--- title: "70 Years of Artificial Intelligence: 14 Lessons from the Boom and Bust Cycle" type: "News" locale: "en" url: "https://longbridge.com/en/news/296090160.md" description: "Deutsche Bank reviews the 70-year development trajectory of AI, distilling 14 historical lessons. AI exhibits exponential, non-linear growth, with declining costs expected to stimulate greater demand; however, frequent iterations in technological pathways are revealing hardware and supply chain bottlenecks. While consumer adoption is rapid, enterprise commercialization remains in its early stages. With current market valuations approaching historic highs, investors should remain vigilant regarding risks associated with technological shifts and supply chains" datetime: "2026-08-17T08:39:13.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/296090160.md) - [en](https://longbridge.com/en/news/296090160.md) - [zh-HK](https://longbridge.com/zh-HK/news/296090160.md) --- # 70 Years of Artificial Intelligence: 14 Lessons from the Boom and Bust Cycle Deutsche Bank's latest research report outlines the development trajectory of artificial intelligence since its inception in 1956, distilling 14 key lessons from historical patterns to provide a reference for investors assessing the direction of the current AI boom. This August marks exactly 70 years since the Dartmouth Summer Research Project on Artificial Intelligence in 1956—the birthplace of AI. According to Zhuifeng Trading Desk, Adrian Cox, Thematic Strategist at Deutsche Bank Research, pointed out in his latest report that the 70-year journey of AI has been characterized by alternating periods of boom and bust. The current wave of investment and valuation enthusiasm is replaying the paradigm of technological revolutions seen multiple times in history. The report argues that "context" is crucial to understanding the future direction of AI. From non-linear growth and infrastructure bottlenecks to the expansion and bursting of valuation bubbles, historical signals are clearly discernible. The report states bluntly that while some may argue "this time is different," data from the past 70 years offers an alternative frame of reference—for investors betting on the AI sector, these insights directly impact asset allocation logic and risk assessment. ## Growth Is Not Linear and Is Often Severely Underestimated **The report begins by highlighting the core characteristic of AI progress: non-linearity.** Presenting historical data on training compute power on a logarithmic scale clearly shows that since 1956, the compute used to train major AI systems has grown by dozens of orders of magnitude, a trend almost completely obscured by linear chart visualizations. Exponential growth is intuitively easy to underestimate, representing the first cognitive hurdle in understanding the AI wave. Closely related is the fact that the pace of AI progress has surpassed Moore's Law. Traditional computing power doubled every 18 to 24 months, but in the era of deep learning, the annual growth rate of compute has reached approximately 4x, far higher than the ~1.4x annual growth rate prior to deep learning. The reason lies in the simultaneous improvement of multiple factors—such as increased system scale, enhanced memory, and algorithm optimization—creating a compounding effect. ## Technological Pathways Continue to Iterate; Today's Leaders May Not Be Tomorrow's Winners **The report presents the evolution of pathways over the past 70 years through an AI technology genealogy map: from symbolic logic and expert systems to statistical machine learning and deep learning, and finally to the currently dominant Large Language Models (LLMs).** Each generation of mainstream technology has experienced a cycle from rise to replacement. Some pathways (such as Recurrent Neural Networks) have been surpassed, while others continue to evolve in parallel. The report points out that LLMs may eventually give way to new paradigms such as "World Models," and technological generational shifts do not depend on the will of current leaders. Historical shifts in market share corroborate this point. Internet Explorer once overwhelmed Netscape, only to be subsequently replaced by Chrome. In the current competitive landscape of generative AI platforms, ChatGPT leads in monthly visits, but Google Gemini, DeepSeek, and Claude are catching up rapidly. Early advantages do not equate to long-term moats. ## R&D Accumulation Determines the Competitive Landscape; DeepSeek's Rise Is No Accident **The sudden emergence of Chinese AI models—represented by DeepSeek—may appear superficially as "overnight fame," but it is actually the result of years of accumulated R&D investment.** Data shows that China surpassed the United States in total R&D expenditure in 2024, and the pace of catching up in the number of major AI models is equally significant. In terms of the number of AI patents granted, China's growth curve also significantly leads other economies. For investors, this means that changes in the competitive landscape often accumulate beneath the surface for years before becoming visibly apparent. ## Declining Costs Do Not Compress Demand; Instead, They Expand It The report cites the "Jevons Paradox" to illustrate that **a significant decrease in the cost of using AI will not lead to a reduction in total expenditure, but will instead stimulate a surge in demand.** Since 2006, the cost of GPU compute has dropped by over 99%, but according to forecasts by the International Energy Agency (IEA), global data center electricity consumption will double from 2024 to 2030. Lower marginal costs mean more application scenarios and higher aggregate demand. ## Software Revolutions Require Hardware to Lead; Current Bottlenecks Are on the Supply Side Every technological revolution has relied on large-scale hardware investment, and AI is no exception. Since ChatGPT launched in November 2022, the total return of industries related to data centers, hardware, and chips within the Russell 1000 Index has far exceeded that of the software industry. Unlike historical constraints caused by the adoption of consumer devices, the current bottleneck lies more on the supply side of AI chips: the hourly rental price of Nvidia's H100 GPU chips has continued to fluctuate over the past few quarters, reflecting structural tensions between supply and demand. ## Globalization Lowers Technology Costs but Also Embeds Supply Chain Risks As the AI boom heats up, US semiconductor imports have surged. Meanwhile, global supply chain concentration continues to rise, with products becoming increasingly dependent on single sources. The report points out that while globalization has made technology cheaper, it has also exposed supply chains to higher geopolitical and concentration risks. ## Technological Dividends Take Time to Materialize; Commercial Implementation Is Still in Early Stages Technological revolutions typically exhibit a "J-curve" effect on productivity: costs appear before benefits. Citing data, the report notes that currently, less than half of US employees use AI to complete work tasks, spending an average of 6% of their work time on AI-related tasks, which has saved only about 2% of work hours. Consumer adoption speed has hit historical records, with generative AI spreading faster than the internet and personal computers. However, the promotion of enterprise-side applications that can be directly monetized is significantly slower—downloading an app takes only minutes, whereas restructuring enterprise operational processes around new technologies takes years. AI application rates lead in the information, professional services, and financial insurance industries, while manufacturing lags behind. ## Valuations Are at Historic Highs; Revenue Expectations Need to Be "More Different" to Be Realized From the perspective of the Shiller Cyclically Adjusted Price-to-Earnings Ratio (CAPE), the current valuation level of the S&P 500 is close to the peak of the 2000 internet bubble, situated in the highest range in 150 years. **Historically, every valuation peak has almost corresponded to a transformative technological wave: electrification in 1899, radio and automobiles in 1929, the electronics wave in 1966, the internet in 2000—and AI in 2026.** At the revenue level, the current revenue growth rates of OpenAI and Anthropic have already exceeded the historical peak levels of comparable companies during the same period. However, the report also points out that to realize the long-term revenue expectations implied by current valuations, the growth curves of these two companies need to be "more different" compared to the historical trajectories of tech giants like Google, Meta, and Nvidia. ## Long-Term Trends Are Resilient, but Extreme Scenarios Have Entered the Discussion The report concludes with the long-term trend of S&P 500 earnings per share, showing that this metric has grown at an annual rate of approximately 6.5% since 1935, maintaining trend stability through multiple wars and recessions. At the GDP level, the report outlines three AI scenarios: the baseline trend, a moderate acceleration driven by AI (with an average annual growth rate of about 2.1% over the next 10 years), and two extremes under the "Singularity" scenario—ranging from a technological utopia to human extinction. The report maintains a neutral stance on this but prompts investors: between historical trends and extreme scenarios, which end of the spectrum current market pricing is closer to is a core issue worth continuously tracking. ### Related Stocks - [FAS.US](https://longbridge.com/en/quote/FAS.US.md) - [FAZ.US](https://longbridge.com/en/quote/FAZ.US.md) - [SEF.US](https://longbridge.com/en/quote/SEF.US.md) - [SKF.US](https://longbridge.com/en/quote/SKF.US.md) - [UYG.US](https://longbridge.com/en/quote/UYG.US.md) ## Related News & Research - [3 arguments for and against AI watermarks](https://longbridge.com/en/news/296205500.md) - [Fortinet acquires AI runtime security firm Virtue AI](https://longbridge.com/en/news/296110529.md) - [The AI boom has echoes of Enron — but that might be okay, a tech guru says](https://longbridge.com/en/news/296021200.md) - [Alphabet vs. Oracle: Which is the better AI stock to own for the next 5 years?](https://longbridge.com/en/news/296001795.md) - [In the AI Era, Duolingo Is Betting on Humans](https://longbridge.com/en/news/296129188.md)