How DeepSeek is upending AI innovation and investment

The tech world experienced a costly and highly consequential wake-up call this week with the revelation that Chinese newcomer DeepSeek had developed an advanced AI model requiring just millions — rather than billions — of dollars in development costs.

Despite concerns about DeepSeek security and that it possibly copied rival ChatGPT, the news sent US AI leaders reeling, causing them to lose more than $1 trillion in total market value — including nearly $600 billion from chip king Nvidia alone.

The DeepSeek news also raised the possibility that the existing model of AI investment and development might soon be ready for a rethink.

Indeed, rather than a bastion of small-scale disrupters and entrepreneurs, the vast majority of AI funding comes from tech giants such as Microsoft, Alphabet and Meta.

It’s the “tech-industrial complex” former President Joe Biden warned in his farewell address last week, lorded over by an emerging oligarchy that he believes has grown so powerful it threatens basic institutions and perhaps even democracy itself. 

Today’s Internet is dominated by giant technology companies like Google, Facebook, Apple, Microsoft, and Amazon.

Taken together, these digital conglomerates are worth $12 trillion, lording over the US stock market, and earning the moniker “The Magnificent Seven.”

And even as DeepSeek threatens to upend their reign, “The Magnificent Seven” has staked much of their fortunes’ future on AI. Should we be worried?

“If Big Tech continues to dominate the AI era, we risk [cultivating a culture] where users are products and their data is the most valuable commodity,” says Tom Serres, co-founder of Nautilus Asset Management, referring to Big Tech’s insatiable appetite for data, which it uses to target ads at users.

AI may appear to be at the forefront of innovation and opportunity, but it is mostly funded by nation-scale investments from legacy tech giants.

Microsoft recently announced that it will spend $80 billion on AI data centers this year, more than the United Kingdom’s entire defense budget. It has also invested billions in OpenAI, maker of ChatGPT. 

Over at Google, Demis Hassabis, CEO of Google’s AI-focused DeepMind division said the company will spend more than $100 billion to develop AI technology. Amazon is developing its own AI chips and has already invested $8 billion into ChatGPT competitor Anthropic while funneling billions into its own data center buildout.

Facebook owner Meta recently projected that it would invest $35-40 billion in AI and its metaverse arm Reality Labs, including billions on chips made by Nvidia. 

Apple’s spending spree saw it acquire DarwinAI, WaveOne, and dozens of other companies in the past few years. Elon Musk said that any company that doesn’t spend a minimum of $10 billion a year on AI like Tesla won’t be able to compete.

Big Tech even got a boost from the new Trump administration, which announced a $500 billion AI “moonshot” initiative backed by several of these companies.

Only governments can come close to matching Big Tech’s largesse. To wit, sovereign wealth funds from Saudi Arabia, and the Gulf Countries have joined the party, plowing billions into high-profile AI deals.

If these companies can buy up all the key hardware, vacuum up the best people, and leverage their market position in search, social media, e-commerce, and robotics to cross-sell their AI products, can anyone else compete? 

A splashy Beijing-backed upstart like DeepSeek, perhaps, but there are only so many splashy upstarts — and so many governments with the resources and political motivations like the one in China. 

Perhaps Western governments should enact new regulations to tame the tech behemoths. But government intervention alone will not solve the problem of tech dominance, and may even make it worse. By codifying new laws under the guise of “AI safety,” for instance, Washington may increase legal and regulatory burdens and make it harder for smaller companies to compete.

AI needs those smaller companies to evolve into larger companies to fuel ongoing innovation cycles. Don’t forget: “The big players we think of as having massive advantages in the age of AI were themselves upstarts not so long ago that took on the incumbents of their day,” says Douglas Heintzman, chief catalyst of the BRI, a think-tank. By investing so heavily in AI today, companies like Meta and Microsoft are actually lowering the barrier for others to compete and spurring future innovation. 

Big Tech so heavily dominates AI investment because the raw computing power needed to build AI models like ChatGPT was traditionally thought of as scarce and expensive, pricing out smaller players, says Serres, of Nautilus Asset Management. But that seeming lead is illusory.

DeepSeek, after all, was reportedly trained and built with only a $6 million investment, a far cry from the billions of dollars many assumed necessary to achieve an AI model that can match or even exceed ChatGPT from OpenAI. The DeepSeek model is open source, meaning anyone can audit the code and build on top of it. 

Tech entrepreneurs and venture capitalists are applauding DeepSeek while cautioning that it reveals China is far more advanced than we thought in AI.  Legendary venture capitalist Marc Andreessen described the arrival of DeepSeek as AI’s “Sputnik Moment.”

The launch is causing tremors across Big Tech; DeepSeek’s debut caused the stock of AI chipmaker Nvidia to crash more than 15% in a single trading day, and now analysts are questioning whether Big Tech was just overspending on AI — throwing money at a problem without understanding it at a deeper level.

Joseph Geraci, founder and both chief strategy and technology officer at NetraMark, adds that the AI that “currently dominates cannot be trained on consumer-level computers,” instead requiring hundreds of supercomputers known as “GPUs” that can cost $40,000 a pop. DeepSeek’s showstopper announcement proves that deep AI work can be done with far less equipment and financial outlay. 

Darwin.AI CEO Sheldon Fernandez says “AI can be used as a substitute for human creativity and logical reasoning,” posing both a risk and opportunity to many professions and opening new doors for start-up founders. Mark Zuckerberg recently admitted as much, saying AI can perform like a “mid-level” software engineer. 

How much longer until they can perform better than the best? Indeed, recent reports surfaced that OpenAI was working on an AI “agent” with PhD level intelligence. Sam Altman, CEO of OpenAI, once said AI would enable the first “billion dollar one person company” with AI handling much of the workload such as finance, marketing and logistics.

What’s more, thanks to AI, entrepreneurs may not even need to know how to code. Your AI ”programmer” can do it for you as the industry increasingly “lower[s]  the barrier to value creation,” says Heintzman. All of this will require vast sums of investment, though now with the arrival of DeepSeek, the industry’s financing model may be ready for a rethink.

Startups can also succeed by specializing, leveraging the tech built by others, and creating new applications. Fernandez of Darwin.AI says the best new startups will “train and augment core models in highly specific and technical ways to achieve their goals.”

The result will be AI startups for law, engineering, construction, and countless other industries. This raises a separate issue of platform risk, where building on someone else’s technology, whether their AI model or their cloud, makes you beholden to them. 

It is the technology behind Bitcoin, called blockchain, however, that promises to shake up Big Tech’s platform investment economics once and for all.

Indeed, it is often the combination, or convergence, of two or more technologies that have the biggest impact — and bang for the investment buck.

Consider how smartphones combined with wireless networks and GPS led to mobile apps, location-based services, and more. Today it may be AI and crypto. 

This plays out in a few ways. Most importantly, in crypto, users can pool their computers together and make them available to the public as “decentralized clouds,” that can compete with centralized systems, such as Microsoft Azure or Amazon Web Services. This will increase access and lower costs for developers who want to train and run AI models. Newer and cheaper AI models will increase the number of AI agents, who, since they can’t open a bank account, must use crypto to do transactions. 

The technology industry is in a constant state of reinvention. Clunky mainframe computers gave way to the PC, which ushered in the Internet era, and the mobile web.

Today, AI is shaking the windows and rattling the walls of incumbent technology companies.

It remains to be seen whether DeepSeek will be the company that upends the status quo with its low cost development and investment model. 

After all, DeepSeek is small, but it has China at its back — which is both a boon and a blessing.

As financial columnist Charles Gasparino noted this week, “I’m skeptical about the DeepSeek threat. I’m not saying it’s a deep-fake, but I just don’t trust anything that comes out of China.” The fight for the AI future will be competitive, but one thing we can count on is change.

Alex Tapscott is the author of “Web3: Charting the Internet’s Next Economic and Cultural Frontier” and managing director of the Digital Asset Group, a division of Ninepoint Partners LP