The stock market earthquake caused by the Chinese artificial intelligence (AI) breakthrough DeepSeek and its promise of a low-cost path for training and using this technology has not changed the strategies of the major U.S. tech companies one bit. Following the February earnings reports, the path shown by Microsoft, Amazon, Google, and Meta is unanimous: continue with multi-billion-dollar investments to prepare their infrastructure for a skyrocketing demand.
The numbers are staggering. If we combine the investment promises of these four companies, they will allocate approximately $330 billion this year to grow their infrastructures, with the majority earmarked for AI (for comparison, in 2024, the Spanish state collected around $280 billion in taxes).
This calculation does not include other operations like the much-discussed project Stargate, a multi-year investment of $500 billion carried out by Oracle and OpenAI, endorsed by U.S. President Donald Trump.
Amazon will be the most aggressive in deploying investments. The company surprised the market early last Friday by announcing that it would spend $100 billion in 2025, "the vast majority" to expand its AI infrastructure.
The group already put $25 billion on the table in the last quarter of the year, and its CEO, Andy Jassy, stated that Amazon Web Services needs to integrate new data centers into the network, such as those planned in Aragon, as the lack of capacity currently limits its growth. "This is a once-in-a-lifetime opportunity," he said regarding what AI means for the company.
Microsoft is not lagging behind either. The company will inject $80 billion in investment this year, mostly for data centers geared towards AI, facing the same issue of accelerated growth as AWS.
"Customers who purchased Copilot in the first available quarter have increased their demand tenfold in the last 18 months. Novartis is buying thousands of licenses every quarter and has 40,000. Barclays, Carrier Group, Pearson, or the University of Miami have each purchased 10,000 just this quarter," revealed Microsoft's CFO, Amy Hood, who also mentioned that the group has been buying land for three years to build data centers.
Google's executive was even more open about the challenges they are facing in meeting the growing demand for AI solutions with their current infrastructure. "We ended the year with more demand for capacity than we have. We are in a moment of tension between supply and demand and working very hard to secure more capacity," stated Alphabet's CFO, Anat Ashkenazi.
In Google's case, it is also noteworthy that they are the only one of the major cloud providers in the race (at least for now) to compete with their own models against companies like ChatGPT with the Gemini family.
"We invested more in 2024 and will invest more in 2025," said their CFO. The tech giant plans to deploy $75 billion, a significant figure considering they are the third player in the cloud market, considerably behind Microsoft and Amazon.
Meta, a bit more cautious [but only slightly], plans to deploy between $60,000 and $65,000 million in investment, acknowledging that the final investment will depend on how the scenario evolves in the coming months regarding the adoption of their AI models.
There was a broad consensus that the costs of using AI will continue to decrease. Beyond the well-known Jevons Paradox, where lower costs lead to higher production, Silicon Valley's leading minds provided insights into how they see the future of AI and its costs.
Google's CEO, Sundar Pichai, acknowledged that cost reduction would be beneficial for companies to see returns on their AI investment. "One reason we are so excited about the AI opportunity is because it will lead to extraordinary use cases, and for that, it's good that costs are decreasing," he emphasized.
"We have been seeing significant efficiency gains in training and inference for years. For each hardware generation, the inference price drops by half, while in the case of models, they become ten times cheaper due to software optimization," explained Microsoft's top executive, Satya Nadella.