According to a September report, artificial intelligence-related capital expenditures (CapEx) and operating costs for Microsoft (MSFT), Meta (META), Google (GOOGL) and Amazon (AMZN) totaled a staggering $125 billion in the first eight months of 2024 Dollar. The cumulative capital expenditures of these four technology giants alone are expected to exceed the US$200 billion mark in 2024 as a whole.
Meanwhile, artificial intelligence startups are receiving unprecedented amounts of funding from investors eager to capitalize on the technology’s lucrative potential. OpenAI is expected to become the most well-funded artificial intelligence company in 2024, with a recent valuation of $157 billion. Its rival Anthropic is preparing a new round of financing that is expected to value it at $40 billion.
Cash-rich leading AI companies now face the challenge of proving to investors and the public that their expensive bets on new technologies will pay off. From the ongoing shift to “agent AI” to the emergence of new scaling laws and the widespread exploration of AI’s myriad capabilities, here’s what 2025 will bring to the world of AI:
Agent AI will be ‘the next big breakthrough’
This buzzword refers to autonomous artificial intelligence assistants capable of completing tasks without human supervision. The potential of artificial intelligence agents to improve the workplace and daily life quickly caught the attention of Silicon Valley, with companies like Salesforce targeting agents as their next major product.
Microsoft has also launched a number of artificial intelligence agents in recent months. In November, it launched several artificial intelligence assistants customized for its Microsoft 365 suite, including an agent that can provide translations in nine different languages.
OpenAI is also on the “agent AI” train, with upcoming models expected to be able to perform tasks such as booking travel and writing code. OpenAI CEO Sam Altman said in a recent AMA on Reddit that artificial intelligence agents “feel like the next big breakthrough.”
According to research firm MarketsandMarkets, the global market for artificial intelligence agents is currently worth more than $5 billion. That number is expected to soar to $47 billion by the end of the decade, driven in part by corporate client demand for agents.
Test-time computing may be the solution to the AI training data crisis
One of the key factors in the success of artificial intelligence in recent years has been feeding large amounts of data into artificial intelligence models. But the amount of text, images, and videos on the Internet is limited. To avoid stagnating technology development, AI companies are turning to other methods to train their models. One of the most promising solutions is test-time computing, where AI models improve through reasoning and spend longer thinking about potential answers before responding—a theory recently demonstrated by OpenAI’s o1 model.
During a November earnings call, Nvidia (NVDA) CEO Jensen Huang described OpenAI’s new model as “one of the most exciting developments” in terms of scaling, noting that “the longer it thinks about , the better and higher quality the answers produced.
Huang is not alone in his optimism. Microsoft CEO Satya Nadella also noted in November that test-time computing is a new scaling law, while OpenAI co-founder Ilya Sutskever said earlier this month It is sometimes emphasized that this is progress in the era of artificial intelligence pre-training.
Synthetic data is another promising solution
Another solution to the AI data crisis is to replace traditional data with information generated by the technology itself. According to BCC Research, the synthetic data market is expected to surge to $2.1 billion by 2028, an increase of more than 450% from 2022.
A year ago, Altman hinted at the potential of synthetic data when discussing the dwindling data supply for artificial intelligence, saying in an interview that “as long as you can span the range of synthetic data events, that is, the model is smart enough to generate good Synthetic data, I think should be fine. According to reports, OpenAI and competitors such as Anthropic, Meta, Microsoft, and Google have begun using synthetic data in some way to train and fine-tune models.
In October, AI startup Writer launched a new AI model trained entirely on data generated by AI. This approach allowed the company to slash model development costs, which totaled just $700,000 compared to the millions doled out by other companies. For example, OpenAI’s GPT-4 model costs more than $100 million to train.
“Big World Model” will create a 3D AI world
Until now, most visual output from artificial intelligence has remained two-dimensional, something technology pioneers hope to change in the coming years. “Big World Models” are an emerging form of artificial intelligence designed to build interactive three-dimensional scenes that advance the worlds of movies, games, and simulators.
One of the biggest players in the field is World Labs, a startup founded by Stanford artificial intelligence pioneer Fei-Fei Li that raised $230 million earlier this year. The company is seeking to build large-scale world models with “spatial intelligence,” a form of intelligence that understands and interacts with the real world. To demonstrate this concept, Lee had previously used the example of a cat reaching down and knocking over a glass of milk, and the human ability to predict the consequences of this event and thus take action to prevent the glass from falling.
In early December, Google DeepMind launched its own large-world model in the form of Genie 2, which simulates a virtual environment and will be used to train and evaluate AI agents. This area is likely to be a key focus for the lab going forward, as evidenced by the recent hiring of former OpenAI researcher Tim Brooks to oversee its video generator Sora. Google DeepMind CEO Demis Hassabis welcomed Brooks to his team in an X post, saying he was excited about “working together to make the long-held dream of a world simulator a reality.”
Artificial intelligence search engines will reshape online search
Google has long held a seemingly unshakable dominance of the search market. But with the advent of artificial intelligence, the proliferation of AI-powered search engines is shaking the tech giant’s footing.
It’s not that Google itself hasn’t embraced the technology. In 2024, it launched AI Overviews, a feature that provides users with AI-generated summaries instead of links. Chief Executive Sundar Pichai predicted the feature would attract more than 1 billion users per month and has “increased overall search usage and user satisfaction,” he told Wall Street analysts in October.
But as companies like OpenAI and Microsoft move into the field with the help of artificial intelligence, Google will have to contend with an increasingly crowded search tools industry.
According to reports, Meta is also preparing to launch its own artificial intelligence search engine, and startup Perplexity AI has emerged as a particularly strong player. Its artificial intelligence search tool, last valued at $9 billion, has processed about 20 million queries per day, up from 2.5 million in early 2024.
