game Engine: Know This First: 5 Key Q&A Q1. What is a Game Engine? Is it necessary to use one? A. Think of a game engine as a set of development tools that bundles essential functions required for making games, similar to a car engine. It includes features like rendering graphics on-screen ( Rendering ), physics effects for characters colliding with walls or jumping ( Physics Engine ), and playing background music or sound effects ( Audio System ). While it's possible to create a game from scratch without an engine, it requires an immense amount of time and effort, which is why most developers today use game engines, especially beginners! Q2. I want to start game development; are there any good free engines available? A. Yes,...
Comprehensive Guide to AI Stocks: Key Strategies You Need to Know Now
AI Stock Overview: From Nvidia to Power Infrastructure, Key Insights You Need to Know
AI Stocks, What You Should Know: 5 Key Q&As
Q1. Where’s the real money flow in the AI investment craze?
It comes from massive capital investments by major IT companies like Microsoft, Google, and Amazon, known as 'hyperscalers'. For instance, Microsoft plans to spend over $30 billion (about 40 trillion won) on AI infrastructure in a single quarter. This money eventually flows into semiconductor, equipment, and power companies.
Q2. Who are the key players in the AI semiconductor market?
Currently, the strongest lineup includes Nvidia (AI chip design) - TSMC (manufacturing) - ASML (essential equipment) - SK Hynix (HBM memory). TSMC manufactures Nvidia's AI accelerators (GPUs) using cutting-edge processes, which require ASML's EUV equipment from the Netherlands. Additionally, high-bandwidth memory (HBM) from SK Hynix or Samsung is essential to accompany the GPU.
Q3. Who is winning in the HBM memory market?
In terms of technology and market share, SK Hynix is considered to be leading. They were the first to announce the completion of the next-generation HBM4 development, solidifying their position as market leaders. Meanwhile, Micron in the US is also rapidly catching up, reporting a surge in HBM sales.
Q4. What is the biggest bottleneck in the AI industry right now?
The issue is 'power'. AI data centers consume an enormous amount of electricity, but there’s a lack of a stable power grid to supply it. In the US, plans for large-scale transmission network investments have been announced in anticipation of surging power demand. In the future, power infrastructure will be as crucial as semiconductors.
Q5. Will the semiconductor equipment or foundry market continue to thrive?
In the short term, it looks very promising. TSMC has announced plans to double its advanced packaging (CoWoS) production capacity, and ASML is confident that EUV equipment sales will grow by over 30% by 2025. However, ASML mentioned that market visibility beyond 2026 is still low, so we need to keep an eye on how long this hot cycle will last.
A massive wave known as the 'AI revolution' is sweeping through industries worldwide. Where does this wave's power come from? It's the astronomical flow of 'money'. The hundreds of billions of won invested by major companies like Microsoft and Google start from Nvidia's small chips, flowing through factories in Taiwan, and reaching huge data centers and power grids in the US. Today, we will follow this immense flow of money to uncover who the real beneficiaries of the AI era are and which variables we should pay attention to in the future.
Key Points by AI Value Chain
AI investment is not about a single company, but rather an ecosystem where multiple companies work together like gears. Let's take a look at the key players and the current situation in each segment.
1. Accelerator (Compute): The Brain of AI
The semiconductors, like GPUs, responsible for AI learning and reasoning, are the heart of the AI industry. Nvidia is the absolute leader in this market. Their recently announced Q2 results for the 2025 fiscal year show that data center revenue alone reached a whopping $41.1 billion, far exceeding market expectations. Although competitors like AMD are trying to catch up, it is still Nvidia's era.
2. Memory (HBM): The Short-term Memory of AI
No matter how fast the GPU is, it’s useless without memory that can quickly deliver data. This is the role of HBM (high bandwidth memory). Currently, SK Hynix is leading technically, having announced the completion of next-generation HBM4 development, while Micron in the US is also rapidly increasing HBM sales, igniting competition.
Episode: Analysis by Semiconductor Analyst Choi Min-jun
Semiconductor analyst Choi Min-jun explains the importance of HBM in AI investments: "If Nvidia's GPU is the engine of an F1 racing car, HBM is like the ultra-fast fuel pipeline supplying that engine. If the pipeline is narrow, no matter how good the engine is, it can't achieve its speed. Right now, all hyperscalers are in a battle to secure this pipeline."
3. Foundry & Equipment: The Factories and Facilities Making AI Chips
The foundries, or semiconductor contract manufacturing plants, are where Nvidia's designed chips are actually produced. The strongest player in this field is Taiwan's TSMC. With a surge in AI chip orders, TSMC announced plans to double its production capacity for the key technology, CoWoS (advanced packaging), by 2025. Additionally, ASML's EUV equipment from the Netherlands is essential to operate these factories.
Episode: Data Center Engineer Kim Seo-yeon's Concerns
Data center engineer Kim Seo-yeon is losing sleep over the 'power' issue. "Every time a new AI server comes in, we need to bring in as much electricity as a small building. While it's great that chip performance is improving, the power grid and substations are woefully inadequate. Right now, securing land and power for building data centers is becoming a bigger task than ordering servers."
Like the engineer's concerns, the biggest variable hindering the AI revolution has emerged: 'power'.
In-depth Exploration: The Major Bottleneck in AI Investment, Power Issues
The AI revolution ultimately thrives on electricity. The International Energy Agency (IEA) predicts that global data center electricity consumption will double by 2030, equivalent to the power consumption of an entire country. This 'power bottleneck' phenomenon is emerging as the most critical variable that will influence the future of the AI industry.
To address this issue, investment flows are now spreading beyond traditional IT companies to energy and infrastructure firms. Major US utility company PG&E has announced plans to invest a staggering $73 billion in transmission networks by 2030, while companies like Microsoft and Google are signing long-term power supply contracts with utility companies before building data centers.
Furthermore, new alternatives are gaining attention for stable power supply, such as small modular reactors (SMRs) or constructing self-generation plants. As we look at AI investments in the future, it has become crucial to also consider who will supply the electricity to run these chips and how.
12-24 Month Watch Points (FAQ)
Q. What indicators should I look at to understand the flow of AI investments?
The most important indicator is undoubtedly the quarterly capital expenditure (CapEx) plans of hyperscalers. If their investment levels are maintained or increased, there’s a high likelihood that the overall industry growth will continue. Additionally, the progress of TSMC's CoWoS expansion and the timeline for SK Hynix's HBM4 mass production are also crucial indicators to read the flow of the supply chain.
Q. What is the biggest risk scenario?
The biggest risk is that power grid bottlenecks take longer than expected, delaying data center construction. Additionally, if high interest rates persist, the investment burden on companies could increase, or further tightening of US semiconductor regulations against China could also contribute to market uncertainty.
Q. Will Nvidia continue its dominance?
For the time being, that seems likely, but competition will become increasingly fierce. AMD is catching up with the next-generation accelerator MI350, and many companies are developing their own AI chips. While Nvidia's software ecosystem (CUDA) is incredibly strong, making it difficult for their dominance to be easily broken, changes in the competitive landscape are always points to watch.
Author Information: The content of this article has been compiled objectively from official IR materials from companies like Nvidia, Microsoft, and TSMC, primary articles from major foreign media outlets like Reuters and The Wall Street Journal, and data from research institutions such as the International Energy Agency (IEA) to summarize the current and future state of the AI value chain.
Comments
Post a Comment