After the uncertainty that the new Chinese AI model Deepseek brought to the markets, investors were eager to understand from big tech companies how this is affecting their strategy and what measures are being taken to keep ahead in the game.
Microsoft, Meta and Tesla – all part of the “Magnificent Seven” companies – reported their earnings last night, and some expressed their thoughts on the new, more efficient, open-source AI model, determined to stay ahead and continue the investments, writes eToro analyst for Romania, Bogdan Maioreanu.
The efficiency of Deepseek, which was much cheaper to train than the current large language models, took the stock markets by storm. According to Anthropic CEO, Dario Amodei, AI models in development can cost up to $1 billion to train. Current models like ChatGPT-4o cost about $100 million, but he expects the cost of training these models to go up to $10 billion or even $100 billion in as little as three years from now. And then comes Deepseek with an estimated cost of training of around 6 million dollars. Now we have seen the statement from Open AI mentioning that Deepseek might have used Chat GPT for training. Regardless, once the door to cheaper AI models that are open source and can work on local hardware has been opened, it will be very difficult to close.
Microsoft – a major investor in Open AI – published its earnings report that beat expected revenues, up 12%, but investors weren’t convinced. Notably, Azure’s cloud growth slowed, and the stock plunged 4% in after hours trading. Satya Nadella, Microsoft CEO insists that the company is “positioned for long-term AI growth,” but with competition heating up, the market isn’t convinced. However, he is convinced that Deepseek innovation will lead to computing prices for inference going down, meaning people will be able to consume more. This means that AI will be more ubiquitous and for a hyperscaler like Microsoft, a PC platform provider, this is all good news.
Meta Platforms CEO Mark Zuckerberg mentioned that it’s too early to predict how new competitors like DeepSeek will impact Meta’s AI investments, but he expects continued heavy capital expenditure spending on both pre-training and inference to keep its strategic advantage. Zuckerberg noted that introducing AI capabilities in a platform that is serving billions of users may be costly but it also sets Meta apart, and the rapidly evolving AI field requires constant learning from every new model release. Meta posted very good financial results. Revenue hit $48.39 billion, beating expectations. AI-powered ads and Instagram Reels are booming, and the stock rose 2%. Meta is doubling down on AI, planning to spend up to $65 billion on infrastructure. Mark Zuckerberg says, “This is the year AI assistant reaches more than 1 billion people and I expect Meta AI to be that leading AI assistant”
AI was also a focus point for Tesla, but Deepseek was not specifically mentioned in the conference call on results. However, Tesla’s CEO Elon Musk said the company is more than doubling down on autonomy, and that it will see a payoff from investments in AI. In its financial reporting, Tesla missed revenue expectations, with its auto business under pressure from competition, but despite this, the stock jumped 4% after hours. The reason might be that Elon Musk is all in on AI and robots. With his known optimism Musk sees a path for Tesla to have a market cap more than the next five most valuable companies combined, pointing to the potential for an “epic” 2026 and a “ridiculous” 2027. As for the FSD – full self-driving autopilot, as a first step, Musk said that the Texas capital, Austin will see an initial launch of an unsupervised FSD in June and expects California and other states to follow. But we have heard the same statements before and yet the FSD is not ready to be used unsupervised.
For Nvidia, the main chip provider for AI operations whose stock has fallen sharply since the launch of Deepseek, cheaper AI models might be an opportunity. One of the things that has slowed AI adoption within big organizations so far has been how expensive these models are to run. This has made it hard for businesses to justify high AI expenses without having a clear case for a positive return on investment. But as we have seen from statements last night, the demand for computing power might skyrocket. And this is good news for Nvidia and the data center ecosystem. With a more efficient AI model, in the short term, Nvidia might earn less from the sale of GPUs used for training, but the massive future revenue from selling significantly more GPUs for inference might have just come closer. While the chip manufacturer stock is still under pressure following the uncertainties regarding the impact of more efficient models on its sales, some analysts see the move as a correction rather than a selloff. But the uncertainty might continue as in the race for AI supremacy, another giant – Alibaba – launched a new AI model, which it claims to beat Deepseek in performance.
While stocks in the companies at the center of the data centers ecosystem have suffered recently, this year might still be a good year for AI as the competition intensifies and the leaders in this emerging sector are fighting to stay ahead. In fact, according to the latest eToro Retail Investor Beat survey, 55% of global investors and 58% of the Romanian ones are seeing AI stock prices increase in 2025, while 20% consider that these will stabilize. _
Three suspects detained in Netherlands over museum theft. No word on the treasure
