If you’ve been following the news lately, you’ve probably heard about DeepSeek ,or at least about how China recently sent shockwaves through the global AI ecosystem.
But before we dive in, let me address the elephant in the room: why is a real estate consultancy writing about artificial intelligence?
The answer is simple. Whether we like it or not, AI is going to impact all of our lives. And as someone who’s been tracking news and developments in this space, I realized something important: most coverage focuses on the drama ("Chinese company wipes out trillions of dollars from the AI bubble!") rather than explaining what actually happened and why it matters.
So, here’s my attempt to connect the dots and break it down for you as i believe this is an important incident.
I am no expert in technology or AI or geo-politics. This is based on the information i have at the moment.
So, what is DeepSeek?
DeepSeek is an AI chatbot, a type of technology known as a Large Language Model (LLM). It competes with industry giants like OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude. But there’s a key difference: DeepSeek is open-source .This means its creators have made the "recipe" for building the AI publicly available, setting it apart from companies like OpenAI, which operates as a closed-source, for-profit entity.
You can download DeepSeek and use it privately if you’re tech savvy. Your data stays with you,it’s your AI, your rules. Anyone can take DeepSeek’s code, tweak it, and build upon it. Unlike ChatGPT, which charges 20 usd per month (Rs.1,950) for its subscription and 200 usd per month for its advanced reasoning model, DeepSeek offers a free version that rivals even the priciest models.
Now, this wasn’t supposed to happen, at least not now - not according to Sam Altman (OpenAI) or the U.S. government.
Let me explain why.
DeepSeek reportedly achieved its breakthrough with an investment of just 6 million and some change, using less powerful hardware than what’s typically required. Compare that to OpenAI’s plans to build an Artificial General Intelligence (AGI) / a 'digital god' through a project called Stargate, with an investment of 500 billion .
When DeepSeek came along, people started questioning whether such astronomical sums are really necessary to create cutting edge AI. Maybe LLMs aren’t the holy grail they were made out to be. May be it's just a commodity. Panic ensued, and stock values of key players like Nvidia (the company that makes GPUs essential for training AI models) plummeted.
Deepseek R1 is what they call as a Balck Swan Event. A black swan event is something unexpected but with profound consequences.
Now, lets see why this is a big deal. Tech Experts if you read, please be kind to me.
Large Language Models (LLMs) are very advanced computer programs that has the ability to learn from huge amounts of text, similar to how a person learns by reading a lot of books. These programs recognize patterns in words and sentences to predict what comes next.
To train these programs, computers need a lot of power, just like how a car engine needs fuel to run. LLM trainings are done via something called GPU [ by Nvidia ] which need high end semi-conductor chips. The best of these chips are made by companies like TSMC, Samsung, and Intel, with TSMC being the industry leader. These chips are built using special machines ( Extreme Ultra Violet Lithography Machines ) from ASML , a Dutch company, which helps make them smaller and more powerful.
Here’s the chain:
LLMs -> need GPUs -> need advanced chips -> need TSMC -> need ASML’s EUV machines -> need global suppliers -> controlled by geopolitics.
Each link in this chain matters:
LLMs : Can’t function without powerful GPUs.
GPUs : Can’t exist without cutting-edge chips.
TSMC : Factory making these chips at scale.
ASML : Is the only company that makes Euv machines TSMC needs.
Global Suppliers : Without them, ASML can’t build its machines.
Geopolitics : Decides who gets access to this technology.
In other words, think of creating advanced AI like building a Car:
GPUs = The engine
TSMC/ASML = The factory that builds the engine
Geopolitics = The guard controlling who gets to use the factory.
The ultimate bottleneck in this chain is ASML’s EUV machines. These machines are so advanced and complex that no one else in the world makes them.
Here’s where geopolitics comes in. The U.S. government lobbied hard to ensure that China couldn’t get access to the best chips. They succeeded to an extent. But there was a loophole: lower-tier versions of these chips were still available for them.
Think of it like this: Imagine you’re preparing for an exam, and your competitor wants to stay ahead by making sure you don’t get textbooks. That’s essentially what the U.S. tried to do to China.
And this i realized is the actual competitive advantage of companies like OpenAI, not innovation.
But here’s the twist, the Chinese got creative. Deepseek claims, despite being cut off from top tier hardware, they figured out how to build highly capable LLMs at a fraction of the cost. And they didn’t stop there - they shared their technical papers with the world. If you or I had the expertise, we could use this knowledge to build our own AI.
Hence, the surprise and market panic.
When asked if a small team with 10 million dollars from India could compete in creating a foundational AI model, Sam Altman famously replied, “It’s hopeless.” Deepseek paved the way for more players to enter the AI space. In simple terms, DeepSeek democratized Language Models.By releasing their work as open-source, they’ve leveled the playing field. Now, anyone with the right skills can contribute to or benefit from AI development.
U.S. investors should actually be thanking DeepSeek. Because, there will be pressure to do more with less resources going forward. Because, people now think it is possible. Even if Deepseek is a scam, this profound change in thinking will have a positive impact.
The U.S will definitely come up every possible way to de-credit them, saying national security, data safety, they cheated, copied, smuggled Chips, don't answer queries about Chinese politics [ chatgpt has its own share of biases ] etc. May be the Chinese are not transparent in their claims of using less funds or using less powerful hardware or even the number of GPUs they used. I do not know. But the fact that it is Open Source is still a huge deal, if am not mistaken. One thing i realize is, if the US government is shielding AI companies then it is only logical to assume that the US government's intelligence assets must be already in place within these organizations and there will be checks and balances to companies like this by organizations like NSA. So - no point in worrying about data safety at least.
As a small business owner, I have always have to get more done more with less. Whether it’s stretching a tight budget, making do with a lean team, or finding creative solutions to problems, resourcefulness is the lifeblood of any small enterprise.
Earlier I thought that OpenAI would introduce a ladder of pricing and intelligence capabilities in the coming years. Eventually, I assumed that access to the most powerful AGI (Artificial General Intelligence) would be restricted to the core OpenAI team and the U.S. government. And diluted versions will be available for those who can pay based on the amount they charge.
This was my expectation until I learned about Deepseek.
May be the company that is going to have a profound impact via AI is yet to be started or the best use cases of AI tech is still under development. If you look at history, when something new and profound comes - there will be a bubble first - then it pops - people do reality check and eventually from the remains of the previous destruction a solid industry forms. Easy to understand references, Phones : Nokia - Apple, Search : Yahoo - Google, Social Media : Myspace - Facebook etc.
AI is going to be a layer on almost everything that we are going to have. And its definitely going to have an impact in Real Estate too in various forms. May be few years down the line - you may not need a real estate consultant. You tell your AI what you need, it will identify, negotiate and close deals with other AIs on your behalf. It may be as simple as ordering food on a delivery platform. Where we fit in that world, i am yet to figure out.
Also, in this new day and age you don't need a big team or a large office or millions to build a substantially large business interms of impact. One person multi million companies should be norm in few years. And it should be the way.
While the U.S. may feel “DeepSeeked,” it leveled playing field to an extend.
Whatever it is - it is just the beginning of a different world - age of Synthetic intelligence. A world where we are going to co-exist with alien intelligence. One thing is almost certain, our species won't be the top of the intelligence food chain in the long run.
LLMs in its present form may not mean much on a grand scale of things, but what it will evolve into is what needs to be seen. Also in the near future - in a world where US, China or Israel has its own AGI, where will India stand. There is a race going on for sure - but where do we stand?