After businesses and consumers spent 2024 testing out new AI tools and last year rolling them out widely, the technology now faces tough questions about whether it can deliver on its promises. While hundreds of thousands of companies and several hundred million people put AI to work in various ways, some found it helpful while others ran into problems that led to costly mistakes. The coming months will see an intense examination of whether AI systems work reliably enough and make financial sense for the massive amounts of money being poured into them. Investment in AI equipment and infrastructure could reach $500bn in 2026 , making it crucial for the industry to answer three major challenges. Growth strategy hits a wall as investors demand results The first issue centers on whether AI’s growth strategy has reached its limits. In 2019, researcher Rich Sutton published a piece called “ The Bitter Lesson ” that explained how feeding more information and computing power into deep learning systems proved the best way to make them stronger. Companies like OpenAI proved this approach right by creating increasingly powerful systems that needed more and more computing resources. However, Sutton now joins other researchers in believing this method is losing steam. This doesn’t mean AI development will stop making progress. Instead, companies will need to show investors they can write better computer programs and find other ways to advance the technology that uses less energy. Experts predict neurosymbolic AI, which combines current data-based systems with rule-following programs, will get much more attention this year. The second challenge involves whether major players can make money as AI becomes more common and ordinary. Tech giants like Alphabet, Amazon, and Microsoft will keep using AI to lower costs and improve services that already reach billions of people worldwide. But newer companies such as OpenAI and Anthropic, which plan to go public this year, must prove they can build lasting advantages that keep competitors away. Business values across the sector shot up in 2025, but companies will soon be judged more carefully on their individual merits. Chinese competitors win users with cheaper, open systems The third question concerns how American tech companies will handle the growing success of Chinese AI systems that anyone can modify and use. About a year ago, a Chinese company called DeepSeek surprised the industry by releasing a high-quality thinking model that cost far less to train than similar American products. Since then, Chinese systems that are more focused, cheaper, and easier to adjust have grabbed significant market presence. Research from the Massachusetts Institute of Technology and Hugging Face showed that Chinese-made systems, which anyone can access, jumped ahead of American ones, making up 17 percent of all downloads. Even Sam Altman, who runs OpenAI, said his company may have picked “the wrong side of history” by mainly building expensive, private systems that users cannot modify. American firms are now putting out more open systems to compete in this space. AI holds real promise when used carefully. It can make business operations smoother, help workers get more done, and speed up scientific research. But users and investors will now separate services and companies that provide genuine value from those simply riding the wave of AI excitement. Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free .
