By A new AI model from China rattled US markets this past week.
DeepSeek represented a breakthrough for the AI industry, delivering results on par with OpenAI’s best model while using significantly less computing power.
The implications have been substantial. Nvidia stock shed nearly $600 billion in market value in a single day as investors worried about future demand for its GPU chips, which are the main fuel source for large language models.
Suppose large language models from OpenAI, Anthropic, and Meta adopt some of the techniques utilized by the open-sourced DeepSeek and become more efficient. Will they need as much computing power as initially thought?
On the flip side, AI adopters, particularly software companies, surged on the prospect that the cost of AI technologies would decrease significantly, leading to higher profit margins.
But there are even broader lessons investors can take from this week’s black swan event, and they impact everything from bonds to stocks and even crypto.
Watch your concentration
The concentration of a handful of mega-cap tech companies dominating the stock market has crept up to historic levels over the past few years.
According to Goldman Sachs data, the top five stocks in the S&P 500 made up about 29% of the index as of December 31, and they’re all highly exposed to similar technology trends.
“Concentration in several large names is a concern when the drivers of success are the same for most of the names,” Chris Fasciano, chief market strategist at Commonwealth Financial Network, told Business Insider.
Concentration is a double-edged sword. It can work great in bull markets, but a simple disruption to investors’ overarching narrative could lead to a painful decline, as seen on Monday, when the index tracking mega-cap tech stocks plunged 3%, compared to a slight gain for the equal-weighted S&P 500.
The concentration speaks to the idea that investors may not be as diversified as they believe.
“This is an underappreciated consideration,” Steve Sosnick, chief strategist at Interactive Brokers, told BI.
“It is common — and understandable — for investors to believe that they are adequately diversified when they buy an S&P 500-linked fund,” Sosnick said, but that’s not the case based on the extreme concentration levels.
“Those seeking true diversification need to look beyond SPX,” Sosnick said, adding that the Nasdaq is not the answer, with a handful of stocks making up about half of that index.
The AI story has layers
Nvidia has dominated the AI story since ChatGPT was released in November 2022, as it turned into the key “picks and shovels” supplier to companies developing their own large language models.
But Monday’s price action revealed that there are second-derivative beneficiaries to the AI trade, which includes the adopters of AI technologies like software companies.
If DeepSeek’s claims of using less computing power are accurate, that means costs should come down considerably for AI adopters, pushing profit margins higher.
Among the software winners amid Monday’s bloodbath was Salesforce, which gained as much as 10% while the market swooned.
“If DeepSeek has truly shown us that an open-source solution is far less resource-dependent than the ChatGPT paradigm, then the benefits of AI might be easier to reap from companies unwilling or unable to afford to partner up with one of the current firms that currently dominate the arena,” Sosnick said.
Crypto is no safe haven
Crypto volatility was on full display this week, even though DeepSeek’s appeared to mainly impact tech firms.
Bitcoin dropped as much as 7%, in tandem with the sharp decline with the Nasdaq 100.
The sell-off highlights that cryptocurrencies are highly correlated to moves in the broader tech space, and absent any crypto-specific catalysts, bitcoin is a high-octane bet on tech’s continued outperformance of
“Safe havens don’t have an average daily volatility of 2%,” Sosnick said. “Over the past 6 months, the correlation of the price levels of bitcoin and NDX is about 80%.”
Don’t count out Apple
Apple stock bucked the DeepSeek sell-off, rising about 8% in the week as investors determined that the company’s decision not to spend tens of billions of dollars on building its own large language model was a good one.
Instead, Apple is focused on leveraging ChatGPT’s AI model and using edge computing to deliver quick answers to iPhone users. That edge computing is key, as it could become even more ubiquitous if AI models see a decline in their compute consumption thanks to the breakthroughs identified by DeepSeek.
And a decline in compute consumption ultimately means a decline in costs, which could fuel even more stock buybacks and dividends for the company.
“Cheap AI means a lot more capital to return to shareholders, either in the form of dividends or stock buybacks. Every US Big Tech company except Amazon already has both in place. Now they can get much, much bigger,” Nicholas Colas, co-founder of DataTrek Research, said in a note this week.
Cheap AI could fuel a bond market rally
If DeepSeek’s open-source model delivers the expected cost savings to other large language models, it could ultimately mean higher efficiency and lower costs.
According to tech CEOs, like Microsoft’s Satya Nadella, cheap AI models should further accelerate the adoption of AI across various industries and lead to a surge in productivity.
If that happens, economic growth could continue without sparking a rebound in inflation, which would put interest rates back on a downward trend.
“If DeepSeek’s processes show that the AI revolution can be achieved at a lower cost and with lower energy usage, that would be good for macro conditions (such as productivity and inflation) going forward,” LPL chief strategist Jeff Buchbinder wrote this week. “In theory, this would help reduce the Fed’s neutral rate a touch and keep interest rates lower than they otherwise would be if big AI spending continued unabated.”
Lower interest rates mean higher bond prices, sparking a potential rally in an asset class that was hurt by the Fed’s rate hikes in 2022 and 2023.
“More egalitarian AI can and should boost productivity throughout the economy and thus have benefits for price levels and interest rates,” Sosnick said, though he cautioned that such an outcome is “not going to happen immediately.”
