January 27, 2025, marked a watershed date for the tech world.
That was when DeepSeek, a relatively new Chinese startup, announced that it was able to develop an AI model at a fraction of the cost relative to estimates of major players in the industry.
DeepSeek said it spent only $5.6 million to power an AI model with capabilities similar to those of products developed by more famous rivals. It was a marked contrast to the major companies in the tech world that had been planning investments of hundreds of millions, if not billions, on their own AI models.
“Driving new cost efficiencies and innovation is important in any tech cycle,” says Morgan Stanley’s U.S. Internet analyst Brian Nowak. “The market’s strong reaction to the news was based on the fact that this breakthrough came from a company out of China that a lot of people didn’t know about. But it’s actually a good sign for the next three to five years of the Gen AI cycle.”
Since then, DeepSeek has begun rapidly integrating into multiple enterprises and sectors, such as data centers and telecom. Companies in industries such as automotive and financial services have adopted these advancements as well. It is also seeing accelerated adoption by consumers, given its very low cost and users’ ability to download a simple version of the model in PCs and smartphones.
Morgan Stanley Research analysts are watching the most significant shifts following the DeepSeek breakthrough, the potential overreactions and, most important, the opportunities that can arise from that event.
1. The AI Market Is Splitting
Shawn Kim, who heads the Asia Technology research team for Morgan Stanley Research, says it’s no longer the case that only a few companies would be able to afford powerful chips and heavy infrastructure to efficiently develop AI.
“The AI market appears increasingly bifurcated,” Kim says. “One segment remains a premium for entities willing to pay substantially to be the first, but for the rest, AI is a cost center, and everyone wants cheaper tokens.”
The premium segment, he notes, is focused on using data and supercomputers that would lead to existential developments, such as the replacement of workers by AI. By contrast, the low-cost AI market, which became more visible after DeepSeek’s announcement, features affordable entry costs, with AI models converging and commoditizing very quickly. That’s good news for the consumers of AI, but it is going to be more difficult for AI providers to make money in that segment.
Kim notes there’s also a third possibility in this new AI economic structure: a market in which large tech companies will likely offer AI for free so that people stay on their platforms longer.
2. Investors Question AI Spending
Given the progress that DeepSeek made with a relatively low budget, investors are scrutinizing companies’ AI investments, while corporate leaders question whether it’s really necessary to spend billions of dollars to reach their AI goals.
But recent earnings reports and calls show that, so far, the major tech companies are sticking to their aggressive plans for capital expenditures (CapEx).
For Stephen Byrd, Morgan Stanley’s Head of Research Product for the Americas & Head of Global Sustainability Research, DeepSeek hasn’t changed the view on AI infrastructure growth. He expects announcements of big data centers and training facilities to keep coming, especially because the demand for AI is expected to increase.
“There has been steady progress in terms of efficiencies, and DeepSeek represents an improvement in that regard,” Byrd says. “This happens in every technology revolution: We see continuing cost reductions and then dramatic increases in demand for the product.”
Before the DeepSeek event, a Morgan Stanley Research survey of chief information officers showed an uptick in IT budget growth expectations for 2025. The survey, conducted between October and December, pointed to AI/machine learning as the main priority for CIOs.
3. AI Drives Clean Energy Demand
As analysts expect companies to maintain CapEx plans for AI infrastructure, that infrastructure will have enormous energy needs—and Morgan Stanley Research still expects that demand to drive investments in clean energy.
Andrew Percoco, Head of North America Clean Tech at Morgan Stanley, says the outlook for power demand associated with AI in the U.S. hasn’t changed as a result of the DeepSeek announcement. He notes there is still a strong need to develop more renewables, including solar and wind, to power AI data centers.
“Overall, it was a scary moment in the market for the AI narrative,” Percoco says. “But I think investors have found it as an opportunity to re-engage in some of these stories.”
Byrd says multiple different sources of energy, such as nuclear and natural gas, will continue to be in demand to power AI data centers.
“When you look at the magnitude of power needs, we’re going to see everything from tiny 20 MW projects to multi-thousand MW data-center projects. We still believe that nuclear very much fits the purpose,” Byrd says. “We do expect to see quite a few announcements of data centers that can be linked to natural gas, as well as nuclear deals and bitcoin-conversion deals.”
4. Humanoids Advance in China
An important lesson from the DeepSeek event is that China is moving very aggressively in data collection. Among other impacts, this will boost its development of humanoid robots – AI “brains” trained on vast sets of real and simulated robotic data to help them understand natural language, learn from and imitate human action, and perceive their dynamic environments.
Morgan Stanley’s research suggests China continues to show the most impressive progress in humanoid robotics, with startups benefitting from established supply chains, local adoption opportunities and strong support from the national government.
Electric vehicles with autonomous capabilities, which require similar technologies as humanoids, are also evolving quickly in China. “If you can solve autonomous cars, you can solve autonomous anything, whether it’s drones, aviation, humanoids and everything else in between,” says Adam Jonas, Head of Global Autos & Shared Mobility Research.
5. Governments Weight Action
The DeepSeek event was also a wake-up call for Western nations.
“Some investors think, rather dismissively, that the Western world will dominate all aspects of AI because it has access to superior computing, and China does not,” Jonas says. “DeepSeek is a symbol of what can be done when a centrally planned government puts very high priority on achieving dominance or leadership in rapidly moving AI-enabled technologies.”
Jonas notes that it’s important to watch very closely what China is doing on the AI front because it affects not just U.S. tech companies’ ability to compete, but even more importantly, government policies that can be supportive of the private sector.
