In fact, $NVIDIA (NVDA.US)$ The second quarter performance, which was released earlier, exceeded analyst expectations, and the performance outlook for this quarter also exceeded analyst expectations. However, it may not have met the increasingly high expectations of investors. Last Thursday, the company's stock price fell more than 6% on the second day after the earnings release. Since the company announced earnings, the stock has fallen 14% in the past three trading days.

Nvidia's volatility fully reflects investors' increasing caution towards the emerging artificial intelligence technology that has driven the stock market to a sharp rise this year. While Nvidia is in turmoil, the entire Nasdaq and U.S. semiconductor stocks are also ready for action.

Dragged down by Nvidia, U.S. chip stocks have plummeted across the board.

During the same period of Nvidia's sharp drop, $PHLX Semiconductor Index (.SOX.US)$It also fell 7.75%, marking the largest single-day decline since 2020. Among them, the stock prices of all 30 companies in the Philadelphia Semiconductor Index fell by at least 5.4%, among which $ON Semiconductor (ON.US)$,$KLA Corp (KLAC.US)$and $Monolithic Power Systems (MPWR.US)$fell more than 9%. $NASDAQ 100 Index (.NDX.US)$ Fell by nearly 3.2%. Intel fell by nearly 8%, Marvell fell by 8.2%,$Broadcom (AVGO.US)$Lost about 6%. AMD fell by 7.8%,$Qualcomm (QCOM.US)$Dropped nearly 7%. The index tracking semiconductor stocks$VanEck Semiconductor ETF (SMH.US)$ also fell by 7.5%, the largest single-day drop since March 2020.

Over the past year, chip stocks have been rising as people optimistically anticipate increased demand for semiconductors and memory from enterprises in order to meet the growing computational needs of artificial intelligence. Leading in this sector is Nvidia, which dominates the AI data center chip market. By 2024, the stock is projected to rise by 118%.

Other chip companies are also joining this growth trend. Intel and AMD sell AI chips, but so far the market penetration is limited. Broadcom is working on Google's TPU chip, and Qualcomm is promoting its chip as the best chip for running Android phone AI.

Last week, Nvidia announced quarterly revenue of $30 billion for the month of July, exceeding Wall Street's already high expectations. The company's data center business (including AI processors) saw a YoY revenue growth of 154%, partly due to a few cloud computing and internet giants purchasing billions of dollars worth of Nvidia chips every quarter.

Nvidia expects sales to grow by 80% this quarter. Some investors found Nvidia's forecast disappointing, which temporarily impacted chip manufacturers supplying memory and other components to the company. According to the latest data, foreign and institutional investors sold 4 trillion South Korean won ($3 billion) worth of Samsung Electronics and SK Hynix stocks in the past month.

According to data from the Korea Exchange on September 1st, foreign institutions sold 2.088 trillion Korean won worth of Samsung Electronics stocks in August. Domestic institutions sold 1.378 trillion Korean won worth of Samsung stocks.

Analysts attribute these sales to Nvidia's recent financial report. Although the tech giant's revenue more than doubled in the previous quarter, investors were disappointed because the growth was the smallest compared to the past six quarters. Nvidia's expectations for the third quarter also show a smaller YoY growth, with expected revenue of $32.5 billion.

Analysts explained that Nvidia's lack of specific details about its next-generation AI chips further eroded investor confidence. Nvidia announced that the Blackwell processor it manufactures will begin mass production in the fourth quarter, potentially bringing in billions of dollars in sales, but did not disclose a specific release schedule. The industry had previously expected Blackwell to start shipping in the third quarter of this year.$Taiwan Semiconductor (TSM.US)$The Blackwell processor manufactured by Nvidia is expected to begin mass production in the fourth quarter, which could generate billions of dollars in sales, but the specific release timetable has not been revealed yet. The industry previously anticipated that Blackwell would start shipping in the third quarter of this year.

Investors are also concerned that Nvidia's future demand for high-bandwidth memory (HBM) may decline, which could negatively impact SK Hynix, as SK Hynix currently supplies most of the HBM chips used by Nvidia to produce graphics processing units (GPUs).

On Tuesday, the August data released by the ISM Manufacturing Index was lower than widely expected, sparking concerns about economic strength, but also possibly increasing the likelihood of a Fed rate cut, further leading to a subdued market performance.

Antitrust investigations are becoming increasingly stringent, aggravating concerns.

While facing performance below expectations, Nvidia is also increasingly facing strict antitrust investigations, further highlighting the future uncertainty of the company's stocks.

Thanks to the investments in GPUs and CUDA in recent years, Nvidia has built a strong ecosystem, making it a huge moat that competitors find difficult to surpass. Nvidia also occupies a dominant position in the AI chip market – estimated to hold 70% to 95% of the market share in the AI training chip market – indicating that its products are crucial to this rapidly growing industry. In addition, Nvidia has strong pricing power, reflected in its impressive gross margin of about 78%.

It is this tying of software and hardware, and the method of selling systems, that has put Nvidia under antitrust investigation.

In July this year, insiders revealed that Nvidia is likely to face charges of anti-competitive behavior from the French antitrust regulator, making it the first enforcement agency to take action against the chip manufacturer. Back in September of last year, France had carried out a surprise inspection of the graphics card industry, and sources said that the surprise inspection of Nvidia was the so-called statement of objections or accusation by France. This surprise inspection is the result of a broader investigation into cloud computing.

Reports have indicated that the French regulator issued partial dissenting statements to companies, but not all. Nvidia declined to comment. The company stated in last year's regulatory filings that regulatory authorities in the EU, China, and France had requested information about its graphics cards. Other sources said that since French authorities are investigating Nvidia, the European Commission is currently unlikely to expand its preliminary review.

The French regulatory authority highlighted the risks of chip suppliers' abuse in a report on AI competition published at the end of June. It expressed concern about the industry's dependence on Nvidia's CUDA chip programming software, which is the only system that is 100% compatible with GPUs essential for accelerated computing. It also mentioned concerns about Nvidia's investments in AI-focused cloud computing service providers such as CoreWeave.

If found to be in violation of French antitrust regulations, companies could face fines of up to 10% of their global annual revenue, but they can also make concessions to avoid penalties.

Meanwhile, the United States is also intensifying its antitrust investigation into Nvidia.

The latest report from Bloomberg shows that the US Department of Justice has issued subpoenas to Nvidia and other companies in search of evidence of this chip manufacturer's violation of antitrust laws, escalating the investigation into this major supplier of AI processors. Informed sources revealed that the Department of Justice had previously issued investigative questionnaires to the company and is now sending legally binding requests for information, bringing the government one step closer to filing formal complaints.

According to informed sources, antitrust officials are concerned that Nvidia makes it more difficult for businesses to switch to other suppliers and penalizes buyers who do not exclusively use their AI chips. Following Bloomberg's report on the subpoenas, Nvidia's stock price fell further in the afternoon trading session.

In response to questions about the investigation, Nvidia stated that its market dominance is derived from the quality of its products, which have faster performance. The company, in an email statement, said, "Nvidia prevails on its own merits, which is reflected in our benchmark results and the value we provide to customers, who can choose the solution that best suits them."

Nvidia, founded in 1993, gained fame by selling graphics cards to computer gamers. However, its chip manufacturing methods eventually proved useful for building AI models, a process that involves bombarding software with data. The company also quickly expanded its product line to include a range of software, servers, networking, and services—all aimed at accelerating AI deployment from Nvidia's perspective.

The success of its products and competitors' struggles to launch alternative chips have made Nvidia a critical player in the supply chains of some of the world's largest companies. For example,$Microsoft (MSFT.US)$Company and $Meta Platforms (META.US)$More than 40% of its hardware budget has been spent on equipment from this chip manufacturer. During the peak shortage of Nvidia's H100 accelerator, the retail price of a single component reached up to $90,000.

Since becoming the world's most valuable chip manufacturer and a major beneficiary of the AI spending boom, Nvidia has been under regulatory scrutiny. Sales have more than doubled every quarter, surpassing former chip leaders like Intel. As mentioned above, Nvidia's practices have raised broader regulatory concerns.

Is there a bubble in generative AI? When will it burst?

As generative AI evolves, industry analysts are starting to openly question whether the massive investments in GenAI will bring returns, leading to doubts about the discussion on generative AI. In a letter, Goldman Sachs' research department stated that the lack of "killer applications" beyond writing co-pilots and chatbots is the most urgent issue, with data availability, chip shortages, and power issues also posing resistance.

Since OpenAI introduced a new large language model (LLM) called ChatGPT to the world at the end of 2022, the hype around neural network AI, especially GenAI based on transformer networks, has been eerily similar to previous tech booms.

It is worth noting that in the long history of technological development, some of these major moments have ultimately become true turning points, such as mobile and cloud computing, some make us ask ourselves, "What are we thinking?" (blockchain, 5G), while the full experience and lessons learned from other technological breakthroughs take years to emerge (the internet bubble, and even Hadoop-style computing).

Therefore, the biggest issue we face now is: five years from now, which category will we assign GenAI to? One of the people who believe that AI might follow the same path as 5G and blockchain is Goldman Sachs. In the June edition of Goldman Sachs Research Communiqué, an extensively read report titled "Gen AI: Too Much Investment, Too Little Return?" explores the question of whether AI will succeed.

She writes, "Generative artificial intelligence technology has the potential to change companies, industries, and society, and this prospect continues to be hyped, leading tech giants, other companies, and utility companies to spend around $1 trillion in capital expenditure over the next few years, including significant investments in data centers, chips, other AI infrastructure, and the power grid." "But so far, apart from reports of improved developer efficiency, there is little to show for these expenditures."

Daron Acemoglu, a professor at MIT, said, "Generative artificial intelligence has the potential to fundamentally change processes like scientific discovery, research and development, innovation, and testing of new products and materials, and create new products and platforms. But given the current focus and architecture of generative AI technology, these truly transformative changes are not likely to happen quickly, and they occur infrequently over the next 10 years (if at all)."

"Attempting to accelerate the progress of GenAI by increasing the production of the two core elements, data and GPUs, may not be feasible, as data quality is a significant part of it," Acemoglu bluntly stated.

The shortage of chips suitable for training GenAI models is another factor that contributes to Goldman Sachs' pessimistic (some would say realistic) view of GenAI.

Jim Covello, Global Head of Equity Research at Goldman Sachs, writes in the report, "Today, Nvidia is the only company capable of producing the GPUs necessary for AI. Some believe that the semiconductor industry or mega-scale companies (such as Google,$Amazon (AMZN.US)$And Microsoft) itself will become a competitor to Nvidia, which is possible. But considering that chip companies have been trying to overthrow Nvidia's dominant position in the GPU field for the past 10 years, but have always ended in failure, this is a huge leap compared to today.

Kovello said that the huge cost of training and using GenAI will be the resistance to GenAI's ultimate productivity or efficiency improvement.

Finally, the electricity required to train LLM and other GenAI models must also be included in the formula, which will also be a hindering factor. It is estimated that AI currently consumes about 0.5% of the world's energy, and this number is expected to increase in the future.

However, some people are still bullish on GenAI's long-term prospects for business and society, and of course, the future of Nvidia. Some analysts believe that Nvidia is expected to exceed a market cap of 10 trillion dollars.

Nvidia, a promising future?

JPMorgan analyst raised the target price of Nvidia's stock by nearly 35%, stating that the strong demand for the company's Grace Hopper series of chips offset the delay of its next-generation Blackwell system by 2 months. Analysts from Jefferies and Wedbush also noted that the report indicates a persistently strong demand for Nvidia chips.

Bank of America analysts pointed out that the increased cost of launching the Blackwell system and its impact on profit margins are potential obstacles faced by the company. However, analysts maintained a 'buy' rating and raised the target price, advising investors to 'ignore quarterly noise' and focus on Nvidia's dominant position in the generative AI chip market.

Despite recent pullbacks, Nvidia remains the top-performing stock in the S&P 500 index this year, with its stock price rising more than 120% since the beginning of 2024.

Todd Sohn, the ETF strategist at Strategas Securities, said: "In the past 12 months, a large amount of funds have flowed into the technology and semiconductor industries, leading to completely imbalanced trading." This has also driven the chip index to rise 14% year-to-date, slightly lower than the S&P 500's increase of 16%.

However, in a client report on Tuesday, a strategist at BlackRock wrote: "Recent research questions whether income from artificial intelligence alone can ultimately justify this wave of capital expenditure. When evaluating individual companies' AI capital expenditure, investors must consider whether they are fully utilizing their balance sheets and capital."

From a global perspective, there is no doubt that the prospects of GenAI still look very promising, even if the outcome may not be realized. But the biggest current issue is whether the returns of GenAI will increase before time runs out, as time is ticking away. Additionally, the emergence of low-cost Nvidia chip alternatives will also be a key factor affecting the future of GenAI and Nvidia.

As mentioned in Fortune: "While Nvidia's profit margin has been at historical highs, there is evidence that this level of profitability may not be sustainable. Over time, competition could weaken its pricing power, leading to a steady decline in gross margin and net profit margin. Nvidia's net profit margin is well above 50%, which is impressive, but it may fall back to the range of 40% to 45% seen in more historical records. Although this decline could harm the stock price in the long run, such deflation is unlikely to occur sharply."

Editor/Emily