|
|
|
|
|
|
Posted: Tue Jan 28, 2025 1:57 pm
Luria addressed inquiries about OpenAI's potential strategies to deter competitors like DeepSeek from emulating their advanced techniques. He noted that OpenAI's current approach, which includes concealing chain-of-thought reasoning tokens, might be the extent of their defensive capabilities. The analyst pointed out that simply delaying the release of new developments would not be effective, as competitors have already made significant advancements.
DeepSeek has recently launched Janus-Pro-7B, a text-to-image model that surpasses other models like Stable Diffusion and DALL-E 3 in performance. This development exemplifies the intense competition in the AI space and the pressure on OpenAI to continually innovate to maintain its edge.
According to Luria, OpenAI's most viable strategy is to accelerate its innovation cycle and outpace open-source labs. Additionally, finding ways to reduce inference costs is crucial for OpenAI's success. The analyst's remarks reflect the dynamic nature of the AI industry and the need for continuous innovation to stay ahead. NVIDIA's financial strength is evident in its remarkable 152% revenue growth and 76% gross profit margin in the last twelve months, demonstrating its dominant position in the AI chip market.
|
 |
 |
|
|
|
|
|
|
|
|
Posted: Tue Jan 28, 2025 1:57 pm
Luria addressed inquiries about OpenAI's potential strategies to deter competitors like DeepSeek from emulating their advanced techniques. He noted that OpenAI's current approach, which includes concealing chain-of-thought reasoning tokens, might be the extent of their defensive capabilities. The analyst pointed out that simply delaying the release of new developments would not be effective, as competitors have already made significant advancements.
DeepSeek has recently launched Janus-Pro-7B, a text-to-image model that surpasses other models like Stable Diffusion and DALL-E 3 in performance. This development exemplifies the intense competition in the AI space and the pressure on OpenAI to continually innovate to maintain its edge.
According to Luria, OpenAI's most viable strategy is to accelerate its innovation cycle and outpace open-source labs. Additionally, finding ways to reduce inference costs is crucial for OpenAI's success. The analyst's remarks reflect the dynamic nature of the AI industry and the need for continuous innovation to stay ahead. NVIDIA's financial strength is evident in its remarkable 152% revenue growth and 76% gross profit margin in the last twelve months, demonstrating its dominant position in the AI chip market.
|
 |
 |
|
|
|
|
|
|
|
|
|
|
Posted: Tue Jan 28, 2025 7:37 pm
Insurer Chubb Ltd. has estimated its costs related to the Los Angeles wildfires to be around $1.5 billion before taxes, to be incurred in the company’s first quarter.
Chubb CB
-1.53% Chief Executive Evan Greenberg said late Tuesday that Chubb employees have been on the ground trying to help policyholders as the tragedy is still unfolding.
|
 |
 |
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:04 am
OPENAI ACCUSES CHINA’S DEEPSEEK OF USING ITS MODEL FOR TRAINING. - FT. Surprise surprise. OpenAI has found evidence that Chinese AI startup DeepSeek used its proprietary models to train its own open-source model, according to the Financial Times. The company says it detected signs of "distillation"—a technique where developers use outputs from larger models to improve smaller ones, allowing them to achieve similar results at a lower cost. OpenAI declined to provide further details but emphasized that its terms prohibit copying its services or using outputs to develop competing models.
|
 |
 |
|
|
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:06 am
OPENAI ACCUSES CHINA’S DEEPSEEK OF USING ITS MODEL FOR TRAINING. - FT. Surprise surprise. OpenAI has found evidence that Chinese AI startup DeepSeek used its proprietary models to train its own open-source model, according to the Financial Times. The company says it detected signs of "distillation"—a technique where developers use outputs from larger models to improve smaller ones, allowing them to achieve similar results at a lower cost. OpenAI declined to provide further details but emphasized that its terms prohibit copying its services or using outputs to develop competing models.
|
 |
 |
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:07 am
OPENAI ACCUSES CHINA’S DEEPSEEK OF USING ITS MODEL FOR TRAINING. - FT. Surprise surprise. OpenAI has found evidence that Chinese AI startup DeepSeek used its proprietary models to train its own open-source model, according to the Financial Times. The company says it detected signs of "distillation"—a technique where developers use outputs from larger models to improve smaller ones, allowing them to achieve similar results at a lower cost. OpenAI declined to provide further details but emphasized that its terms prohibit copying its services or using outputs to develop competing models.
|
 |
 |
|
|
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:09 am
eek the biggest near term determinant for many AI stocks will be META and MSFT's Capex numbers. I am optimistic that they will maintain their CAPEX, but I admit I am not sure. So yes, the CAPEX number for META and MSFT in particular will be the main driver of price action for these AI hardware stocks, including Nvidia as well as data centre stocks like NBIS, and energy stocks like OKLO.
If CAPEX is maintained, and comes in line with expectations, then it will be a suggestion that the mag7 firms are shrugging off the claims that Deepseek have built their LLM with only $6M. It will suggest that they are not worried that they have overspent on their AI infrastructure and ambition, which will be seen as a massive positive for these AI firms, as all that fear that these firms will now pullback their CAPEX in light of Deepseek's so called efficiency, will be dismissed.
On the other hand, if CAPEX guidance is cut significantly, it will be seen as an admission that these firms are worried that they have overspent on their AI ambitions in light of Deepseek, and are now curtailing spend until they can figure things out. This will be seen as a massive issue for AI hardware and we can see another day like Monday, as fears will reignite that the biggest buyers of AI hardware are pulling back their spending.
In my opinion, the former scenario is significantly more likely. Firstly, if we consider META. Well, Zuckerberg was on the Joe Rogan podcast recently and was already talking about Deepseek. He mentioned that it was an impressive LLM. Presumably, he knew of their efficiency claims. Yet, despite this, last Friday he announced a MASSIVE capex increase. By almost a third! He wouldn't do this if he thought that Deepseek may have shown META up for overspending. And he likely wouldn't have done this, only to walk back on the increased CAPEX the next week, due to the existence of a Chinese LLM that he was already aware of.
It is clear that META's capex is building towards something far more than LLM. It is building towards fully fledged AGI. Additionally, Deepseek may even be seen as a positive for META as it validates their open source Llama ambitions.
I don't see it likely that META cuts capex at all.
Then when we look at MSFT, although Deepseek is a more direct competitor for OpenAI, I again see it likely that they will maintain their CAPEX. They too are using their CAPEX for much more than just LLMs. And they will likely understand that Deepseek's claims of building their model for just $6m are near impossible.
I think then that if these companies maintain their CAPEX as I expect, we can see another ripper day for AI hardware stocks, particularly Nvidia.
Nvidia is the biggest beneficiary of META capex. If Meta maintain their CAPEX, as I believe they will, it will mean more revenue for Nvidia. It will also set a precedence that despite Deepseek, one of the biggest spenders on NVDA GPUs is continuing to spend on them at the same rate. This will bring other firms in line with that same move.
My view that they will hold CAPEX in line with expectations is the prediction of Wedbush too.
This is their note:
We expect Meta and Microsoft to reiterate their massive 2025 AI driven Capex numbers of $60 billion-$65 billion and $80 billion respectively with a firm tone on tomorrow night’s conf call. This is what the Street is focused on after the DeepSeek LLM/model heard around the world.
I have no inside scoop on META or MSFT so I can only hypothesise as well, but this is where my thinking is leading me right now.
|
 |
 |
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:13 am
Multiple expansion moves from industry to industry. For example, software companies saw a significant re-rating without any underlying improvement in fundamentals after the release of deepseek: investors rotated from hardware to software. As a trader, you need to follow the trail of multiple expansion
|
 |
 |
|
|
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:16 am
Multiple expansion moves from industry to industry. For example, software companies saw a significant re-rating without any underlying improvement in fundamentals after the release of deepseek: investors rotated from hardware to software. As a trader, you need to follow the trail of multiple expansion
|
 |
 |
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:18 am
Multiple expansion moves from industry to industry. For example, software companies saw a significant re-rating without any underlying improvement in fundamentals after the release of deepseek: investors rotated from hardware to software. As a trader, you need to follow the trail of multiple expansion
|
 |
 |
|
|
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:25 am
Oppenheimer downgrade AAPL to perform from outperform. Cites issues with increased competition in China and lack of compelling Apple Intelligence features. I agree with both headwinds We lower our FY26 EPS estimate by 4% to $7.95, below consensus of $8.23. Our revision is based on reduced estimates for iPhone sales over the next 12-18 months. We see a twofold challenge ahead for iPhone growth: 1) Stronger competition in Greater China. 2) Lack of compelling Apple Intelligence and generative AI apps to accelerate near-term device replacement.
|
 |
 |
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:26 am
Oppenheimer downgrade AAPL to perform from outperform. Cites issues with increased competition in China and lack of compelling Apple Intelligence features. I agree with both headwinds We lower our FY26 EPS estimate by 4% to $7.95, below consensus of $8.23. Our revision is based on reduced estimates for iPhone sales over the next 12-18 months. We see a twofold challenge ahead for iPhone growth: 1) Stronger competition in Greater China. 2) Lack of compelling Apple Intelligence and generative AI apps to accelerate near-term device replacement.
|
 |
 |
|
|
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 4:35 am
Bernstein take on STRONG ASML earnings. I agree with the points here. Gives a better summary than the overall numbers but will post them soon too "Q4 and 2024 came in ahead of expectations due to stronger installed base services. Gross margin was supported by a strong upgrades business and better-than-anticipated costs to reach customer acceptance for two High-NA EUV systems. Bookings were the standout once again, coming in well ahead of expectations at €7.1B (vs. consensus €3.5B). While a very solid set of results, it’s far too soon to have visibility on any impact from changes in AI demand driven by DeepSeek news. Q4 revenues were a solid beat at €9.3B (vs. guide €9.0B / consensus €9.02B / Bernstein €9.09B). Strong performance in installed base management (services) was also notable at €2.1B, supporting a stronger-than-anticipated gross margin at 51.7% (vs. guide/consensus/Bernstein 50.6%). EPS followed strongly at €6.85 (vs. consensus €6.67 / Bernstein €6.61). Investors will likely be comforted by resilient memory performance in Q4, at 39% of systems sales and 39% of bookings (memory was 46% of sales in Q2 and 36% in Q3). This puts full-year 2024 revenues at €28.3B (vs. consensus/Bernstein €28.1B) and gross margin at 51.3% (vs. guide/consensus/Bernstein 50.6%). The Q1 sales guide was also positive, ahead of expectations at €7.75B at the midpoint (vs. consensus €7.25B / Bernstein €7.11B), with margin guided to a midpoint of 52.5% (vs. consensus 51.2% / Bernstein 51.9%). The full-year guide for 2025 was maintained at €32.5B at the midpoint, with gross margin still guided at 52%. Looking ahead, ASML has indicated they will stop providing bookings numbers on a quarterly basis and will only update the backlog annually. Investors had been increasingly focused on ‘lumpy’ quarterly bookings as a means to underpin guidance, but ASML reiterated that guidance is based on regular customer reviews. China was just below our expectations as a share of systems revenue at 27% (vs. just under 50% of revenues in the first nine months of 2024). Overall, this put China system sales at 41% for the full year 2024, up from 29% in 2023. $ASML continues to expect this to normalize back to pre-2023 levels at around 20% of total revenues in 2025. While overall a solid set of results, and the strong bookings number should reassure investors about 2025, we believe there is still a lot to play out in AI demand as the DeepSeek impact becomes clearer. Longer-term, we remain bullish on the opportunity for ASML, but caution may prevail until near-term implications are better understood."
|
 |
 |
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 1:11 pm
Bernstein take on STRONG ASML earnings. I agree with the points here. Gives a better summary than the overall numbers but will post them soon too "Q4 and 2024 came in ahead of expectations due to stronger installed base services. Gross margin was supported by a strong upgrades business and better-than-anticipated costs to reach customer acceptance for two High-NA EUV systems.
Bookings were the standout once again, coming in well ahead of expectations at €7.1B (vs. consensus €3.5B). While a very solid set of results, it’s far too soon to have visibility on any impact from changes in AI demand driven by DeepSeek news.
Q4 revenues were a solid beat at €9.3B (vs. guide €9.0B / consensus €9.02B / Bernstein €9.09B). Strong performance in installed base management (services) was also notable at €2.1B, supporting a stronger-than-anticipated gross margin at 51.7% (vs. guide/consensus/Bernstein 50.6%). EPS followed strongly at €6.85 (vs. consensus €6.67 / Bernstein €6.61).
Investors will likely be comforted by resilient memory performance in Q4, at 39% of systems sales and 39% of bookings (memory was 46% of sales in Q2 and 36% in Q3). This puts full-year 2024 revenues at €28.3B (vs. consensus/Bernstein €28.1B) and gross margin at 51.3% (vs. guide/consensus/Bernstein 50.6%).
The Q1 sales guide was also positive, ahead of expectations at €7.75B at the midpoint (vs. consensus €7.25B / Bernstein €7.11B), with margin guided to a midpoint of 52.5% (vs. consensus 51.2% / Bernstein 51.9%). The full-year guide for 2025 was maintained at €32.5B at the midpoint, with gross margin still guided at 52%.
Looking ahead, ASML has indicated they will stop providing bookings numbers on a quarterly basis and will only update the backlog annually. Investors had been increasingly focused on ‘lumpy’ quarterly bookings as a means to underpin guidance, but ASML reiterated that guidance is based on regular customer reviews.
China was just below our expectations as a share of systems revenue at 27% (vs. just under 50% of revenues in the first nine months of 2024). Overall, this put China system sales at 41% for the full year 2024, up from 29% in 2023. $ASML continues to expect this to normalize back to pre-2023 levels at around 20% of total revenues in 2025.
While overall a solid set of results, and the strong bookings number should reassure investors about 2025, we believe there is still a lot to play out in AI demand as the DeepSeek impact becomes clearer. Longer-term, we remain bullish on the opportunity for ASML, but caution may prevail until near-term implications are better understood."
|
 |
 |
|
|
|
|
|
|
|
|
|
|
Posted: Wed Jan 29, 2025 1:21 pm
Premarket report 29/01 ahead of FOMC and Big Tech earnings. FOMC decision today - expectation is for a dovish FOMC.
DEEPSEEK:
OPENAI accuses China's deepseek of using its model for Training: OpenAI has found evidence that Chinese AI startup DeepSeek used its proprietary models to train its own open-source model, according to the Financial Times. The company says it detected signs of "distillation"—a technique where developers use outputs from larger models to improve smaller ones, allowing them to achieve similar results at a lower cost. OpenAI declined to provide further details but emphasized that its terms prohibit copying its services or using outputs to develop competing models. U.S. Navy bans use of DeepSeek due to ‘security and ethical concerns Semianalysis, who are one of the best at semiconductor research IMO, corroborate that Deepseek had 10s of thousands of Nvidia GPUs, spending over $500M ASML on Deepseek: “For AI to be everywhere, we need to see major progress on costs and power consumption. Lowering costs also leads to more volume...So I will say any technology, whatever it is, that will contribute to cost reduction will increase the opportunity.”
Main things to watch in the market:
Right now, the key thing to watch is the CAPEX number for META and MSFT when they announce earnings tomorrow. If it is massively down from expectations, this will be seen as another selling trigger. Capex guidance in line will be seen as a reiteration that these firms are not massively concerned and will lead to a rip in semiconductors. The 2nd scenario is my base case. MAG7:
AAPL - Apple (AAPL) has been in secret cooperation with SpaceX and T-Mobile (TMUS) to support the Starlink satellite service in its latest iPhone operating system, which would serve as an alternative to Apple's current in-house satellite communication service, Bloomberg's Mark Gurman reports, citing people familiar with the matter. AAPL - Oppenheimer downgrades AAPL to perform from outperform, lowers FY26 EPS estimate by 4% to $7.95, below consensus of $8.23. Our revision is based on reduced estimates for iPhone sales over the next 12-18 months. Said main headwinds for the company are stronger competition in China, and lack of compelling AI features META - CEO Mark Zuckerberg is exploring the purchase of a property in Washington, D.C., according to sources cited by the Financial Times. Notable in implying improved relationship with Trump. NVDA - Ming Chi Kuo, one of the top semiconductor analysts, says that While scaling laws are hitting limits, Nvidia remains the dominant player. Kuo suggests its edge will grow when scaling laws regain momentum, but short-term supply chain issues with the GB200 NVL72 could impact expectations. TSLA - NORWAY SOVEREIGN WEALTH FUND CEO: NO PLANS TO PULL OUT OF TESLA
EARNINGS:
ASML:
ASML BOOKINGS COME IN MORE THAN DOUBLE EXPECTATIONS -- POSTS $7.40 BILLION IN TOTAL BOOKINGS VS. $3.68 BILLION ESTIMATE.
Bookings is a precursor for future revenue, so this is a massive tailwind.
EPS: €6.84 (Est. €6.6 cool ; UP +31.5% YoY 🟢 Revenue: €9.26B (Est. €9.02B) ; UP +28% YoY 🟢 Net Bookings: €7.09B (Est. €3.53B) 🟢 WTF DOUBLE EXPECTED Lithography Systems Sold: 132 (Est. 121) ; UP +6.5% YoY 🟢 Gross Margin: 51.7% (Est. 49.6%) ; +35 bps YoY 🟢 Gross Profit: €4.79B (Est. €4.47B) ; UP +28.9% YoY 🟢 Operating Profit: €3.36B (Est. €3.09B) ; UP +40.3% YoY 🟢 Operating Margin: 36.2% (Est. 34.3%) ; +317 bps YoY 🟢 Operating Cash Flow: €9.55B (Est. €3.66B) ; UP +198.6% YoY 🟢 Backlog stands at €36 billion, and remain confident in their outlook for 2025. Q1’25 Guidance
Revenue: €7.75B (Est. €7.25B) ; UP +46.5% YoY 🟢 Gross Margin: 52.5% (Est. 51.2%) ; +150 bps YoY 🟢 SG&A Expense: €290M (Est. €300M) ; UP +2.1% YoY 🔴 R&D Expense: €1.14B (Est. €1.10B) ; UP +9.5% YoY 🟢 FY’25 Outlook
Operating Profit: €32.50B (Est. €32.19B) ; UP +15% YoY 🟢 Gross Margin: 51%-53% (Est. 52.0%) ; +70 bps YoY 🟢 Net Sales: €30B-€35B (Est. €32.19B) 🟢 CEO Christophe Fouquet’s Commentary:
“Our fourth-quarter was a record in terms of revenue, driven by additional upgrades and first revenue recognition of two High NA EUV systems. 2024 overall was another record year, with total net sales of €28.3 billion and gross margin of 51.3%. Looking ahead, we see Q1 ’25 net sales in the range of €7.5 billion to €8.0 billion, gross margin between 52% and 53%, and full-year sales of €30 billion to €35 billion. AI remains a key growth driver for our industry, though it creates shifting market dynamics among our customers. We’re confident our lithography leadership will continue to serve as a critical enabler for advanced semiconductor roadmaps.”
SBUX:
Results were't great, however restructuring plans have piqued investor optimism enough for the stock to be up in premarket. Same-store sales fell 4%, marking four straight quarters of decline. U.S. sales dropped 4% with an 8% transaction decline, while China saw a 6% drop. To counter this, Starbucks will cut its menu by 30% by the end of 2025, aiming to simplify operations and focus on core offerings. CEO Brian Niccol outlined the "Back to Starbucks" strategy, balancing customer focus across Gen Z to the 50-60+ demographic. Also expanded with 377 new stores and is testing a mobile order algorithm to smooth out rush periods. Reports Q1 U.S. gift card loads of $3.5B, maintaining its #2 ranking for gift card sales in the U.S.
Revenue: $9.4B (Est. $9.42B) ; Flat YoY 🔴 EPS: $0.69 (Est. $0.67) ; DOWN -23% YoY🟢 Operating Margin: 11.9% (Contracted 390 bps YoY)🔴 Global Comparable Store Sales: DOWN -4% YoY (Est. -4.06%) 🟢 Segment Highlights:
North America Segment Revenue: $7.07B; DOWN -1% YoY Comparable Store Sales: DOWN -4% YoY Transactions: DOWN -8% YoY Ticket Size: UP +4% YoY Operating Income: $1.18B; DOWN -22% YoY Operating Margin: 16.7% (Contracted 470 bps YoY) International Segment
Revenue: $1.87B; UP +1% YoY Comparable Store Sales: DOWN -4% YoY Transactions: DOWN -2% YoY Ticket Size: DOWN -2% YoY Operating Margin: 12.7% (Contracted 40 bps YoY)
OTHER COMPANIES:
SEmicodncutor equipment companies such as KLAC, LRCX etc are all higher on stellar ASML earnings. MSTR - Mizuho initiates with outperform rating, PT of 515. Said for Bitcoin, rising global adoption, slowing rate of bitcoin supply growth, and favorable political environment support price appreciation. BABA - just launched its new Qwen2.5-Max AI model, outperforming DeepSeek V3 in benchmarks like Arena Hard, LiveBench, LiveCodeBench, and GPQA-Diamond. MDB and SNOw - Bernstein says that they could be key beneficiaries of more inferencing. If the cost of inferencing decreases and the availability of models, especially SLM increases then more enterprises over time will build Gen AI apps which is incrementally positive for MongoDB and Snowflake RDDT - appears to have added Meta campaign import feature, allowing advertisers to instantly import campaigns from meta into Reddit. This reduces friction and makes Reddit more convenient for new users to test out LYFT - Lowers PT to 19 from 21. Says that Waymo’s future launch of freeway rides to the public will likely be a negative for LYFT but the Street could be overestimating the timeline to public launch. HIMS - will run its first superbowl ad on feb 9th MNDY - JPM opens positive catalyst watch on MNDY, reiterates overweight rating, with PT of 350. Said they are positively biased following analysis of qualitative feedback from channel partners, quantitative data on headcount growth and website visits. Said partner conversations indicate that after a brief slowdown in September, demand in the U.S. enterprise segment has recovered meaningfully through December and into 2025. VKTX - HC Wainwright & Co. Reiterates Buy on Viking Therapeutics, Maintains $102 Price Target RIVN - initiated at 'Underperform' at Bernstein, with a $6.10 price-target, implying over -50% of downside: it faces slower market growth, rising competition, and limits to the segments its brand can address.breakeven years away and mounting risks, BA - UBS raises PT to 217 from 208, maintains buy rating. MAX production and deliveries have resumed at a level higher than we anticipated, and Boeing communicated good progress on both the supply chain ramp-up and FAA KPIs and cooperation. We still see supply chain risk, given the strike stoppage—but engine and fuselages are doing better, and inventory buffer should help stabilize Boeing's own ramp. while a lot still needs to be done, we believe new CEO Kelly Ortberg has laid out a sound strategy PATH - dropped a PR yesterday highlighting some stats about AgenticAI. 90% of U.S. IT executives believe agentic AI can enhance business processes; 77% of them are planning to invest in it this year. It seems like they are trying to position themselves as an agentic AI play. DDOG - downgraded to hold from Buy, lowers PT to 140 from 165. foresee revenue growth and margin headwinds throughout FY25. Combined with a fairly full valuation (~13.5x CY26e EV/Revenue and ~47x EV/FCF) following the stock’s strong performance (+~10% relative to IGV) since the early November 3Q report, the risk-reward profile for the stock appears less favorable in the coming quarters. KSS - plans to reduce its corporate workforce by 10%, according to the WSJ. NVO's Ozempic Ozempic is now indicated to reduce the progression of kidney disease. COIN - Mizuho upgrades Coinbase to Neutral on correlation to bitcoin price PT $290 up from $250 RKLB - Rocket Lab price target raised to $32 from $27 at KeyBanc OTHER NEWS:
President Trump is offering buyouts to federal workers unwilling to return to the office, provided they resign by February 6th, Axios reports. DOGE just announced that they are saving the US Government $1 billion PER DAY
|
 |
 |
|
|
|
|
|
|
 |
|
|
|
|
|