Four Key Issues AI Must Face in 2025
Analysis suggests that the improvement methods of AI have already produced a "diminishing returns" effect, and the development of AI has begun to encounter bottlenecks. Meanwhile, the "killer applications" of AI are still a long way off. However, next year, NVIDIA's GPUs are still expected to continue dominating the market, and if there is sufficient liquidity, the bull market for AI stocks will also continue
On December 27, the Financial Times reported that there are no surprises in the AI field for 2024 like the emergence of ChatGPT, and it is expected that the momentum of AI development will further weaken in 2025, although advancements in underlying technologies will keep the field vibrant. In 2025, AI will face four key issues:
Will the development of artificial intelligence encounter bottlenecks?
As the cost of training large models continues to rise, even the most optimistic individuals in the tech industry have begun to worry that relying solely on increasing data volume and computing power to enhance AI performance is showing diminishing returns, a method that was previously a reliable means of improvement.
Analysis suggests that this indicates that the development of AI is beginning to encounter bottlenecks. However, the analysis also points out that at least in the next 12 months, advancements in other areas of AI may compensate for this shortcoming.
Some models, for instance, perform a series of operations before outputting answers, allowing them to query and refine initial responses to provide more "reasonable" results. OpenAI's o3 system has demonstrated the potential to optimize AI responses through multi-step reasoning, while Google is also developing AI agents that can complete tasks for users by tracking their browser behavior.
However, the report also cautions that whether AI "reasoning" can truly compare with human reasoning remains debatable, and it will take time for these system demonstrations to translate into real products.
Will there be a "killer application" for artificial intelligence?
Currently, generative AI has been applied in areas such as copywriting and image editing, but some analysts believe these tools are only "occasionally useful" and far from changing human life significantly.
The report suggests that next year may see some smarter AI applications, such as virtual memory banks or life assistants capable of absorbing all user digital information and learning behavior patterns. However, due to concerns about the reliability of the technology, both tech companies and users may adopt a cautious attitude. This also means:
"2025 is more likely to be a year where AI functions continue to permeate various applications, without the emergence of a true AI 'killer application', nor are there sufficient signs indicating that the AI era has arrived."
Will NVIDIA's GPUs continue to "dominate" the tech world?
As a leader in the AI chip market, NVIDIA faces challenges from major tech companies, many of which are beginning to design their own AI chips.
According to chip manufacturer Broadcom, three large tech companies plan to use their internal chip designs to create supercomputing clusters containing 1 million chips each by 2027, which would be the largest AI chip cluster to date, ten times the size of Musk's Colossus system.
However, the report believes that NVIDIA is developing rapidly, and with continuous technological innovation and the upcoming Blackwell product line, the company is expected to maintain its market leadership in 2025, with other competitors still unable to match it.
NVIDIA's strong software ecosystem will remain an important moat, and the company is expected to welcome a new round of significant product cycles by the end of the year
Will the AI Boom in the Stock Market Continue?
Analysts believe that in 2025, the AI competition among large technology companies will continue, becoming one of the main drivers of AI capital expenditure. At the same time, some companies are beginning to claim significant results from applying AI in their businesses, which may stimulate more enterprises to increase AI investment.
However, there is uncertainty about whether investors will continue to favor AI stocks, including the market's confidence in the Trump 2.0 government's tax cuts and deregulation policies, as well as the direction of the Federal Reserve's monetary policy.
Reports suggest that this means the stock market may experience significant volatility and some important adjustments in 2025. However, if liquidity remains abundant, Wall Street may continue to bask in the AI boom for some time.