Anna's Deep Dives

_{Just facts, you think for yourself}

A. What AI Chips Are: A Quick Look at GPUs, ASICs, and More

AI gets its power from special chips, not like regular silicon. Four main types of these chips are important for AI now: GPUs, ASICs, FPGAs, and SoCs.

Graphics Processing Units (GPUs) are great for training AI because they can do many calculations at once. For example, Nvidia’s H100 GPU costs between $27,000 and $40,000. Big tech companies like Google and Microsoft use a lot of these.

Application-Specific Integrated Circuits (ASICs) are made for specific AI jobs, making them fast and power-efficient. Google’s Tensor Processing Unit is an ASIC. Broadcom's sales of custom AI chips grew 220% in 2024, reaching $12.2 billion, which shows they're in demand.

Field-Programmable Gate Arrays (FPGAs) are flexible. You can change what they do even after they're made. The FPGA market was $12.1 billion in 2024 and is expected to be $25.8 billion by 2029, growing 16.4% a year.

System on Chips (SoCs) put many parts on one chip, which saves power. Ambarella said it shipped over 30 million of its AI SoCs in early 2025. Intel’s Xeon 6700P processors are also made for AI systems in data centers.

Neuromorphic chips are newer and designed like the human brain for efficient AI work. This market was worth $8.36 billion in early 2025 and could grow to $47.31 billion by 2034. The U.S. Air Force uses these in experimental mini-drones.

The whole AI hardware market is growing very quickly. It was worth $66.99 billion in 2024 and is predicted to be $334.50 billion by 2034. That's a 17.44% growth each year. Big names like NVIDIA, Alphabet, Intel, and AMD lead this market, but new startups like Tenstorrent and Mythic are also coming up with new ideas.

B. How Chips Are Made: The Complicated Global Supply Chain

Making semiconductors is very complex and involves the whole world. Politics and national plans always affect this important network. Tools like AIPNET track over 5,000 goods in this chain. Problems, like China controlling rare earth mineral exports, often test how stable it is.

Making chips has many steps: design, making them (fabrication), assembly, testing, and then sending them out worldwide. Every step is important.

Special companies make key equipment. For example, ASML in the Netherlands makes EUV lithography systems needed for chips smaller than 5 nanometers.

Taiwan Semiconductor Manufacturing Company (TSMC) is the biggest contract chipmaker. It makes about 60% of all semiconductors globally. And it makes over 90% of the most advanced chips needed for AI. TSMC plans to invest up to $42 billion to stay ahead.

Countries are focusing on different areas. India has about 20% of the world's chip designers. Its government plans to help fund 70% of the country's first big new semiconductor factory in 30 years.

Southeast Asia's semiconductor market aimed for $576 billion by the end of 2024. Vietnam has 6,000 semiconductor engineers but needs 50,000 for its goals. India's Semiconductor Mission (ISM) wants to grow its market from $26.3 billion to $271.9 billion by 2032.

The European Union wants to make 20% of the world's chips by 2030. That's a big goal since the EU currently imports about 80% of its chips. Japan, once a leader, is also investing over $25 billion to boost its chip industry.

The United States uses export controls to limit China's access to advanced chip-making tech for national security reasons. The U.S. wants to make 28% of advanced chips globally by 2032. The CHIPS and Science Act puts $280 billion toward this, with $52 billion for research and production.

China put new semiconductor rules in place in April 2025. These rules sort chips by where they're made, which affects taxes and trade. The global supply chain has many weak spots. Shipping through the Panama and Suez Canals recently fell by over 50% due to conflicts and climate change. Taiwan's political situation is also a big concern because it's so important for chipmaking.

Demand for semiconductors keeps growing. It's expected to be a $1 trillion yearly market by 2030. This growth shows we really need more varied and stronger supply chains.

C. U.S. vs. China: Why They Don't Trust Each Other on Tech

The tech relationship between the U.S. and China has gotten worse. It used to be mostly about business competition, but now it's more like a direct fight. Semiconductors and AI are the main areas of this fight.

China's tech progress makes the U.S. nervous. The DeepSeek-R1 AI model, from a Chinese company, launched on January 20, 2025. It reportedly runs 95% cheaper than similar U.S. tech. This surprised many in the AI field. China's "Made in China 2025" plan aimed to make 70% of its own semiconductors by 2023. In 2024, it still depended on others for 41%, but its goal was clear. China has invested about $150 billion in its semiconductor industry in the last ten years. Because of government help, some Chinese companies can sell chips 20-30% cheaper than U.S. ones.

U.S. worries about Chinese tech aren't new. For example, the Trump administration suggested banning TikTok in June 2020 due to security concerns. In January 2025, the U.S. Supreme Court said TikTok could stay in the U.S. if its U.S. business was sold off by its Chinese parent company. TikTok has over 170 million U.S. users, so it's a big deal.

Because of these trends, the U.S. has made its export controls tighter. These rules have stopped advanced AI chips, like Nvidia’s H20 series, from going to China. A 2025 U.S. Senate bill called the "Decoupling America’s Artificial Intelligence Capabilities from China Act" suggests harsh penalties, like up to 20 years in prison for U.S. citizens caught using banned Chinese AI products. Besides these rules, the U.S. CHIPS Act set aside $52 billion to support America's own semiconductor industry.

Cybersecurity problems make the distrust between the two countries even deeper. In February 2025, China's government said it was investigating claims of NSA cyberattacks on Chinese groups. This came after Western cybersecurity firms reported that hackers, possibly linked to China, had broken into U.S. Treasury Department systems. These kinds of events, and the blame that goes back and forth, keep the tech fight going.

D. The AI Race: Why Being Top in Chips is a Must for World Leaders

By early 2025, the global push to lead in AI and chip tech had become like an arms race. Countries now see being the best in chips not just as good for business, but as something they absolutely need for national power and global leadership.

Global spending on AI tech was expected to be over $500 billion in 2025. About 20% of this huge amount is for defense and military use. China has set aside $150 billion for its national AI development. The U.S. invested about $4.6 billion in AI from 2021 to 2023 through different programs.

AI is being used more and more by the military. Automated Weapons Systems (AWS) that can attack targets without direct human control bring up serious ethical questions. It's thought that governments worldwide will buy over 80,000 advanced spy drones, many using AI, in the next ten years.

The U.S. wants to make three times as many semiconductors by 2032. The $50 billion in the CHIPS Act is a big part of this plan. Where advanced chips are made is very important; Taiwan's TSMC currently makes over 90% of the world's most advanced microchips. Relying so much on one place can be risky.

China aims to be the world leader in AI by 2030. It's investing over 10 trillion yuan (about $1.4 trillion) in advanced tech to get there. The U.S. defense budget for AI jumped from $10.3 million in 2022 to $320.4 million by 2023. This extra money now funds 685 different AI projects in the U.S. military.

Other countries are also spending a lot. Saudi Arabia has promised $14.9 billion to build a huge AI data center and plans 42 more AI centers in the country.

Companies are spending a lot on research too. Amazon spent $24.3 billion on capital, much of it for AI. Nvidia planned to spend $12.9 billion on R&D in 2025 alone. The AI chip market could hit $400 billion in yearly sales in the next five years, showing how much money is involved.

U.S. export controls are an important part of this race. These rules try to stop rivals from getting key advanced tech, and they affect about 120 countries. This global tech race is changing international security and how countries work together economically.

We don’t take shortcuts, chase headlines, or push narratives. We just bring you the news, straight and fair. If you value that, click here to become a paid subscriber—your support makes all the difference.

Table of Contents

(Click on any section to start reading it)

I. AI Chips: The New Cold War

• Why AI Chips Are Key to National Power

• The "Silicon Curtain" and the New AI Cold War

• DeepSeek: A Chinese AI Company Shakes Things Up

II. AI Chips: What They Are, How They're Made, and Why There's Tension

• What AI Chips Are: A Quick Look at GPUs, ASICs, and More

• How Chips Are Made: The Complicated Global Supply Chain

• U.S. vs. China: Why They Don't Trust Each Other on Tech

• The AI Race: Why Being Top in Chips is a Must for World Leaders

III. U.S. Moves: Export Bans, Policy Changes, and Trying to Shape the Future

• U.S. Export Controls on China's AI: How They've Changed

• The "AI Diffusion Rule": What It Was, Why It Was Criticized, and Its Repeal

• Biden's AI Rule: What It Tried to Do

• Trump's New Plan: Simpler Rules or a Different Strategy?

• The "Chip Security Act" Idea: Tracking Chips for Better Control

• How Location Tracking Could Stop Illegal Chip Shipments

• Chip Tracking Problems: Privacy, National Rights, and Practical Issues

• What's the U.S. Strategy? Stopping, Separating, or Just Slowing Down China?

• Side Effects and Unexpected Results: How U.S. Policies Are Really Affecting Things

IV. Nvidia's Challenges, Industry Effects, and the Cost of Following Rules

• Nvidia: Stuck Between U.S. Rules and the Chinese Market

• Nvidia's Modified Chips for China (H20, etc.): New Ideas or Just Weaker?

• Weaker Chips for China: How This Affects Their AI Tech

• How Chinese Customers Are Adjusting

• The Cost: Billions Lost by Nvidia and Others Due to Export Rules

• How Export Rules Affect Chip Companies Worldwide

• Shaky Markets: Investor Worries, Stock Swings, and Rethinking AI Value

V. China's Answer: DeepSeek, Local Chips, and Aiming for AI Independence

• DeepSeek Shakes Things Up

• Smarter, Not Harder? DeepSeek's Efficient AI

• DeepSeek vs. Big Spenders: A New Way for AI?

• DeepSeek's Open Approach: Good for All or Smart Business?

• More Than DeepSeek: China's Own AI Chip Makers Are Growing

• Huawei, Alibaba, and Other New Chinese Chip Makers

• China's AI Chips: Catching Up or Doing Things Differently?

• China's Big Plan: Self-Sufficiency with Top Companies and Big Money

• Big Hurdles: Problems for China's Chip Independence Plan

• China's AI: Copying or Creating New Things?

VI. A Changed World: New Politics, Innovation Challenges, and What the AI Chip War Leaves Behind

• Broken Supply Chains: New Tech Groups and a More Split Digital World

• What's Next for Innovation: A Boost or a Bust from Competition?

• Tech "Islands": A Risk to Global Progress?

• Separate Work, Surprise Discoveries?

• Power Shifts: How AI Chips Are Changing Global Politics

• Ethics in the AI Race: Will Security Beat Safe Development?

• After 2025: What Could Happen with AI, U.S.-China, and World Peace?

• Lasting Effects: The AI Chip Cold War's Unavoidable Impact and Facing an Unsure Future

Baked with love,

Anna Eisenberg ❤️