news

tencent technology news, september 12, according to foreign media reports, generative artificial intelligence has triggered one of the largest consumer booms in modern american history. companies and investors are betting hundreds of billions of dollars, believing that this technology will reshape the global economic landscape and has huge potential for profit. but the question is: whether and when will this huge investment bring returns?

applications such as chatgpt, a chatbot owned by openai, have attracted hundreds of millions of users, but the user base willing to pay for advanced services is still limited. at the same time, the business community is still in the exploratory stage, committed to exploring the potential of generative artificial intelligence in improving production efficiency. despite this, technology giants are not stingy and are injecting funds at an unprecedented level, focusing mainly on the development of cutting-edge hardware to support the development and operation of artificial intelligence models.

“we would rather take the risk of overinvesting than the risk of underinvesting,” sundar pichai, ceo of google and its parent company alphabet, stressed during the company’s latest earnings call.

venture capitalists widely expect that the valuations of at least a few artificial intelligence startups will jump to hundreds of billions or even trillions of dollars in the next few years, even though most of them are not yet profitable.

so far this year, ai startups have raised $64.1 billion in venture capital, a figure that is approaching the all-time peak set during the broader investment boom in 2021, and the share of venture capital in ai this year has climbed to an all-time high.

the results of these massive investments are starting to show up in new data centers across the u.s. unlike traditional data centers that primarily store data and run non-ai software, ai-optimized data centers are equipped with cutting-edge chips designed specifically for developing and running generative ai applications.

specifically, microsoft has more than doubled its data center footprint since the beginning of 2020, and google is not far behind, growing by 80% over the same period.oracle corporationit has also focused its strategic focus on data center business and plans to build 100 new data centers.

compared with traditional data centers, ai data centers consume more energy because ai chips require an uninterrupted and stable energy supply to maintain their efficient operation. any short-term fluctuations in power supply may have an adverse impact on the "training process" of ai models, which optimizes their performance through massive data analysis. this risk is particularly prominent for large models that are extremely expensive and cost tens or even hundreds of millions of dollars per training.

since 2015, the amount of electricity ordered by data centers in the united states and canada from energy companies has surged nearly ninefold, a trend that directly reflects the sharp increase in power demand for data centers due to the development of artificial intelligence.

nvidianvidia has become the dominant force in the field of ai model training and running chips. although its gpus (graphics processing units) originally served the video game field, their excellent performance has led to high-end gpus costing tens of thousands of dollars. today, technology companies dedicated to building and hosting ai models are competing for nvidia's chip resources to meet the growing demand.

ceo of metamark zuckerbergmark zuckerberg has publicly announced that his company aims to have 600,000 gpus by the end of 2024 to support its ai strategy.teslaceo and founder of artificial intelligence startup xaielon musknot to be outdone, elon musk said he plans to purchase 300,000 gpus by next summer.

highly skilled talent has also become a scarce resource in the market. despite the recent wave of layoffs in silicon valley, tech giants are still spending millions of dollars to recruit research scientists who can lead the exploration of new frontiers in artificial intelligence. many of these experts previously worked in academia. now, they are among the world's highest-paid technical talents.

even if you mastermachine learningprofessionals with basic knowledge can easily get six-figure salary positions. it is worth noting that compared with the same period last year, the number of new recruitments for ai-related positions in july surged by nearly 50%, which is in sharp contrast to the slight decline in overall recruitment in the technology industry during the same period, highlighting the market's high demand for ai talents.

investors’ patience with silicon valley’s massive investments in artificial intelligence is wearing thin, especially as meta, microsoft and other companies continue to increase spending on ai even as revenue growth lags. this has been reflected in the stock prices of these companies.

sequoia capitala partner recently analyzed that in order to justify the investment in data centers and chips this year, the artificial intelligence business will eventually need to generate annual revenue of up to $600 billion. although most companies do not disclose their revenue from artificial intelligence, analysts estimate that the total annual revenue is at most tens of billions of dollars, which is far from expectations.

doubts about the prospects of artificial intelligence can't help but remind people of the internet bubble period 25 years ago, when companies blindly invested in fiber optic networks in order to support overly optimistic expectations for the popularization of the internet, but the actual development was far below expectations.

faced with such doubts, senior executives of technology giants have called on people to remain patient. zuckerberg admitted at the earnings conference that the commercialization of artificial intelligence applications will take several years to show results. pichai also said: "there is a time curve in taking basic technologies and turning them into meaningful solutions." (compiled by jinlu)