This article first appeared in The Edge Malaysia Weekly on February 10, 2025 - February 16, 2025

IN October 1957, at the start of the Cold War between the US and the Soviet Union, the launch of Sputnik, a low earth orbit satellite by Moscow, sent shockwaves through the world. Sputnik was seen as the ultimate humiliation of America. The narrative at the time was that the US had fallen behind its then arch-rival on a key military technology with a secretive system whose capabilities other governments could not comprehend. Sputnik sent radio signals back to earth for three weeks before its batteries became depleted. Just eight weeks later, an aerodynamic drag caused the satellite to fall back into the atmosphere. 

These days, Sputnik is remembered as the moment of national reckoning rather than a knee-jerk response to a Cold War humiliation. Americans recognise it as a wake-up call that spurred the nation into urgent action. For them, it embodies the realisation that the US significantly lagged its adversary, which helped prompt a surge in innovation and effort that led to the US landing a man on the moon in 1969, unleashing of a personal computer revolution in the 1980s, the collapse of the Soviet Union in 1992, a cellular phone revolution in the 1990s, an internet boom in the late 1990s, a smartphone and mobile internet revolution in 2007 and a battery electric vehicle (EV) revolution, with the first Tesla cars hitting the road in 2008. 

Will the launch of Chinese artificial intelligence (AI) chatbot DeepSeek, which sent Wall Street reeling, hammering tech stocks two weeks ago, be characterised by America’s ability to unify national efforts, pivot its priorities, indeed even catalyse transformation? Is DeepSeek, an open-source AI model built for a fraction of the cost of its larger US competitors, a real threat to the business models of big tech hyperscalers like software giant Microsoft, search supremo Alphabet’s Google, e-commerce behemoth Amazon.com and social media firm Meta Platforms, or just a flash in the pan?  

Two weeks after the Jan 27 sell-off on Wall Street following DeepSeek’s launch, most AI stocks have bounced back from the lows, though they are still 3% to 4% below their recent peaks. Hong Kong-based French strategist Louis-Vincent Gave of Gavekal Research says it would be a mistake for the US or its tech giants to ignore the significance of DeepSeek. China, he notes, has pulled ahead in high-speed rail, with new trains going at 450km per hour, battery EV exports, EV batteries, solar panels and drones, and is seen as the likely leader in the next-generation or 6G telecom technology with satellite-to-earth breakthroughs. “Given that China now graduates more science, technology, engineering and mathematics students each year than the rest of the world combined, fighting a tech battle against China always seemed a short-sighted strategy. With all its capital and human resources, why wouldn’t China be able to catch up with — and perhaps eventually surpass — the West’s technological advances?” he asked in a research note last week. “Few investors imagined that China could leapfrog Western countries when it comes to AI solutions, and certainly not as soon as 2025.”

For their part, American tech leaders say writing off US big tech companies such as Apple, Microsoft, Tesla, Nvidia, Amazon, Google and Meta, and betting that China is now pulling way ahead would be a mistake. “One swallow doesn’t make a summer,” is an oft-repeated line in Silicon Valley. Autonomous vehicles or robo­taxis have been plying the streets in several cities in California and Arizona for over three years and are likely to spread to another dozen or so cities with the launch of Tesla’s Cybercab later this year. 

UberEats is already delivering food in Los Angeles and other cities in California through a partner called Serve Robotics. You might want to Google search “Serve Robotics Hollywood Street YouTube” to get a sense of what I am talking about. Serve’s robots are everywhere. Want a Shake Shack burger, dim sum, curry or pizza? These robots will bring the food to your door faster than the delivery guy on the bike. (Full disclosure: I have a small amount of shares in the firm and, as such, apologise for the shameless plug. I am not qualified to give anyone investment advice.) Though they have been very volatile, Serve shares are up 102% over the past three months. Humanoid robots, now being tested by Tesla, Boston Robotics, Agility, Sanctuary AI and others, could be commercially sold as early as late next year or early 2027. No country has a monopoly on innovation.

For now, the betting in Silicon Valley is that DeepSeek is a net positive for the AI ecosystem and US AI players. More likely than not, Deep­Seek will accelerate AI adoption. Even if the narrative is that the capital-intensive model will be replaced by software-centric AI development, US companies retain an edge. Indeed, until now, Chinese companies have been known for their hardware advantage while American ones were seen to have an edge in software. Clearly, the cost of developing AI models is becoming cheaper. What that means is that new doors are being opened for companies that have not had strong enough training models but have maintained a strong distribution and user ecosystem. Apple has a huge customer base as does Meta even though Instagram’s owner has been spending on AI chips like there’s no tomorrow. “DeepSeek may create opportunities for internet companies to train and fine-tune its model on proprietary data, leading to higher accuracy and relevance in their applications such as personalised recommendations, chatbots, search and content creation,” notes Angela Hong, an analyst  for Nomura. “Video game companies could explore AI-powered content generation and real-time player interactions by using the cost-effective AI model.”

If there was any worry that DeepSeek’s “big reveal” on Jan 20 would dampen demand for Nvidia’s AI chips, earnings calls from top tech firms this past week have strengthened the narrative that it is business as usual. There was no rush to scale back in the aftermath of DeepSeek’s announcement. Alphabet announced it expects to lay out US$75 billion (RM332.7 billion) on capital expenditure this year, up 43% over 2024. Microsoft CEO Satya Nadella last week said his firm would spend US$80 billion on AI-related capex this year while Meta has vowed to spend at least US$65 billion. Of the larger players only Amazon.com, which spent US$75 billion on capex last year, has yet to disclose its AI chips budget for the current year, though CEO Andy Jassy has previously said he expects to increase spending on the chips. Add all that together and you get US$300 billion in AI capex. Add to that spending by Oracle, Elon Musk’s xAI, as well as OpenAI and Anthropic and the final tally might be over US$400 billion this year. Most, though not all, of the capex has been on Nvidia’s cutting-edge chips to enable the foundation models they host to be trained as effectively as possible. Even if you assume that Nvidia will only get 75% of that money, 2025 spending on AI chips works out to US$300 billion. In recent years, Nvidia has had over 80% of the AI infrastructure spending pie. 

“It is far from clear that the development of new and cheaper LLMs (large language models) is necessarily bad news for all tech firms,” says Neil Shearing, chief economist of Capital Economics in London. “One point generally overlooked is that this is exactly how transformational technologies are supposed to work. Most start with a breakthrough technology, which others then innovate around.” That helps bring costs down, which is crucial to encouraging the adoption of new technologies and embedding within economies. 

Competition, Shearing notes, “also helps to ensure that the value created is spread between both consumers and shareholders, rather than concentrated in the hands of a few early investors”. The new technologies create large and durable profits for firms responsible for the initial technological breakthrough. “But they also lead to disruption and the creation of new challenger firms that help innovate and embed the technology in every process,” he says. That is what ultimately raises productivity.

Until the AI chip arms race began two years ago with hyperscalers spending tens of billions of dollars every year on top-of-the-line chips like Nvidia’s latest Blackwell, US big tech firms, or the “Mag 7”, were known for their high growth, big margins and capital-light business models. The dawn of the AI age changed everything. Mag 7 firms were seen to have a huge advantage over smaller or mid-tier firms because they had the balance sheets to outspend all others. Indeed, detractors of US big tech firms have long argued that they were using capital as a weapon to keep rivals at bay. 

Gave of Gavekal Research says DeepSeek seems to bring things back to what tech should be: open-source and capital-light. “This is ironic since it is coming out of China, where the popular perception is that growth is always capital-intensive and policy is anything but open-source!” DeepSeek, he notes, “destroys the idea that although having a few tech titans controlling the broader tech ecosystem might be socially pernicious, at least it makes for technological progress and ensures that the US remains technologically dominant”. In the current Sputnik moment, Gave says “the tech war between China and the US, between open-source and closed-system, and between capital-light and capital-intensive models has just taken a very interesting turn.”.

So, what’s next? Edison Lee, tech analyst for Jefferies & Co in Hong Kong, believes that the US government will soon ban DeepSeek. Already, the US federal government, US Navy, US Defense Department, or Pentagon, Nasa and the state of Texas have banned the use of the DeepSeek app. Cybersecurity and national security experts have raised concerns about the risks of using the app. Several other US states are also looking at a ban. Outside of America, Taiwan and Australia were among the first to announce a ban. France, Belgium, and the Netherlands have also raised concerns about the app’s data collection practices and hinted that they might take action. “It is highly likely the US would use the law that bans TikTok to ban DeepSeek,” Lee says. 

He argues that in order for America to ensure its global AI dominance, the Trump administration would likely ban DeepSeek, at least for consumers. However, he says, AI developers may not be restricted, as they could self-host the model and thus eliminate the security risk. “That will allow US AI developers to take advantage of DeepSeek’s low-cost capabilities,” he adds. “Even if no US companies are allowed to host DeepSeek, US AI players could still replicate its tech to develop low-cost AI.” That would not change the likely result that model efficiency and lower AI cost would become the key focus of US AI players. Indeed, even OpenAI stressed its latest o3-mini model is 63% cheaper than its o1-mini, although still almost twice as expensive as DeepSeek’s R1.

Banning DeepSeek’s app might only be the first step. Dario Amodei, CEO of Anthropic, an AI start-up whose largest investors include Amazon and Google, says while DeepSeek’s breakthrough seems impressive, it is merely “an expected point on an ongoing cost reduction curve”, which American firms like his would soon match. Amodei argues that DeepSeek’s sudden emergence makes chip “export control policies even more existentially important than they were a week ago”. If the US restricts the sale of all kinds of graphics processing units to China, including consumer-grade GPUs, and strictly enforces the AI diffusion policy, it will be hard for China to not only do any AI training but also inferencing. If that’s the case, AI growth in China will be severely limited, including Edge AI, which can be used in smartphones. 

So far, China has had access to Nvidia’s lower-end H20 chips but there have been calls in recent days from US politicians for President Donald Trump to ban the sale of low-end chips as well. Without access to data centres outside China for training purposes, the key for China is access to H20 chips and some of the consumer-grade gaming chips that can be used for inferences, says Lee. Over the longer run, China, the Jefferies analyst adds, needs to focus on expanding its own 7nm chip foundries even as Nvidia relies on TSMC’s cutting edge 2nm foundries.  The latest Sputnik moment will accelerate AI adoption and advances in robotics, software and services.

Assif Shameen is a technology and business writer based in North America 

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's App Store and Android's Google Play.