Tech Sector Price-to-Earnings Discrepancies
I am not a financial advisor and these are personal observations. However, there is value in evaluating trends from outside the inner world of academic or professional orthodoxies. According to established metrics, the market is currently overvalued. The ratio of total market capitalization to gross domestic product, which Warren Buffett suggested was the best indicator of market misevaluation, has recently hit all-time highs. Using a modified version (the Wilshire 5000 divided by gross national product), the value as of May 2025 is almost double its historical average.1 Another metric, Professor Robert Schiller’s cyclically-adjusted price to earnings ratio of the S&P 500, currently matches record high values previously set by Dot-com Bubble and the 2021 COVID response.2 Although counter-inflationary policies contributed to the inverted yield curve across 2023 and 2024, which is the occurrence of short-term treasury bonds having higher yields than long-term ones, it has been a reliable predictor of future recessions since the 1970s.3
However, many people are still optimistic about the short-term growth of the stock market, and the tech sector is at the center of this belief. The so-called Magnificent Seven companies (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla) have accounted for 40% of the S&P 500’s growth in April and May of 2025, with artificial intelligence models being a large explanation for the growth.4 Goldman Sachs chief equity research argued in 2024 that comparisons with the Dot-com Bubble are unjustified due to the higher earnings and proven organizational capacity of modern tech giants.5 As of May 2025, these companies have risen to 35% of the S&P 500’s capitalization, a far cry from the youthful startups which historically defined the industry.6 I wanted to look at the tech sector as a whole to see if it had an evaluation issue that might help explain the recent surge in Magnificent Seven stocks.
Using industry-wide data compiled by Professor Aswath Damodaran at NYU, I compared trailing PE and forward PE ratios for average firms across the same time period. For example, I calculated the difference between trailing PE from 2019 and forward PE from 2018. Trailing PE represents observed PE from previous quarters, whereas forward PE is made based on the predictions of analysts, who might be too optimistic given the current economic climate. A large positive ratio would entail overvaluation, either from underwhelming earnings or overwhelming prices. The sectors which I included in tech were: electronics (general), electronics (consumer & office), computer services, computers/peripherals, healthcare information and technology, software (entertainment), software (internet), software (system & application), semiconductor, semiconductor equipment, telecommunications equipment, and information services.
For the first graph, I created a forecast ratio (e.g. (2019 trailing – 2018 forward) / 2018 forward). For the second graph, I used a simple difference ratio (e.g. (2019 trailing – 2018 forward) / 100). The large positive values suggest an underperformance from either lackluster earnings per share growth or a high price increase that exceeds earnings growth. The difference between the market en masse and tech firms is conspicuous. Outside of the COVID pandemic’s peak from 2020 to 2022, when many tech sectors benefited from the remote transition, tech evaluations tend to be overvalued relative to the market. Concerningly, the data from 2024 shows the strongest underperformance yet.
At first I was surprised at the consistently large underestimation of the PE future ratios, which indicates a likely overvaluation of the firms. After contacting the professor, I learned the explanations. Yearly economic growth led to increased earning projections across all firms. Importantly, not all firms could have their forward PE calculated due to a lack of information, which would self-select more prosperous firms. This almost certainly causes variation in accuracy among different sectors, with tech sector forward earnings likely to be relatively understated due to absence of startups and recently public companies. This limits the accuracy of analysis at an absolute level, but will still be informative when comparing year-to-year variations.
Because the PE ratio has two main variables, I was uncertain if higher predicted earnings rates relative to the aggregate market was the main contributor to the mismatch. While short-term projected earnings per share data would have been ideal, the dataset contained 5 year projections, which I used as a proxy of general investor sentiment. I shifted the values to use predictions from the end of the preceding year, as they would be calculated in the same investment climate as the forward PE ratios, being more explanatory for potential overvaluation.
This graph tells a story: optimism about tech’s long-term growth prospects has consistently caused an overvaluation of its short-term stock, with an exception during the COVID pandemic. A positive slope, indicating an expectation of growth corresponding with overvaluation at the year’s end, can be seen in 7 years, with 5 of them being substantially steep. Consistently high overvaluations suggest that the sector failed to reach long-term expectations.
Interestingly, swapping the forecast error ratio for a simple difference makes for a noticeable difference. The graph does not necessarily show persistent overvaluation, but it often is overvalued due its incredibly wide variance. When you exclude the pandemic data points, it suggests a consistent overvaluation similar to the previous graph.
While long-term sentiment may give some clues into short-term consequences, short-term fluctuations also shift long-term outlook. For this inquiry, I used the 5 years earnings per share predictions at the end of the year, rather than the start. There is a noticeable negative slope for both tech and the market as a whole, which makes total sense. When people especially overvalued the market in a given year, they would revise their long-term predictions to be slightly less optimistic. While tangential to the question of tech overvaluation, it is noteworthy.
The all-important question is how much were the COVID era tech undervaluations a result of circumstances peculiar to the crisis, or a sustainable breakthrough. It is too early to definitively tell, but the data from 2023 and 2024 suggests a return to pre-COVID trends of overvaluations. While the economic benefits reaped from AI will exist, the extent for companies to profit off of it is a total mystery. Given the historical trend, I am worried that these current developments are viewed with rose-tinted glasses. In 2015, one could reasonably say that drone delivery and augmented reality would be aspects of daily life in 2025. Since nobody knows the future, historical patterns remain our best guide.
It is worth noting that the drivers of recent growth are not synonymous with the entire tech sector. The Magnificent Seven are all unique companies, and companies like Microsoft and Apple have shown success throughout decades of technological changes. I hope that my pessimism is proven wrong, that people’s savings and pensions are not damaged. However, blind optimism can cause more pain than sober reasoning. There are some concerning valuations of Magnificent Seven companies that people want to believe because of the supposed exceptionalism of our time.
I agree with Nivida CEO’s statement that “you’re not going to lose your job to an AI, but you’re going to lose your job to someone who uses AI.”7 I would have been unable to create my website without it, yet it could not create the website without myself. Large language models are not the first new technology which has disrupted the market and they will not be the last. Having reasonable expectations about its implementation in business is essential: AI implementation may take years even with existing models, the tasks which are appropriate to automate may be less than expected, labor cost saving may be limited as LLMs could function more as a tool than a replacement for most workers, potential security and legal risks could restrict its implementation scope, and new models may plateau in their performance or capacity.
Sources:
Used Prof. Damodaran’s dataset PE Ratios, PEG Ratios, and Expected Growth Rates by Sector for graphs (https://pages.stern.nyu.edu/~adamodar/)
1. https://www.gurufocus.com/economic_indicators/60/buffett-indicator
2. https://www.multpl.com/shiller-pe
3. https://fred.stlouisfed.org/series/T10Y2Y
4. https://www.reuters.com/business/finance/investors-see-us-stocks-rally-broadening-even-magnificent-seven-rebound-2025-05-30/
5. https://www.goldmansachs.com/insights/articles/ai-stocks-arent-in-a-bubble
6. https://www.nasdaq.com/articles/35-sp-500-concentrated-magnificent-seven
7. https://timesofindia.indiatimes.com/technology/tech-news/nvidia-ceo-jensen-huang-issues-urgent-ai-warning
© Peter Derrah 2025. All rights reserved.