Why are US stock markets recovering a bit after a day of bloodbath linked to China’s AI wonder, DeepSeek? To understand the question, we need to do some deep-seeking to figure out the intricate interplay between technological wonders, economic behaviour and stock market movements. But, first up, something about India’s thali meals—they can tell us something about the open-source-based artificial general intelligence (AGI) engine that has set tongues wagging as it rivals ChatGPT and Gemini at a fraction of development costs.
In India, you can buy a cheap thali at a roadside stall that serves a wholesome meal at Rs 50 a plate, or go for a lavish unlimited thali at a five-star hotel for Rs 1,500. There are differences in their features, style, and details. At the roadside stall, you may get a small thermocol plate with just a couple of dishes to go with the main course rice, while liveried waiters will fawn on you with silver cutlery in an air-conditioned environment when you order a thali at a five-star place.
DeepSeek Is Food For Thought
What the Chinese-backed hedge fund, High-Flyer, has given us through DeepSeek is like a decent, affordable thali. It is based on open-source technology that allows people to tweak its features, much like a thali that allows a second helping after a plate is served. It is based on limited specifications, but it also cleverly circumvents some advanced, ambitious features, and matches up to in-vogue engines such as ChatGPT.
DeepSeek is food for thought, even in literal terms, as it does the work of a human brain. Its R1 engine built at an estimated cost of only $5.6 million by a one-year-old startup that matches ChatGPT offers a lesson in utilitarian economics.
An easier way to understand the phenomenon is to compare DeepSeek with a Chinese-made Android phone equipped with features quite comparable to an iPhone that runs on Apple’s proprietary iOS. By developing Android as an easily usable mobile OS platform, Google gave the world’s phonemakers a cheaper alternative to iOS. DeepSeek seems to be returning the favour with a cut-rate AI engine that may help proliferate the use of AGI, much like how Asians seized on Android while the US-centric iPhone got blissfully isolated in its elite, high-profit ecosystem. Another way to look at DeepSeek is to see it as an equivalent of the open-source Linux OS platform that challenged Windows in its prime. Windows survived and grew with its own strategy, but Linux changed the way the industry behaves. It is still a great technological force behind the scenes.
DeepSeek Challenges ‘Core’ Players
Understandably, US President Donald Trump called DeepSeek’s arrival a wake-up call for American giants. I want to call DeepSeek the Android of AI for its potential power to proliferate its use and challenge established technology leaders.
A key point is that AGI, which learns things and articulates expressions like humans, needs a humongous amount of processing power, which, in turn, consumes tonnes of electricity to make the volumes of data come together in a cohesive form. AI needs a lot of “training”; it is helped by advanced GPUs (graphic processing units) for high-speed mathematical calculations that power up machine learning. Nvidia, once a leader in humbler graphics cards that enhanced the video or gaming experience on your PC, is today a leader in AI, thanks to its GPU clout, which has helped its stock grow 20 times in the past five years.
The Nvidia Wipeout
However, after Nvidia’s stock fell 17%—nearly $600 billion—in history’s biggest single-day wipeout on Monday, the jury is out on whether various US plans to invest hundreds of billions of dollars in AI-linked ventures is a smart move. Nvidia’s computing power is undeniable, but what DeepSeek is proving is that you can build a good-enough monster without necessarily having the best-in-class GPUs. Its training methods also differ. Picture it like a dog-whisperer who can tame pets at cheaper rates to do smarter things.
The market has recovered somewhat after the Monday mayhem, and one of the reasons cited for it is the end-user perspective on AI. Just like you can buy a cheaper thali or an Android phone, DeepSeek can make companies embrace Generative AI that can create content at more affordable budgets. And that is good for their bottom lines. As investment bank UBS noted, the arrival of DeepSeek implies that AI applications and the intelligence layer that goes with them will gain at the cost of the enabling layer (GPU power) in terms of cost advantage.
China’s Clever Timing
Having said that, the US still retains its leadership as the creator of core technologies that power the AI revolution. But the timing of DeepSeek’s release is such that President Trump may think twice about slapping trade restrictions on China in some spheres, while investors all set to bet their farms on American AI may ponder on the most optimal way to spread their bets.
AI, as an across-the-board phenomenon, is headed for interesting times.
(Madhavan Narayanan is a senior editor, writer and columnist with more than 30 years of experience, having worked for Reuters, The Economic Times, Business Standard and Hindustan Times after starting out in the Times of India Group.)
Disclaimer: These are the personal opinions of the author