Gavekal: Another Sputnik moment

In 2018, the arrest of Huawei CFO Meng Wanzhou in Canada and the ensuing US ban on high-end semiconductor exports to China transformed Donald Trump’s “trade war” into a “tech war.” At the time, the US clearly felt it had a comparative advantage in technology, and that if it had to fight a battle against China, then picking tech as the battlefield made good sense.

In September 2021, US commerce secretary Gina Raimondo, declared that “If we really want to slow down China’s rate of innovation, we need to work with Europe,” so roping Europe into a cold war most European businesses— not least ASML—would have rather avoided.

Since then, China has pulled ahead in 5G (even announcing a 6G satellite-to-earth breakthrough at the beginning of January), high-speed rail (with new trains going 450 km/h), electric vehicles (triggering the imposition of new trade barriers by the US and Europe), batteries (more trade barriers), and drones.

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?

Still, while China’s ability to leapfrog the West in autos, industrial robots, solar panels and batteries has been broadly accepted by investors as the new reality, few investors imagined that China could leapfrog Western countries when it comes to AI solutions, and certainly not as soon as 2025.

But this is precisely what China has just achieved with the release of DeepSeek, an open-source AI LLM built for a tiny fraction of the cost of its larger, much-hyped competitors; a release which profoundly challenges the market narrative that China is incapable of creativity because of the country’s repressive political regime (see Prejudice And China).

One irony of the DeepSeek release is that China’s AI solution is the kind of open-source system that OpenAI was intended to deliver, until Sam Altman elected to transform the non-profit into a more typical corporation. This triggered a massive feud with Elon Musk, one of the original founders and funders of OpenAI. A feud which might just explain why over the weekend everyone’s Twitter feed was overwhelmed with tweets about the importance of DeepSeek and how OpenAI is essentially heading down the primrose path.

Tech cannibals

This feud increasingly gives the impression that we have reached the stage of the big-tech bull market where in order to keep growing, its leaders have little choice but to cannibalize each other. At least, this the feeling given off by last Tuesday’s press conference, at which Donald Trump, flanked by Masayoshi Son of Softbank, Larry Ellison of Oracle and OpenAI’s Altman openly discussed how to carve up the carcass of TikTok, even though TikTok is still alive and showing little inclination to give in to US demands.

At the same press conference, Trump and the assembled tech CEOs unveiled a US$500bn AI capex spending plan dubbed Stargate. Until just 10 minutes ago, this announcement would likely have triggered a sharp rally to new record highs for Nvidia, Microsoft and other US mega-cap tech stocks perceived as “AI winners.” If nothing else, US$ 500bn in capital expenditure is a massive number.

In the event, AI plays did pop higher on the news. But the rally stopped well short of record highs, and by after-hours trading following Friday’s close, most tech stocks, including Nvidia and Microsoft, had given back their gains to trade down from the Stargate announcement.

This collective shrug of the market’s shoulders might be because the release of DeepSeek undermines one of the core tenets of the AI bull market: that spending tens, if not hundreds, of billions of US dollars will build ever-bigger moats for US mega-cap tech stocks. DeepSeek just shredded this belief. It now looks as if the hundreds of billions spent by the giants of US tech is essentially so much capital that will never see any returns – the “metaverse” fiasco all over again, only larger.

This marks an important shift in market narratives. Essentially, DeepSeek has taken a hot needle to the AI bubble, and the air will now be leaking out. This possibility raises the question whether the deflation of the AI bubble will remain a concentrated market phenomenon, with names like Nvidia coming back down to earth in isolation, or whether it will impact the broader market, with a lower Nasdaq, a weaker US dollar and stronger US treasuries.

The outperformance of US growth stocks started in 2007, at a time when valuations had been thoroughly rinsed out.

This bull market initially had two important drivers: the launch of the first Apple smartphone and the US shale revolution, which meant that the US suddenly had a cheaper cost of energy than almost everyone else and was set to record an improvement in its trade balance.

By 2015, investors realized that returns from investments in US energy would most likely disappoint, and the markets suffered a bit of a hiccup. But the combination of the “Shanghai agreement” with the promise of deregulation and Trump’s first-term tax cuts got things going again. By 2020, the market had entered a typical “blow-off phase,” characterized by sheer silliness: the GameStop saga, the AMC stock surges, exercise bikes with iPads trading at a market cap of US$49bn, Metaverse spending etc.

Then, in 2022, the air came out of a lot of the flimsiest ideas and the more egregious Ponzi schemes, and the market rolled over. So far, the bull market was thoroughly “text book”: 10 years of solid outperformance, a couple of years of exuberance, then the implosion.

What was not text book was the back-to-back very strong years in 2023 and 2024. These two years of impressive returns led by a handful of mega-cap stocks left the broad market indexes more concentrated than at any time in history. And the catalyst for this rebound? The release of ChatGPT and the sudden excitement over all things AI-related. All of which brings me back to the release of DeepSeek, and the following questions.

1.Has AI reached a dot-com moment?

At the end of the 1990s, hundreds of e-commerce platforms were funded by eager VCs. When Pets.com and others started to fail in 2000, and dot-com companies started to run out of money, companies like Lucent and Sun Microsystems that had extended credit to fast-growing customers suddenly found themselves in deep trouble on two fronts.

First, their former clients were fire-selling their new equipment for cents on the dollar. Second, the clients stopped making the payments they owed.

Balance sheets and income statements came under assault at the same time. Could history repeat, but this time with failing VC-funded AI companies selling the high-priced chips they were rushing to buy just months before on the fear of shortages?

2.If the AI bull market stalls, what will take over?

There are plenty of other exciting stories today: the rise of autonomous driving, Chinese stimulus, US deregulation, the impact of steeper yield curves on financials. Some of these themes center on the US. Others less so.

3.If the next “new new thing” is not in the US, what will it mean for the US dollar?

Last week the US dollar gave back some of its recent gains. Arguably, the US dollar’s rollover was linked to Trump’s softening stance on tariffs. But given the news on DeepSeek, it may not be all that surprising. If China can develop a more efficient AI at very low cost, it calls into question a lot of the recent rhetoric about the need for very high-end chips, one key US comparative advantage.

Whatever the reason, the fact remains that over the past year, the US dollar and US growth stock relative performance have been heavily correlated. So if, as seems likely, the Nasdaq now starts to underperform on the back of the DeepSeek news, then given recent trends, the US dollar will most likely follow.

In closing, I would like to highlight that one of the odd things about the AI bull market has been its capital intensity. Up until a couple of years ago, one of the main attractions of tech companies (along with strong growth) was the “capital-light” nature of their business models (see Gavekal’s 2005 book, Our Brave New World). But AI turned this on its head.

All of a sudden, the market seemed keen to buy semiconductor companies on triple-digit multiples, based on the premise that the tech world would now have a capital intensity higher than a steel plant or an oil refiner.

On the positive side, 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!

Perhaps most importantly, DeepSeek 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 this latest Sputnik moment, 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.

Louis-Vincent Gave is a founding partner and CEO of Gavekal Research. This article was originally published by Gavekal, which graciously agreed to its reposting.