Crypto+AI Web3's last hope?
The meme market seems to be facing a collapse, and the market's anxiety is spreading again. What's the reason? The lack of new narratives, the investors are smart, and all went to blue chip NFT.
JinseFinanceLog in/ Sign up
Cycle Capital: optimistic but neutral about the market in the rest of the year
Follow
Market Overview
Since the third quarter, the assets that performed better: RUT (Russell 2000 Index) and XAUUSD (gold price) as well as SPF (financial stocks) and US bonds, the assets that performed worse: Ethereum, crude oil, US dollars. The almost flat pie, Nasdaq 100.
For US stocks, the current market is still in a bull market and the main trend is still upward. However, the trading environment in the last few months of the year will lack performance themes, and the market's upside and downside space will be limited. The market continues to revise down Q3 earnings expectations:
The recent valuation correction, but the rebound was also very fast, 21 times the PE is still much higher than the 5-year average:
The performance of large technology stocks weakened in the middle of the summer, mainly due to lower earnings expectations and the market's waning enthusiasm for AI themes. However, the long-term growth potential of these stocks is still there, and the prices will not fall.
There was a time when the market was indeed very bullish, for example, the period from October last year to June this year experienced some of the best risk-adjusted returns in a generation (NDX sharpe ration reached 4). Today, the stock market has higher PE multiples, economic and financial growth expectations are slower, and the market's expectations of the Fed are higher, so it is relatively difficult to expect the stock market to perform as well as the previous three quarters in the future. And we have seen signs that large funds are gradually switching to defensive themes (for example, both subjective and passive strategy funds have increased their holdings in the healthcare sector, which provides defensiveness and growth unrelated to AI). This trend is not expected to reverse soon, so it is safer to have a neutral attitude towards the stock market in the coming months.
At the Jackson Hole meeting on Friday, Fed Chairman Powell made the most clear statement on interest rate cuts to date, with a September rate cut as a foregone conclusion. He also said that he did not want to further cool the labor market and that he was more confident that inflation would return to the 2% path. But he still insisted that the pace of policy easing will depend on future data performance.
So I personally think that Powell’s statement this time was not more dovish than expected, so it did not cause much splash in the traditional financial market. What everyone is most concerned about is whether there will be a chance for a single 50bp interest rate cut this year, but Powell did not hint at all. So the expectation for interest rate cuts within the year has hardly changed from before:
So if future economic data improves, the current expectation of a 100bp interest rate cut may even be adjusted downward.
However, the crypto market reacted very strongly. This may be due to the squeeze caused by too much accumulation of short positions (for example, the recent position has increased rapidly but the contracts often have negative rates), and the fact that the rhythm of currency circle players' understanding of macro news is not as unified as that of traditional markets, that is, the damping of message transmission is relatively large. Many people may not know that Powell will speak at the JH meeting this week. However, whether the current market environment supports the crypto market to hit new highs may be a question mark. Generally speaking, to hit new highs, the macro environment needs to be loose and sentiment needs to be risk-averse. Crypto-native themes are also essential, including NFT, defi, the liberalization of spot ETFs, and meme craze. At present, the only theme with a strong momentum is the growth of the Tele ecosystem. Whether it has the potential to become the next theme depends on the performance of the latest token issuance projects and the quality of the incremental users they bring.
At the same time, the jump in the crypto market is also related to the sharp downward revision of last year's non-farm payrolls in the United States this week. However, as we have analyzed in depth in the previous video, this downward revision is excessive, ignoring the contribution of illegal immigrants to employment, and these people were included in the original employment statistics, so this revision is of little significance. As a result, the traditional market reacted mediocrely, while the crypto market regarded this as a sign of a sharp interest rate cut.
From the experience of the gold market, most of the time the price is positively correlated with the holdings of ETFs, but the market structure has changed in the past two years. Most retail investors and even institutional investors have missed the rise in gold, and the main purchasing power has become the central bank:
From the figure below, you can see the big pie. The inflow rate of ETFs has slowed down significantly since April. In terms of bitcoin, it has only increased by 10% in the past five months, which is consistent with its price peaking in March. If the risk-free rate declines, it may attract more investors into the gold and bitcoin markets, which is very likely.
From the perspective of stock positions, subjective strategy funds did quite well earlier in the summer, reducing their positions in a timely manner and having the opportunity to attack in August. The figure below shows that the green line subjective strategy funds have recently covered their positions very quickly, and their positions have returned to the historical 91st percentile, but systematic strategy funds have reacted more slowly and are currently only at the 51st percentile:
Stock market shorts liquidated positions during decline:
Capital Flow
China's stock market has been falling, but funds with China concepts have been experiencing net inflows. This week's net inflow of US$4.9 billion hit a five-week high and marked the 12th consecutive week of net inflows. Compared with other emerging market countries, China also has the largest inflows. ="https://miro.medium.com/v2/resize:fit:1400/0*vBaE4qNtIc_lO6Or">
However, structurally, from the perspective of Goldman Sachs' clients, they have basically been reducing their holdings of A shares since February, and the most recent increases have been in H shares and Chinese concept stocks:
Despite the rebound in global stock markets and capital inflows, the money market with low risk appetite has also seen inflows for four consecutive weeks, with the total size rising to US$6.24 trillion, a new historical high. It can be seen that market liquidity is still very abundant:
Weaker US dollar
Over the past month, the US dollar index (DXY) has fallen 3.5%, the fastest rate of decline since the end of 2022, which is related to the market's increased expectations of the Fed's interest rate cuts.
Looking back at the beginning of 2022, the Federal Reserve adopted an aggressive interest rate hike policy to combat inflation, which pushed the dollar to strengthen. However, by October 2022, the market began to expect that the Fed's interest rate hike cycle was coming to an end and might even begin to consider cutting interest rates. This expectation led to a decline in market demand for the US dollar, pushing the dollar weaker.
Today's market seems to be a repeat of that year, except that the hype at the time was too advanced, and today the interest rate cut is about to land. If the dollar falls too much, the unwinding of long-term carry trades could resurface, becoming a force suppressing the stock market:
Two major themes next week, inflation and Nvidia
Key price data include the US PCE (personal consumption expenditure) inflation rate, Europe's August preliminary CPI (consumer price index), and Tokyo's CPI. Major economies will also release consumer confidence and activity indicators. In terms of corporate earnings, the focus will be on Nvidia's earnings report after the U.S. stock market closes on Wednesday.
Friday's PCE is the last PCE price data before the Fed's next decision on September 18. Economists expect core PCE inflation to remain at +0.2% month-on-month, with personal income and consumption increasing by +0.2% and +0.3% respectively, the same as in June, which means that the market expects inflation to maintain a moderate growth momentum and will not decline further, leaving room for possible downward surprises.
Nvidia's financial report preview - the dark clouds have dissipated, and it is expected to inject a shot in the arm for the market
Nvidia's performance is not only a barometer of AI and technology stocks, but also a barometer of the entire financial market sentiment. NV's demand is not a problem for the time being. The most critical topic is the impact of the postponement of the Blackwell architecture. After reading several institutional analysis reports, I found that the mainstream view on Wall Street is that this impact is not significant. Analysts generally maintain an optimistic outlook for this financial report, and in the past four quarters, NVDA's actual announced results have exceeded market expectations.
The most core market expected indicators are:
Revenue 28.6 billion USD, 110% YoY, 10% QoQ
EPS 0.63 USD, 133.3% YoY, 5% QoQ
Data center revenue 24.5 billion USD, 137% YoY, 8% QoQ
Profit margin 75.5%, the same as Q1
The most concerned issues are:
1. Has the Blackwell architecture been delayed?
UBS analysis believes that Nvidia’s first batch of Blackwell Chip shipments will be delayed by 4–6 weeks at most, and are expected to be delayed until the end of January 2025. Many customers have switched to purchasing H200, which has a very short delivery time. TSMC has begun production of Blackwell chips, but the initial production volume is lower than originally planned due to the complex CoWoS-L packaging technology used by B100 and B200, which has yield challenges, while H100 and H200 use CoWoS-S technology.
However, this new product was not originally included in the recent performance forecast:
Since Blackwell will not enter sales expectations until Q4 2024 (Q1 2025) at the earliest, and NVIDIA only provides single-quarter performance guidance, the delay will have little impact on the performance of Q2 and Q3 2024. At the recent SIG GRAPH conference, NVIDIA did not mention the impact of the Blackwell GPU delay, indicating that the impact of the delay may not be significant.
2. Has the demand for existing products increased?
Secondly, the decline in B100/B200 can be compensated by the increase in H200/H20 in the second half of 2024.
According to HSBC's forecast, the production of B100/B200 substrates (UBB) has been corrected by 44%, although deliveries may be partially postponed to the first half of 2025, resulting in reduced shipments in the second half of 2024. However, H200 UBB orders have increased significantly, with an estimated increase of 57% from the third quarter of 2024 to the first quarter of 2025. Based on this projection, H200 revenues of $23.5 billion in H2 2024 should more than offset the potential $19.5 billion loss in B100 and GB200 related revenues – equivalent to an implied revenue loss of 500,000 B100 GPUs or $15 billion, plus additional ancillary facility (NVL 36) revenue losses. 4.5 billion. We also see potential upside from strong H20 GPU momentum, which is primarily for the Chinese market, with possible shipments of 700,000 units or implied revenue of $6.3 billion in the second half of 2024.
In addition, a step-up in TSMC's CoWoS capacity may also support revenue growth from the supply side.
On the customer side, U.S. hyperscalers account for more than 50% of NVIDIA's data center revenue, and their recent comments suggest that NVIDIA's demand outlook will continue to increase. Goldman Sachs' forecast model shows that year-on-year growth in global cloud computing capital expenditures will reach 60% and 12% in 2024 and 2025, respectively, higher than previous forecasts (48% and 9%, respectively). But it can also be seen that this year is a year of growth, and it is impossible to maintain the same level of growth next year:
The following summarizes the recent comments of super-large technology companies on AI capital expenditures, showing these companies’ expectations for capital expenditure growth in 2024 and 2025:
Alphabet: Expects to spend $12 billion or more in capital expenditures per quarter for the remainder of 2024, with total spending likely to reach $12 billion to $13.5 billion.
Microsoft: Expects higher capital expenditures in 2025 than in 2024 to meet growth demand for its AI and cloud products.
Capex is expected to increase quarterly to meet cloud computing and AI demand that currently exceeds Microsoft's capacity.
Capacity constraints on AI in Azure cloud services, in particular, are expected to continue through the first half of fiscal 2025.
Meta: Raised its 2024 capital expenditure outlook to $37 billion to $40 billion, up from a previous range of $35 billion to $40 billion.
Capex is expected to increase significantly in 2025 as the company plans to invest to support its AI research and product development efforts.
Amazon: Capex is expected to be higher in the second half of 2025.
The main part of the capital expenditure will be used to support the company's growing demand for generative AI and non-generative AI workloads.
3. The degree of slowdown in momentum
In addition to the slowdown in spending growth of large companies next year, NV's performance growth will also slow further.
The market consensus expects revenue in fiscal 2025 to be $105.6 billion, compared with $60.9 billion last year, and the growth rate has slowed from 126% last year to 73%. Official guidance Q2 revenue is $28 billion, the market is expected to be more optimistic, but the growth rate of this quarter's performance will slow down further from the 2x% growth range to the 1x% growth range:
It should be noted that there are more and more participants in the artificial intelligence market: AMD's MI300X chip is said to be superior to Nvidia in some aspects. Cerebras has launched chips with a whole-wafer architecture, which significantly reduces interconnection and network costs and power consumption. In addition, major tech companies including Google, Amazon, and Microsoft are developing their own AI chips, which could reduce their reliance on Nvidia products in the future.
However, there are not enough cases to support this concern, and Wall Street still expects Nvidia to maintain its dominant position in data center chips:
4. Focus on China
In the upcoming earnings outlook, NVIDIA's demand trends in China will also be a focus of attention, especially when the market expects an increase in H20 demand. Pay attention to the following information in the earnings call:
How customer interest has changed since the launch of H20.
The company's competitiveness in the face of domestic competitors (mainly Huawei).
The timing of the launch of B20 (a scaled-down version of Blackwell) in 2025.
5. Product Line Changes
Due to the unprecedented production complexity faced by TSMC chip packaging (CoWoS-L vs. traditional CoWoS-S) and ARM-based Grace CPUs (vs. traditional x86 CPUs), it is possible to reduce the number of high-bandwidth memory stacks to reduce packaging complexity (allowing the use of traditional CoWoS-S instead of CoWoS-L), such as the new NVIDIA products B200A and GB200A Ultra may use the old CoWoS-S packaging, and the changes in technical specifications and cooling methods of the new products make it difficult for NVDA to maintain its previous high pricing power in the market:
Technical Specifications: The new A series has reduced performance compared to the standard B100 and B200 GPUs. Lower performance means that the market's price expectations for these new products will also be reduced. As a result, the average selling price (ASP) of the B200A is expected to be between $25,000 and $30,000, compared to the $35,000 to $40,000 ASPs of the previous B100 and B200 GPUs. This reduction in specifications and performance directly leads to a reduction in pricing power.
Cooling method: The upcoming GB200A Ultra NVL36 rack solution is expected to be air-cooled instead of a complex liquid cooling system. This change may result in lower consolidated revenue compared to the previous GB200 NVL36 and NVL72 racks.
This may have an uncertain impact on performance, which may be good for revenue on the one hand, but may also reduce NVDA's pricing power as its own product lines compete with each other.
6. Stock price fluctuations
Due to the postponement of Blackwell, the slight challenge to the AI narrative, and the overall market correction, Nvidia once fell 30% from its peak, but investors bought on dips, causing the stock price to rebound by 30%. The current market value of 3.18 trillion US dollars ranks second in the world, only 7% away from the historical high price
NVDA's current valuation level is basically at the median level of the past three years, neither high nor low.
7. Bull-Bear Hypothesis
The AI narrative encountered some challenges in the first two months, mainly from the perspective of the possible low contribution to corporate revenue. If based on this pessimistic assumption, that is, the investment boom in AI is a one-off, NVIDIA's data center business may quickly return to the trend level before 2023.
That is, assuming that data center revenue declines to US$69 billion in 2025, 38% lower than the current level and 55% lower than the baseline expectation, as the assumption of the "most pessimistic" scenario. Then the most optimistic assumption is that data center revenue growth reaches 100% in 2025.
Goldman Sachs' stock price change forecast under this bull-bear scenario is:
Based on the current stock price of $124.58, the baseline expected stock price is $135, Bull #1 and Bull #2 will rise by 41% and 89% respectively, and Bear #1Bear #2 will fall by 61% and 26% respectively. That is, at the current price level, NVDA's potential return is still less than the risk.
8. Summary: The trend is slowing down, but remain optimistic
NVDA's valuation is currently in a neutral range, and its performance is still satisfactory, but the market's most FOMO time has passed, and the growth rate of performance has entered a downward channel. It is obviously difficult for NVDA prices to replicate the 10-100 increase of that year.
The biggest risk is that the AI narrative is falsified, but as long as this matter does not continue to ferment, the negative impact on NVDA's performance is more of a mood swing. Other negatives come more from macro interest rates and geopolitical uncertainties. For example, the U.S. Department of Commerce will conduct an annual review of semiconductor export restrictions in October, which may prohibit the export of the "special edition" H20 chips that have weakened performance, and even affect the difficulty of obtaining the "castrated version" B20.
Uncertainties in the market may bring some stock price fluctuations, and valuation multiples may also shrink, but at this stage, we can still remain optimistic because the problem is mainly concentrated in the supply chain, not demand. Some supply chain problems can be solved and will not fundamentally undermine NVIDIA's long-term growth momentum. The company will remain attractive in the next few years. For example, the recent 30% rebound in just a few days shows the market's enthusiasm for bottom-fishing.
In particular, the outlook for AI demand may still be in its early stages. For example, Meta expects the computational load of its next-generation Llama 4 large language model to be 10 times higher than that of Llama 3.1, indicating that the long-term demand outlook for AI computing chips may exceed our expectations.
Gain a broader understanding of the crypto industry through informative reports, and engage in in-depth discussions with other like-minded authors and readers. You are welcome to join us in our growing Coinlive community:https://t.me/CoinliveSG
Add Comment
LoginLeave your comments
0 Comments
Earliest
Load more comments
More news about cerebras ai gpuimpossible
- Messari-promoted independent AI product “Messari AI”BullishBearish
- Microsoft May Be Working On an AI Edition of WindowsBullish1Bearish1
- Binance launches AI product BicassoBullishBearish
More news about cerebras ai gpuimpossible
- JinseFinance
Goodbye BTC, hello AI: Crypto mining companies turn to AI
Ali Fenn, president of Lancium, told CNBC that this will be one of the largest AI data center campuses in the world, further proving that the competition to support the development of artificial intelligence is becoming increasingly fierce, while Bitcoin is left behind.
JinseFinanceAI Beauty Pageant with AI Models & AI Judges Debuts Miss AI: Are We Going Too Far with AI?
Miss AI debuted as the first beauty contest with AI models and AI-generated judges just last month. The top 10 shortlist is now out, but are we pushing AI too far and distorting societal beauty standards?
KikyoOpenAI burst Sora detonates Web3+AI
Sora is an artificial intelligence model developed by OpenAI that can generate realistic and imaginative video scenes based on user-entered text instructions.
JinseFinanceTech Stocks Grapple with AI Frenzy's Aftermath
Post-AI frenzy, tech stocks plunge, questioning AI's financial viability. Delayed GenAI revenue raises doubts, signaling a tech industry transformation.
Xu LinA quick overview of Binance’s latest Launchpool project Sleepless AI (AI)
Binance New Coin Mining has now launched the 42nd phase of the project - Sleepless AI (AI), a Web3+AI gaming platform.
JinseFinanceAI x Crypto：盘点AI相关的加密项目
JinseFinanceWeb3 + AI: Crafting the Future of Connectivity
AI and Web3 have the potential to reshape industries. But challenges like data privacy, interoperability, and ethical AI usage must be addressed as these technologies converge.
CatherineAI War: Tron DAO Secures $100-M For AI Project, After ChatGPT And Google Push
In recent years, AI (artificial intelligence) has emerged as one of the most intriguing and quickly expanding technological narratives.
BitcoinistBlockchain and AI Don't Mix, Says Andre Cronje
Blockchain and AI have different underlying principles and are difficult to integrate in a single system, according to Andre Cronje.
Beincrypto