Why It's Getting Harder to Build AI: Inside the Chip Shortage Slowing Down Tech

A group of leaders from different parts of the AI industry met at a conference in Beverly Hills to talk about a serious problem: they can't get enough computer chips to keep up with demand. The shortage is slowing down AI companies that want to build bigger and more powerful systems.

Think of it like this: right now, everyone wants to buy the same popular car, but the factory can only make so many each month. Some people have to wait a long time to get one. That's what's happening with the specialized chips that power AI systems.

ASML, a Dutch company that makes the machines used to manufacture these advanced chips, says the shortage will probably last between two and five years. ASML is especially important because it's the only company that makes the special equipment — called extreme ultraviolet lithography — needed to create the newest, most powerful chips. When ASML's CEO says there's a problem, the whole industry pays attention.

Why the Shortage Keeps Getting Worse

The problem isn't just about making more chips. Companies building AI systems keep making them larger and more demanding. Each new AI model needs more computing power than the last one, like how each generation of smartphones becomes more powerful and uses more battery. This creates a cycle where demand grows faster than manufacturers can possibly keep up.

It takes time to build more chip factories and train workers. You can't just flip a switch and double production overnight. Meanwhile, huge tech companies like Google, Microsoft, and Amazon are all buying as many chips as they can to build the AI systems they need for their services.

New Types of AI Are Adding to the Problem

One of the panelists, Qasar Younis, runs a company that builds AI systems for physical machines — robots, self-driving trucks, drones, and defense equipment. These systems need special types of computing equipment and sensors. It's like the difference between running a software program on a desktop computer and building a robot that has to actually work in the physical world. Both need chips, but the robot also needs motors, sensors, and other hardware.

This company is starting to work more with the military and defense sector. When computer chip shortages happen, companies sometimes look for more stable sources of business. Defense contracts can be more predictable than building products for the general public.

Search Engines Are Changing, and It Uses More Power

Another panelist, Dimitry Shevelenko, works at Perplexity, a newer search company. He explained that the way search engines work is changing. Traditional search engines just look up words in a giant index. Newer AI search systems actually think through your question, reason about the answer, and have conversations with you — more like talking to a knowledgeable person than typing into a search box.

This new approach uses a lot more computing power. It's the difference between looking up a word in a dictionary and having a conversation with an expert. The expert has to think, consider what you've said, and respond thoughtfully. That takes time and energy.

Companies like Perplexity are competing with Google, Microsoft, and Amazon for the same limited supply of chips. Smaller companies might get left behind just because they don't have the money or negotiating power to secure enough equipment.

A Bigger Question: Is This the Right Path?

One of the panelists, Eve Bodnia, is a physicist who founded a startup called Logical Intelligence. She raised a different kind of concern: maybe the entire approach the AI industry is taking is inefficient. Instead of making computers bigger and faster, maybe there's a smarter way to build AI systems.

The truth is, nobody knows if she's right. The AI industry has invested hundreds of billions of dollars in the current approach — using massive computers and a technique called transformer architecture. If that approach turns out to be a dead end, that would be a waste of enormous resources. However, when industries change direction, it usually takes many years. Companies have to keep working with the technology they have now while watching for better options on the horizon.

Looking for Solutions in Unusual Places

Some companies are exploring putting data centers in space. That sounds like science fiction, but when you're facing a serious shortage, even unusual ideas start to look interesting. Satellites can access unlimited solar energy and avoid some of the cooling problems that ground-based computer centers face.

We've seen something similar before. In the late 1990s, when the internet was exploding, companies invested enormous amounts of money in undersea cables to connect continents. At the time, it seemed like too much. But that infrastructure turned out to be essential. Today's push to build more AI capacity, even in unusual ways, might follow the same pattern — what looks excessive now could be necessary later.

What This Means for the Rest of Us

The shortage creates real consequences. Large tech companies with money and influence will get the chips they need. Smaller companies might struggle. Smaller players that can't secure reliable access to chips will face a disadvantage, and this could push the industry toward consolidation — fewer, bigger companies.

For people planning to use AI systems at their companies, it's worth knowing that getting access to this technology will take longer than expected, and it will probably cost more. If you're considering an AI project, building in extra time for planning and budgeting is practical advice.

The overall picture is straightforward: the AI industry grew faster than its supply chain could keep up with. There's real friction in the system right now. But this kind of bottleneck also pushes people to find smarter, more efficient solutions — both in how they design systems and in where they build them. History suggests that pressure often leads to breakthroughs.