Did China’s Deepseek crash and burn President Trump’s $500 billion AI Stargate project?
Shortly after his inauguration on January 20th, President Donald Trump held an event at the White House, featuring some of the biggest names in the tech industry.
Openai CEO Sam Altman, Oracle Founder Larry Ellison and Japanese technological magnate Masayoshi led the billing of an infrastructure project called Stargate, which aims to invest $500 billion in American technology companies over the next four years. It’s there.
The initial response to this initiative was positive, but the happiness was short-lived. A few days after Stargate’s announcement, investors focused on a new AI startup from China called Deepseek.
Below we explore some of the major topics surrounding Deepseek and evaluate whether the latest AI darling poses a threat to Trump’s Stargate agenda.
What makes Deepseek the point of such competition is that it claims it trained its models using older hardware compared to what US AI companies are using. For example, you’ve probably heard business leaders talk about it over the last few years. nvidia‘s (NASDAQ: NVDA) Graphic Processing Unit (GPU) – that is, the architecture of Hopper and Blackwell. These are Nvidia’s most sophisticated and expensive chips.
However, Deepseek’s developers claim they used older GPUs and cheaper infrastructure, primarily from Nvidia, a cluster of H800 chips.
This is important. This is because Deepseek’s team subtly implies that high-scoring AI can be developed because it is far less than what Openai and its cohort spend. If so, it makes sense for investors to expect a big drawback AI infrastructure spending. I acknowledge the logic that supports the bear story here, but let’s look into whether this argument really continues.
Over the past few weeks, several major players in the AI revolution reported full-year revenue for the fourth quarter and 2024.
-
Microsoft: Microsoft is a large investor in Openai, and over the past few years, the company has integrated new AI-powered services throughout its ecosystem. One of the most important areas of Microsoft’s AI utilisation is Azure, the cloud computing business. Over the past 12 months, the company has ploughed over $55 billion into capital expenditures (CAPEX). Many of them are assigned to ongoing AI efforts. It appears that Microsoft executives are committed to these infrastructure projects. This year’s CAPEX budget should be in the $80 billion range.
-
Meta Platform:Meta is not afraid to double its AI vision. In 2023, the company spent $28 billion on CAPEX. Last year, Meta’s infrastructure spending rose 40%. It costs about $39 billion. During the company’s recent revenue calls, Meta management said CAPEX spending in 2025 would range between $6 billion and $65 billion.
-
alphabet:Alphabet has spent the past few years integrating AI services into its own ecosystem, diversifying its business from reliance on advertising, and unlocking new opportunities to compete more directly with Microsoft and Microsoft. Amazon. During the company’s fourth quarter revenue call, investors plan to spend around $75 billion on CAPEX in 2025, representing a 43% increase compared to 2025 spending levels. I learned that.
-
Amazon: Amazon has also spent a significant amount of capital on AI infrastructure over the last few years. In particular, the company has poured billions of dollars into startups, humanity, which competes directly with Openai. In particular, by integrating human integration into the cloud business, the company helped it recapture Amazon Web Services (AWS) sales and expanding profit margins. During the company’s fourth quarter revenue call, management hinted that Amazon could spend around $105 billion this year on AI infrastructure as demand for AWS remains strong. If Amazon spends around the indicated levels, this represents a 27% increase compared to CAPEX in 2024.