I recently read an interesting article by the well-known economist Daron Acemoglu in the Financial Times. He wrote: “Even in uncertain times, it is hard to picture a completely different future.” He then imagined a historian, 25 years from now, looking back at 2025 as the start of America’s decline. What could cause this decline? And how might the current government prevent it?
This is a useful way to think, especially in an industry where we often focus on daily tasks and forget the bigger picture. Looking at history reminds us that what seems extremely important today will eventually become just another part of the long-term story.
With this in mind, I sat down to discuss the long-term impact of DeepSeek with Sumant Wahi, a tech-focused portfolio manager at Man Group and one of the sharpest minds I know in the industry. DeepSeek, in case you missed it, is a language model developed by a Chinese hedge fund incubator. It can perform tasks as well as ChatGPT but at a much lower cost and using far less computing power.
DeepSeek has reportedly trained a model as powerful as GPT-4 for just $6 million. That’s very cheap compared to the “over $100 million” that OpenAI’s CEO, Sam Altman, says GPT-4 cost. This could have a huge impact, but one of the most important—and often ignored—effects might be in robotics. Sumant believes this is where DeepSeek’s breakthrough could make a big difference.
From Theory to Reality
For years, humanoid robots were like flying cars—possible in theory but too expensive, power-hungry, and impractical. People thought that for robots to work outside factories, they would need supercomputers inside them. This would cause overheating, drain batteries quickly, and make them far too expensive to be useful.
DeepSeek has just challenged that idea. If AI models can be trained and run on much cheaper, less power-hungry hardware, then the obstacles in robotics may not be as big as we thought. The same improvements making AI more efficient could also change “Physical AI”—a term used by Jensen Huang for robots that can understand and interact with the real world. This could be the turning point where humanoid robots go from science fiction to a real economic force.
Robot prices have already dropped a lot, while their usefulness has grown rapidly. In the early 2000s, Boston Dynamics’ BigDog, a four-legged military robot, cost several million dollars per unit and had a loud engine like a chainsaw. Their humanoid robot, Atlas, also cost over $1 million per unit. Now, look at Tesla’s Optimus robot—Morgan Stanley estimates its material cost at $50,000 per unit (not including software). That’s already cheaper than traditional industrial robotic arms, and Tesla expects mass production to bring costs below $20,000. In China, Fourier Intelligence’s GR-1 aims to launch at under $30,000.
Industrial vs. Consumer Robots
This is where things get interesting. A humanoid robot that can move around a home, do simple tasks, and connect with smart home systems isn’t just a cool idea—it’s something people might actually want to buy. The debate about whether robots will first become common in factories or in homes might not matter anymore. If AI keeps getting better, the answer could be both.
Industrial robots have always made sense because companies can afford the high costs in exchange for better efficiency. But if the price of a humanoid robot drops low enough, everyday people might start buying them too. Labor shortages in areas like healthcare, food service, and home care make this even clearer. The number of working-age people in the U.S. is no longer growing, as shown in the chart below. If Trump follows through on his plan to deport millions of undocumented immigrants, the labor shortage could become much worse. At that point, selling humanoid robots for $15,000 wouldn’t just be possible—it might be necessary. This could have a big effect on the economy, from productivity to inflation.
Robots Are Getting Smarter and Cheaper
Robots are not just becoming more affordable—they are also getting smarter. In the past, “cobots” (collaborative robots) were pre-programmed to do repetitive tasks. But now, AI-powered robots can handle more complex, hands-on jobs in real-world settings. As AI improves, robots no longer need every movement to be programmed in advance. Instead, they can learn and adapt on their own.
Why DeepSeek Was Bad for Nvidia but Good for Robotics
Another big change is in computer chips. Many people thought advanced robots would always need powerful Nvidia chips to run AI programs. But DeepSeek’s new technology shows AI can work with less memory and simpler hardware. That means robots don’t need expensive, high-power chips, making them smaller, cheaper, and more efficient.
This is why Nvidia’s stock dropped when DeepSeek launched. Investors realized that AI is not just about having the most powerful computers—it’s about using resources wisely. If China can create advanced AI with simpler chips, robotics companies can do the same. Just like China leads in making electric car batteries, it could soon dominate AI-powered robots while Western companies focus on expensive data centers.
Using Robots for Dangerous Jobs
The biggest impact won’t be in home robots but in industries that need automation the most—especially dangerous jobs. A recent Morgan Stanley presentation highlighted this point. These jobs may not seem as exciting to Silicon Valley investors, but they are where robots can truly improve human lives.
People have been talking about self-driving trucks for years, but the real change might happen in delivery and logistics. Fully self-driving cars still have many problems, but robots in warehouses, drones, and small delivery bots are already being used. Another area growing fast is farming robots. In the past, it was hard to use robots on farms because outdoor environments are unpredictable. But now, AI is improving, helping robots adjust in real time, making farm labor with robots more affordable.
The big concern isn’t just that robots will take jobs—it’s that they will do it very quickly. In history, automation has usually happened slowly, giving people time to adjust. But if robot costs drop suddenly, the change could happen too fast, creating big economic problems. This reminds me of the Luddites, who once fought against machines taking their jobs. At one point, there was even a real risk of people pushing back against industrial progress.
Chip Geopolitics
DeepSeek shows something important: The U.S. can’t rely on chip restrictions to stay ahead in AI. The idea was that limiting China’s access to Nvidia’s best chips would slow down its AI progress. But DeepSeek just proved that idea wrong.
If AI can be trained more efficiently, China doesn’t need the best chips—just enough of them. If the same logic applies to robots, China could move ahead of the West in AI-powered manufacturing, logistics, and consumer robotics. Imagine a future where China mass-produces affordable humanoid robots before the U.S. even knows how to regulate them. That could completely change the global economy and power balance.
The main lesson here? Robotics is no longer developing slowly. AI is getting cheaper and better much faster than expected, and robots are benefiting from that. Many people thought humanoid robots would always be too expensive, inefficient, and power-hungry. That belief is now fading.
AI and robotics are merging quickly. The debate about whether humanoid robots will appear first in factories or homes no longer matters. If costs keep dropping, the answer is simple: they will be everywhere. The future of AI isn’t just about software—it’s about physical machines. And that future is arriving much faster than anyone thought.
Published: 13th February 2025
For more such articles please follow us on Twitter, Linkedin & Instagram
Also Read:
iOS 18.3.1: Urgent Security Update for All iPhone Users!
The 10 best PlayStation 5 games to play in 2025
Want Super Soft Skin? Try The Double Moisturizing Trend
