Only AI that makes money is good AI.

The "fireworks" ignited by the emergence of ChatGPT in the US stock market have been burning for half a year. AI has directly pushed the stock market into a bull market during the interest rate hike year. Amidst the lament of the capital winter, AI remains a beautiful sight.

However, after half a year, even Aunt Zhang, who runs a small grocery store at the village entrance, has learned to use ChatGPT to write promotional texts. Everyone is already tired of the battle of large models. Looking back, the only ones making money are NVIDIA - after half a year of gold panning, they haven't found any gold, but they have accumulated quite a few shovels.

The second half of the bubble squeeze has quietly begun.

AI Hype Subsiding

After the US stock market closed on Wednesday, C3.ai, a star AI concept stock whose stock price has risen by more than 200% this year and has a PB ratio of 3.5 times, announced its earnings report. The loss guidance has been expanded from $50-70 million to $70-100 million, and the expected time for profitability has been further delayed from the end of 2024. The company's CEO also stated at the performance meeting that they plan to further expand investment.

On Thursday's opening, C3.ai opened more than 11% lower and closed down more than 12%.

Morgan Stanley analyst Pinjalim Bora further expressed bearish sentiment towards the stock in a report on Thursday, believing that C3.ai's top-line indicator (revenue) has not substantially improved, and increasing investment in AIGC does not make sense.

Affected by C3.ai, other AI concept stocks such as BigBear.ai and SoundHound.ai also experienced significant declines overnight.

Before this round of sharp decline, the momentum for the AI sector to further rise has been insufficient.

Due to limited revenue growth driven by AI, Microsoft, the main supporter of OpenAI, has seen its stock price fall by 7.4% since mid-July, and social media platform Snapchat has suffered from profit squeeze due to AI investments. Bank of America cited EPFR data, pointing out that in the week ending September 6, technology stocks saw a net outflow of $1.7 billion, facing the first sell-off in 11 weeks, which may indicate that the AI hype is subsiding.

As the industry trend becomes increasingly clear, investors are becoming less patient. Citigroup analyst Scott Chronert told the media that the AI hype has now entered the "show me the results" stage.

"NVIDIA's Earnings Fraud" Conspiracy Theory Spreads, Long-Short Battle Deadlocked

Even NVIDIA, the biggest winner of this round of AI hype and a member of the trillion-dollar market value club, has recently encountered "short essays" spreading bearish sentiment, causing its stock price to decline for two consecutive days, wiping out all the gains since the second-quarter report was announced after market close on August 23.

Widely circulated conspiracy theories on X and Reddit believe that Coreweave, an AI startup in which NVIDIA has invested, helped inflate second-quarter revenue. Some even suspect that NVIDIA is involved in earnings fraud. Coreweave, which has placed billions of dollars in orders with NVIDIA, may not even be a real company, but rather a shell organization created and supported by BlackRock, NVIDIA's largest shareholder, in collaboration with NVIDIA.

There are also claims that NVIDIA is colluding with "big players" such as Microsoft, Amazon AWS, and the third-ranked large-scale model startup Inflection. The main industry players have placed a significant number of orders with NVIDIA, and these orders are problematic, designed to boost NVIDIA's revenue.

According to research firm Macrotips Trading, although NVIDIA's Q2 report far exceeded expectations and provided strong guidance for future revenue, there are reasons to be concerned about "suspicious disclosures and management practices." In particular, the involvement of an AI startup directly supported by NVIDIA in driving revenue growth, peculiar timing of stock buybacks, and insider selling have raised red flags.

Market analysis suggests that behind the spread of conspiracy theories, it is evident that the long and short game surrounding NVIDIA is highly contentious, with the underlying concern being the fear that if the AI bubble bursts, it will drag down the entire market.

It is clear that due to the lack of downstream performance validation, some impatient investors have begun to doubt whether AI is truly a revolution.

Making Money from Prospecting, Sustaining the Market

This anxiety is understandable, as compared to AI infrastructure vendors who have already benefited from the enormous demand generated by model training, with orders and performance being continuously validated, B-side applications are still in the early stages. Most AI application vendors have yet to enter the commercialization phase, and based on the time to monetization, it is expected to be 2-3 quarters later than the infrastructure layer.

According to CBInsights data, as of Q2 2023, the investment and financing in generative AI has grown 4.6 times compared to the total of $2.5 billion in the previous year. However, generative AI application layer financing only accounts for 30%: currently, about 70% of the funds are invested in AI infrastructure, including large-scale model development.

However, according to Gartner's technology maturity curve, new technologies typically go through a typical development cycle of rise, hype, decline, and resurgence. Generative AI may now be entering the decline phase of the bubble.

But the validation of downstream application performance may not be far off.

For example, CRM software provider Salesforce has jumped on the AI bandwagon. The company integrated AI into its office IM software Slack and other products last quarter, and introduced a dedicated AI chatbot called Einstein GPT for salespeople, customer support agents, and marketers. They also raised product prices and exceeded market expectations in terms of revenue, profit, and earnings per share.

In addition, Salesforce has also raised its guidance for Q3 and annual revenue and operating profit margins, and stated at the performance meeting that they will "lead our customers into the new era of artificial intelligence." Investment bank Raymond James analyst Brian Peterson pointed out:

"The continuous upward revision of the company's growth prospects for the second quarter reflects the management's increasing confidence in sustainable double-digit growth. Salesforce has experienced price increases and has begun to monetize its AI capabilities."

In addition to Salesforce, another leading enterprise in the application layer, Adobe, will also announce its performance next week. Analysts expect its quarterly earnings per share to reach $3.97. According to the guidance provided by the company, the earnings per share guidance range for the third quarter is $3.95-4.00, and the earnings per share guidance range for the 23 fiscal year is $15.65-15.75, indicating Adobe's strong confidence in future performance prospects.

Huajin Securities pointed out in a recent research report:

"Leading companies in some tracks, including MS, Salesforce, ServiceNow, and Shopify, have released generative AI products. Among them, MS and Salesforce have announced the pricing of AI products, and it is expected that the initial commercialization of B-side applications will be seen from Q4 of this year to Q1 of next year. In addition, some companies such as Palantir and PaloAlto Networks have already applied generative AI products in practical scenarios and have made significant revenue contributions."

Therefore, although this wave of generative AI still has many inherent problems and deficiencies, its long-term technological potential and industrial space have been widely recognized by the market. The widespread outbreak of downstream applications may come in the next few quarters, and the long-term investment opportunities in the industry chain are beyond doubt.

A market where only NVIDIA can make money cannot be considered a good market. Only gold diggers can achieve real returns. Only when the entire industry ecosystem can sustainably continue can it be stable.