Looking at performance, which companies are truly pseudo-AI?

In the year and a half since the emergence of GPT, the entire U.S. stock market has revolved around the bull market centered on Ai. Nvidia is undoubtedly the leader of U.S. stocks and also the company with the most significant increase in market value.

Although many people are wary of Nvidia's performance, believing that such short-term gains are unrealistic, speculative, and at risk of collapse at any time. However, the closer people track Nvidia, the more they realize that the company's valuation is not unreasonable. It is a stock truly driven by performance. Even if the Ai bubble bursts, Nvidia will not be the hardest-hit company because it is indeed a technological monopolist and the company that has made the most substantial gains in this wave of enthusiasm.

The danger lies with companies that are not closely related to Ai but have been following the Nasdaq's upward trend. There are many such companies in A-shares, and of course, many of them have returned to their original state this year. The same goes for many U.S. stocks, with some already reverting to their original state, but many are still soaring. Perhaps it is necessary to carefully examine their performance to determine which companies are truly pseudo Ai.

I. Hardware performance growth is substantial

The Ai craze began in early 2023, so in terms of company valuation, comparing the current cumulative increase with the expected revenue and profit growth can help understand the reasonableness of the increase.

For example, Nvidia has risen by 577% from 2023 to the present, which is quite an astonishing figure. In 2022, the revenue was $14.8 billion, with a profit of $4.3 billion.

Taking the furthest performance forecast given by Nvidia for the latest fiscal year, the revenue expectation for the first half of 2025 is $54 billion, annualized to $114 billion, an increase of 323% compared to 2022's $26.9 billion. Due to the popularity of Ai chips, Nvidia's current profit margin is much higher than in 2022, reaching nearly $14.8 billion in one quarter, resulting in a profit increase of over 10 times.

Therefore, from a PE perspective, Nvidia is currently lower than the end of 2022, and from a PS perspective, it is slightly higher. However, if Nvidia continues to fall, it will indeed reach a new low in valuation in recent years.

It is evident that Nvidia's performance growth is better than its stock price growth. Concerns about the low returns on Ai applications and the delayed shipment of Nvidia's new chips have recently caused Nvidia to retreat by 30%, making its valuation appear more reasonable. If it falls further, Nvidia will reach its lowest PE in five years. Therefore, regardless, a significant portion of Nvidia's multi-year gains will ultimately be value appreciation that will not retract As a comparison, AMD is the second largest competitor in AI chips, and has been pushing for the development of new computing chips in recent years in hopes of gaining some market share from NVIDIA. Almost all downstream players in the industry chain are willing to support AMD, but the fact is that AMD's stock price has risen by 100%, but its revenue growth is actually very poor. After the guidance in the second half of 2024, the revenue is only 25.2 billion, a 6% increase compared to 2022, with a decrease in profit margin. From this growth rate, it can be seen that AMD has been playing a pseudo-AI role in recent years. A 6% revenue growth has resulted in a doubling of the stock price.

If we evaluate the trend of AMD compared to NVIDIA in the past two years, it can be said that AMD follows the decline rather than the rise. Each round of pulse rise is smaller, and the retracement amplitude is larger. Currently, AMD's high point has fallen by nearly 44%. AMD has repeatedly hurt the fear-of-heights NVIDIA, hoping to attract investors who underestimated the Dragon 2.

Looking at the various companies in the NVIDIA chain, including servers, optical modules, and chips, many of them have seen frightening increases. However, there is no major controversy in terms of performance.

Seeing SMCI, as the number one little brother who has risen more fiercely than NVIDIA in this AI bull market.

Although this company's business model is not ideal, it is just a data center assembler with significant competition and has never had a good profit margin. However, the performance trend is real.

Over the past 2 years, Super Micro Computer Inc. has accumulated a 521% increase. Of course, the absolute increase at the peak is higher than NVIDIA's, but the recent sharp drop in stock price due to the significant decline in profit margin in the financial report. However, it is also seen that the company has provided a revenue guidance for the fiscal year 2025 of 26-30 billion US dollars, with a midpoint of 28 billion. In 2022, the revenue was 5.196 billion, a 338% increase. Comparing the stock price increase and revenue increase, the difference is still acceptable.

SMCI's profit margin estimate for 2025 is slightly lower than 2024, but it is also at 8%, which translates to a profit of 2.2 billion, an increase of 785% compared to 285 million in 2022

Based on the ratio of these two data, SMCI has gone through a round of adjustments, and the current stock price increase has been digested. It can't be considered pseudo-AI at all, because the orders are in a backlog state, and profits are somewhat predictable. The probability of a significant error in performance over 25 years is not high.

As a company that sells shovels for computing power, the only point of doubt is the unstable demand. It is possible that after everyone stops investing in AI, orders may plummet. However, this is a demand issue, and the supply side has not made any mistakes. Similar to cyclical stocks and vaccine stocks, their earnings should not be questioned.

However, as seen from AMD, there are also many pseudo-AI companies in the hardware sector. They mostly leverage the concept of AI to drive significant stock price increases based on the expectation of benefiting from existing profits. Representative stocks with high increases include AAOI, ANET, VRT, etc. In the semiconductor sector, there are companies like Broadcom, TSMC, and Applied Materials. Their performance does not match the stock price increase at all.

Which companies will truly benefit from the chain effect, and which are fake, will naturally differentiate. For example, TSMC, which dominates the competitive landscape, will not find it too difficult to benefit from NVIDIA's profits. The expectations for Broadcom are similar to those for AMD, and there is a chance they may repeat the same mistakes.

However, there are a few things to note. Firstly, the US stock market has experienced a significant decline, and the valuations of companies like NVIDIA that have already made profits have adjusted accordingly. Companies with speculative valuations currently lack this adjustment because their valuations will only rationalize based on the performance of the next few quarters.

Secondly, they all have the characteristic of profit delay and subsequent release, which is lagging. If companies like NVIDIA continue to rise with favorable performance, those lagging behind will be fine. But if faced with a collapse in the AI industry and a downturn in NVIDIA's demand, while these companies are still experiencing rapid growth in performance, should we look at NVIDIA's leading indicators or their financial reports? Their stock prices and valuations are waiting for performance releases to catalyze, and in such a scenario, even positive performance may not boost stock prices, which is a drawback.

Therefore, from the performance of the hardware industry, it makes sense for leaders to become leaders. The logic of waiting for profit overflow sharing is not as good. Once they do not progress as expected, they become pseudo-AI, following the decline instead of the rise. Even if you are bottom fishing in AI now and believe that the prosperity will continue, you must understand that there are differences in logic, and choosing the good ones over the bad ones is crucial.

II. Uncertain Profit Trends in Software

The discussion about the AI bubble is already widespread, with the most discussed issue being the mismatch between investment and output. With relevant expenditure data, assuming NVIDIA currently has an annualized revenue of 90 billion, if the investment in computing facilities and other equipment is half of the computing power, then it would require the construction of 180 billion data centers. Calculated based on a 50% profit margin in the software industry, AI would need to drive 360 billion in revenue to be justifiable

However, the current situation is that the revenue growth of the software industry is far from reaching this estimate.

Of course, the 360 billion may not necessarily be an increase in revenue. If Ai saves half of the manpower and maintains the same efficiency, its impact on enterprises is mainly cost-related. By halving labor costs, this part of the profit is released.

For global software companies, after a large investment in Ai, releasing an additional profit increment of 180 billion through profit margin optimization each year is an acceptable figure. However, this excludes the profit growth generated by the development of the enterprise itself.

Currently, the profit improvement of global software application companies has not reached this level.

Microsoft, the largest software company in the world, which holds openai, has shown simultaneous growth in revenue and profit margin from a financial perspective. Specifically, Microsoft's profit in the 2024 fiscal year increased by 21% compared to the 2022 fiscal year, with an increase of around 15.4 billion, and revenue accumulated by 23%. However, the forecasted revenue for the next quarter is flat, which has lowered the revenue growth rate expectation for 2025 to around 15%. So even if the revenue growth rate for 2025 is calculated optimistically, Microsoft's revenue growth over these three years is only about 41%.

While Microsoft's net profit margin is declining, there is reason to believe that the cumulative profit growth rate until 2025 will not exceed 41%.

Currently, Microsoft's stock price has also increased by 70% in two years, and ai pc and copilot have been launched for a long time, theoretically should drive some growth. However, the growth rate slope has not changed much. The growth rates of the three years are not as good as the stock price increase of the two years, indicating that Microsoft is also a company whose performance does not match its stock price.

Google and Meta, as competitors in the arms race with Microsoft's large models, have lower product integration but higher growth. However, in terms of performance, they have indeed outperformed Microsoft slightly. Google and Meta's revenue increased by 37% and 33% respectively over the past two years, but Meta's profit was very low in 2022 due to excessive investment in the metaverse two years ago, with profits rebounding by nearly 160% in the past two years. However, both companies have not seen a significant increase in stock price, whether in terms of revenue or profit.

Of course, the performance of the software industry benefiting from AI is definitely slower than that of hardware, so everyone has more optimistic expectations for the future. However, it should also be noted that Microsoft, Google, and Meta's latest fiscal year profit margins have not improved significantly compared to 2022. Whether it is capitalization or expense, computing expenses will increase costs. The growth in enterprise profits is influenced by business adjustments (such as Meta's withdrawal from the VR business) and follows natural economic growth. Even without AI, growth is reasonable. The profit release brought about by cost optimization by AI should theoretically improve profit margins, but this is not the case now. This makes the goal of generating an additional profit of 160 billion more distant.

Moreover, Microsoft is the company that integrates AI applications the fastest, but it also has the fastest decline in profit margin performance and revenue growth rate. Meta and Google are still exploring and integrating, and their excellent performance currently comes from low base numbers and the superiority of their own business models, not so much from AI. So, is the result of continuing AI transformation sustained growth, rather than becoming like Microsoft? This is a question worth pondering.

For other software service companies, they can be considered pseudo-AI or not part of the AI sector, because the prosperity of AI began with GPT and Vision, and most software services do not have much logic for business growth in this area, perhaps only cloud services have gained some prosperity following Microsoft's AI infrastructure. But looking at their performance, most of them have seen greater growth than revenue growth. From CRM, SAP to NOW, PLTR. Valuations have basically soared.

Saying that AI will drive software prosperity, most companies have seen a decrease in growth rates after 2022, and AI has not changed the downward trend in the industry's growth rate. Occasionally, there are improvements in performance growth rates, such as Palantir and ServiceNow, which have surged, but looking back, with a PS of over 10 and a 30% revenue growth rate, the entire software sector is now stuck at a 30% growth rate. It is worrying.

In such an environment, a bunch of software service companies with declining growth rates have already plummeted. For example, star stocks like Cloudflare and Snowflake have unknowingly become one of the worst-performing sectors in the U.S. stock market, and have already been expelled from the AI sector. This is the fate of pseudo-AI III. Conclusion

The development of AI will greatly promote productivity, but without quantitative analysis, it is just empty talk. If it does not drive performance growth and profit improvement, or if investing 100 billion only yields 10 billion in output, then it is all in vain. From the current performance and stock prices, the market has expelled some AI stocks, but in the hardware and software sectors, there are still many individual stocks with significant bubbles due to the concept of AI.

Regardless of whether the economy continues to thrive or doubts increase, there are several logics that need attention. The logic of delayed benefit spillover itself carries greater risks. Secondly, for some leading internet companies, AI may not necessarily promote future performance. Perhaps when the market can rationally view AI, not considering it solely as a positive business promotion factor, but rather as a factor of coexistence of good and bad, that would be the correct attitude. It is clearly a mistake to make the degree of AI integration or AI-related performance the first evaluation indicator of a company's value