Nvidia: Small LLMs, Bigger Problems

Seekingalpha2023-12-25

Nvidia Corporation has experienced impressive growth in revenue and profits, but the rise of smaller, more cost-effective AI models and advancements by competitors like Intel may hinder future growth....

Source Link

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap

Leave a comment

{"i18n":{"language":"en_US"},"isChannel":false,"data":{"share":"https://ttm.financial/m/news/2393541576?lang=en_US&edition=fundamental","thumbnail":"","is_english":true,"pubTime":"2023-12-25 10:35","share_image_url":"https://static.laohu8.com/5a36db9d73b4222bc376d24ccc48c8a4","id":"2393541576","market":"us","top_or_hot":-1,"title":"Nvidia: Small LLMs, Bigger Problems","media":"Seekingalpha","content":"<div>\n<p>Nvidia Corporation has experienced impressive growth in revenue and profits, but the rise of smaller, more cost-effective AI models and advancements by competitors like Intel may hinder future growth....</p>\n\n<a href=\"https://seekingalpha.com/article/4659206-nvidia-small-llms-bigger-problems\">Source Link</a>\n\n</div>\n","source":"seekingalpha","html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>Nvidia: Small LLMs, Bigger Problems</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\nNvidia: Small LLMs, Bigger Problems\n</h2>\n\n<h4 class=\"meta\">\n\n\n2023-12-25 10:35 GMT+8&nbsp;&nbsp;&nbsp;<a href=https://seekingalpha.com/article/4659206-nvidia-small-llms-bigger-problems><strong>Seekingalpha</strong></a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\n<p>Nvidia Corporation has experienced impressive growth in revenue and profits, but the rise of smaller, more cost-effective AI models and advancements by competitors like Intel may hinder future growth....</p>\n\n<a href=\"https://seekingalpha.com/article/4659206-nvidia-small-llms-bigger-problems\">Source Link</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","isBrief":false,"type":0,"news_type":1,"symbol":"BK4514","symbol_name":"搜索引擎","start_time":0,"source_url":"https://seekingalpha.com/article/4659206-nvidia-small-llms-bigger-problems","article_id":"2393541576","we_media_id":null,"thumbnails":[],"rights":null,"url":"https://stock-news.laohu8.com/highlight/detail?id=2393541576","pubTimestamp":1703471753,"sourceInfo":{"source_id":"seekingalpha","name":"Seeking Alpha"},"weMediaInfo":null,"summary":"Nvidia Corporation has experienced impressive growth in revenue and profits, but the rise of smaller, more cost-effective AI models and advancements by competitors like Intel may hinder future growth.","collect":0,"end_time":0,"defaultTopTitle":"seekingalpha.com","property":[],"viewcount":null,"language":"en","relate_stocks":{"BK4514":"搜索引擎","LU0466842654.USD":"HSBC ISLAMIC GLOBAL EQUITY INDEX \"A\" (USD) ACC","LU0310800965.SGD":"FTIF - Templeton Global Balanced A Acc SGD","BK4529":"IDC概念","LU0097036916.USD":"贝莱德美国增长A2 USD","LU0320765059.SGD":"FTIF - Franklin US Opportunities A Acc SGD","LU0689472784.USD":"安联收益及增长基金Cl AM AT Acc","LU0198837287.USD":"UBS (LUX) EQUITY SICAV - USA GROWTH \"P\" (USD) ACC","IE00BD6J9T35.USD":"NEUBERGER BERMAN NEXT GENERATION MOBILITY \"A\" (USD) ACC","LU0444971666.USD":"天利全球科技基金","BK4515":"5G概念","LU1064131342.USD":"Fullerton Lux Funds - Global Absolute Alpha A Acc USD","NVDA":"英伟达","LU0312595415.SGD":"Schroder ISF Global Climate Change Equity A Acc SGD","LU0820561818.USD":"安联收益及增长平衡基金Cl AM DIS","LU0149725797.USD":"汇丰美国股市经济规模基金","BK4567":"ESG概念","IE00BKVL7J92.USD":"Legg Mason ClearBridge - US Equity Sustainability Leaders A Acc USD","LU0127658192.USD":"EASTSPRING INVESTMENTS GLOBAL TECHNOLOGY \"A\" (USD) ACC","BK4585":"ETF&股票定投概念","BK4576":"AR","LU0557290698.USD":"施罗德环球可持续增长基金","LU0211328371.USD":"TEMPLETON GLOBAL EQUITY INCOME \"A\" (MDIS) (USD)   INC","BK4566":"资本集团","LU0943347566.SGD":"安联收益及增长平衡基金AM H2-SGD","BK4535":"淡马锡持仓","LU0170899867.USD":"EASTSPRING INVESTMENTS WORLD VALUE EQUITY  \"A\" (USD) ACC","LU0642271901.SGD":"Janus Henderson Horizon Global Technology Leaders A2 SGD-H","LU0048573561.USD":"FIDELITY AMERICA \"A\" (USD) INC","BK4538":"云计算","IE00BJTD4V19.USD":"NEUBERGER BERMAN US LONG SHORT EQUITY \"A1\" (USD) ACC","BK4550":"红杉资本持仓","BK4077":"互动媒体与服务","LU0056508442.USD":"贝莱德世界科技基金A2","BK4141":"半导体产品","LU0719512351.SGD":"JPMorgan Funds - US Technology A (acc) SGD","LU0786609619.USD":"高盛全球千禧一代股票组合Acc","LU0310800379.SGD":"FTIF - Templeton Global A Acc SGD","LU0640476718.USD":"THREADNEEDLE (LUX) US CONTRARIAN CORE EQ \"AU\" (USD) ACC","IE00BJJMRY28.SGD":"Janus Henderson Balanced A Inc SGD","BK4551":"寇图资本持仓","LU1242518857.USD":"FULLERTON LUX FUNDS - ASIA ABSOLUTE ALPHA \"I\" (USD) ACC","IE00BMPRXN33.USD":"NEUBERGER BERMAN 5G CONNECTIVITY \"A\" (USD) ACC","BK4561":"索罗斯持仓","LU0820561909.HKD":"ALLIANZ INCOME AND GROWTH \"AM\" (HKD) INC","LU0109392836.USD":"富兰克林科技股A","BK4512":"苹果概念","IE0004445015.USD":"JANUS HENDERSON BALANCED \"A2\" (USD) ACC"},"translate_title":"英伟达：小LLMs，大问题","themeId":"95e3e7595ac2407d82a2e646db9ba6e0","isJumpTheme":false,"ttsUrl":null,"symbols_score_info":{"NVDA":1},"content_text":"Nvidia Corporation has experienced impressive growth in revenue and profits, but the rise of smaller, more cost-effective AI models and advancements by competitors like Intel may hinder future growth.Intel's 5th Gen Intel Xeon Scalable processors offer efficient AI workloads for smaller local models, potentially reducing the demand for Nvidia's high-end GPUs.Open-source models like Mistral 7B and the increase in viable alternatives from competitors are making AI technology more accessible and affordable, posing a challenge to Nvidia's EPS growth.Investment ThesisNvidia Corporation (NASDAQ: NVDA), one of the dominant forces in the artificial intelligence (\"AI\") and data center chip markets, has reached a crucial tipping point in its scale & growth. In 2023, the company has had impressive performance, with revenue growing 206% alone in the for the third quarter of FY 2024 to 18.12 billion. However, the rapid rise of an emerging trend in the AI sector may start to hamper their growth.More efficient Large Language Models (LLMs) capable of running locally and the significant advancements in AI chip technology by competitors like Intel (INTC) propose a possible roadblock for Nvidia. These smaller LLMs require much less data and computing power for training. Not only that, but they are much more cost-effective than the larger models that require Nvidia's high-end GPUs.What makes Intel's 5th Gen Intel Xeon Scalable processors so special is their ability to operate smaller AI workloads efficiently (like Nvidia's GPU solutions) but for smaller local models. This is due to their built-in AI acceleration. Nvidia's current growth predictions rely heavily on the demand for their top line GPU solutions, which is an issue because of these new processors diminishing the demand for Nvidia's GPU solutions. Therefore, I believe that these predictions are an overestimation of future growth for Nvidia. As developments in open-source models like Mistral 7B and viable alternatives from competitors continue, AI technology is becoming more accessible and affordable. Given this, Nvidia's stock growth could face a reality check, so I think it's a hold.BackgroundIn the recent fast-paced rise of AI technologies, Nvidia has emerged as one of the key players. 2023 has been huge for Nvidia, marked by a 206% year-on-year growth in the most recent quarter, with third-quarter FY 2024 revenue hitting $18.12 billion. This performance is based on Nvidia's current dominance in the AI chip market, where they possess approximately 70% market share, and their pioneering role in AI technology due to the success of its CUDA platform.Nvidia has been proactive when it comes to advancing their AI technology (yet another way they've strengthened their current leading position in the AI world). For example in 2021, Nvidia met face-to-face with one of their competitors, Intel, with the release of Grace CPU by Nvidia. This software directly challenged Intel's dominance in AI technology within the data center CPU market, allowing Nvidia to excel in this field. Nvidia later released the BlueField DPU (data processing unit), which is designed for AI and accelerated computing. Along with that they also unveiled the DGX SuperPOD, a turnkey AI supercomputer. Their innovation has been strong, this supports their 70% market share. How much more market share can they get?While Nvidia's growth has been strong, the creation of smaller, more efficient LLMs and advancements by competitors like Intel could add headwinds. The recent trend of localized and cost-effective AI solutions is causing a shift in the demand for LLMs. This could (in turn) shift demand for GPUs.Changes in the Last Month: New, Smaller LLMs Pose a ThreatMy hold on Nvidia stems from the industry's hard shift towards smaller, more efficient LLMs. These models can be run more locally, which reduces the demand for extensive and costly data center infrastructure.The infrastructures used for the development and training of large models like GPT-3 can be expensive, costing millions of dollars. On the other hand these new, smaller models, require less data and computing power. This decrease in power requirement will drastically lower operational costs for developers. For example, GPT-3, one of the more prevalent LLMs in this year's AI sector, cost OpenAI over $4.6 million to train. This cost could now be easily avoided using smaller models which can be developed and deployed at a fraction of this cost. Because of the drop in development cost, these smaller LLMs will be much more accessible to businesses. The opportunity for AI development (as costs drop) will expand to small and medium-sized enterprises (SMEs).Not only do smaller LLMs vastly reduce cost, but they are also domain-specific to a company's needs. By domain-specific, I mean that these LLMs are able to adapt specifically to a company's needs. This can be done by prompt and adapter tuning, which then allows those companies to adjust these models. This increased adaptability allows businesses to achieve more accurate and relevant AI-powered solutions without the overhead associated with larger models.As smaller LLMs gain in popularity large-scale models such as those produced by Nvidia are faced with a serious competitor. The growing trend of developing and applying smaller, more efficient models are reducing the need for high-end GPUs. Intel's new chips are an example of where these models could go.Mistral's Advancements are Lowering the API Costs of Models. These Margins may be too Low for many Firms to Afford New NVDA ChipsThe creation of open-source LLMs like Mistral 7B, a 7-billion-parameter model, has significantly impacted the AI model market, especially when it comes to API costs and hardware requirements. This model is presented as a favorable open-source alternative to other AI solutions, therefore in some cases providing superior performances. The release of this model is consistent with the trend of more affordable, accessible, and cost efficient AI models.As my research of Mistral 7B has advanced, I've become more and more impressed. Mistral 7B is already making its own name in the world of AI, outperforming the Llama 2 13B model in all tests and competing closely with the Llama 1 34B in many metrics. Although this model has similar capabilities to Llama 2, the computational overhead is much lower, therefore presented as a more cost effective solution for businesses. Along with that, Mistral AI's approach to open-source, community-driven development is predicted to provide a more adaptable and ethical oversight in AI applications.With the development of Mistrak 7B, there has been a shift in the charging model. Instead of their cost being calculated per token of GPU usage, it is now being calculated per minute. The change can lead to companies saving vast amounts of money, especially if they are using Mistrak 7B for applications that don't need a continuous high GPU usage.With changes means dropping prices. Major AI model providers like Anthropic and OpenAI have cut their API costs meaningfully over the last two months. With cheaper models comes margin compression and an obvious emphasis to cut costs, especially at smaller AI startups.Nvidia's new H100 chips cost twice as much to train on according to data from CoreWeave (an Nvidia-backed startup). While these H100s can train models faster, many new smaller models do not need such intensive training. This is where Intel's new cost-conscious (yet powerful) chips come in.What's Changed in the Last Week: Intel's New Chips Make Cheaper Hosting PossibleOn December 14th, 2023, Intel revealed their newest technology, the 5th Gen Intel Xeon Scalable processor. The release of this processor resulted in a pivotal moment in AI advancement. It presented a new technology that was capable of a large increase in overall performance for a variety of critical workloads, including artificial intelligence, analytics, networking, security, storage, and high-performance computing. What makes these processors so important is they provide an increased performance per watt, thereby lowering the operating cost. By producing a more cost-effective solution to AI, businesses could start shifting towards the use of Intel's new technologies rather than Nvidia.As I mentioned above, one of the features of the 5th Gen Intel Xeon processors is that performance is being increased per watt. This can be affiliated with the built-in AI acceleration in every core. This allows processors to handle demanding AI workloads, including deep learning inference and fine-tuning on models up to 20 billion parameters, while still producing effective results. This positions the 5th Gen Xeon processors as a viable option for running smaller, more localized AI models, such as LLMs.There are significant improvements in performance in these new processors. For example, they provide a 21% average performance gain for general compute performance and result in a 36% higher average performance per watt. The cost of ownership compared if other processors can decrease by up to 77%. These are incredibly cost conscious.Intel's processors have already shown promising results in various applications. For example, IBM's watsonx.data platform achieved up to 2.7x better query throughput with these processors compared to previous-generation Xeon processors. Google Cloud plans to implement the 5th Gen Xeon next year and is predicting increased performance in AI-based applications. These processors are being used across all types of companies, for example an indie game studio reported a 6.5x improvement in inference performance using Xeon processors over a GPU-based cloud instance.Growth Will Slow and Year-Over-Year Hardware Comps are ChallengingThe growth outlook for Nvidia in the coming years presents a complex picture. The 3rd quarter FY 2024 year-on-year revenue growth of 206% reflects Nvidia's current dominance in the AI chip market, where it holds approximately 70% of the market share. Growth like this will be hard to sustain.However, while Nvidia's current performance is strong, the future might look a little different. The increasing development and adoption of smaller, more efficient LLMs and advancements by competitors in the AI chip market could reduce the demand for Nvidia's high-end GPUs. Because of these new LLMs advancements, Nvidia's growth might not maintain its current performance.In FY 2024 (current year) Nvidia is projected to hit revenue of $58.77 billion. Next fiscal year revenue is supposed to grow an additional 54% to $90.66 billion. Keep in mind that most of Nvidia's revenue comes from Chip sales (which is not a recurring revenue model like a Software as a Service, or SAAS, company). While this is an obvious statement, this part is key: they have to sell the same number of chips they will sell this fiscal year next year, but then plus another 54%. This assumes the pricing mix on these chips does not fall.As models get smaller and costs to run the models are compressed as pricing compresses, AI companies will look to make the most of the hardware they have. This will mean longer cycle times going forward for hardware upgrades, and making the most of the hardware they currently have (in many cases this hardware is sufficient for smaller models).Many of these companies that contributed to the sales of chips this year may not be marginal buyers next year as models get smaller and chip needs are not as strong. This will likely not cause revenue to drop, but growth to slow below the market expectations.ValuationWith Nvidia facing tougher year-over-year comps for growth (and expecting to grow another 54% above that), I think some of their key valuation metrics are high and warrant a 'wait and see' approach, which is why I rated this stock a hold.For example, Nvidia's forward price to sales ratio is 20.22, almost 7 times higher than the industry average of 2.94. While the company's strong growth has helped push this price to sales ratio up, part of the key here is the firm's extremely high margins on their chips, which means while the high forward P/E ratio of 39.11 is notable, it's not as challenging to understand.However, these margins (rumored to be over 90% on some chips or a 1000% markup), are insane margins for hardware. These margins could compress if Nvidia's demand growth starts to slow. This could decrease effective EPS growth making the P/E ratio seem high and justify pressure on the price to sales ratio as well.TakeawayAs Nvidia navigates the evolving AI world, its future is somewhat murky with growth. While 2023 was a great year for them, I am hesitant to predict the same for the future of this company. The growing trend of smaller and most cost effective LLM technology may reduce the growth demand for Nvidia's high-end GPUs. Nvidia's competitors have significantly contributed to this shift in the AI market by producing new technologies such as Mistral 7B and Intel's 5th Gen Intel Xeon Scalable processors.Nvidia has a solid foundation due to their observed strong market position, however in order to stay consistent with their current stock performance, they will need to be very strategic in how they respond to their competitors and the ever-changing world of AI. In essence, Nvidia Corporation will have to keep these incredible growth rates up in order to justify the valuation they have (this is despite increased competition and different needs from LLM AI companies). I am not saying it's impossible, I am just saying it is difficult. In essence, until we have more clarity, Nvidia Corporation stock is a hold.","kind":null,"is_publish_news":true,"is_publish_highlight":false,"is_publish_live":false,"is_publish_wemedia":null,"editions":null,"column":"","sentiment":"0","news_tag":"","news_rank":0,"symbols":[],"gpt_button":0},"commentList":[],"isCommentEnd":true,"newsSizeData":{"likeSize":0,"commentSize":0,"repostSize":0,"favoriteSize":0,"likeStatus":false,"favoriteStatus":false},"APP":{"userAgent":"Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com)","isDev":false,"isTTM":true,"deviceId":"web-server-community-laohu8-v3","version":"4.29.1","shortVersion":"4.29.1","platform":"web","vendor":"web","appName":"ttm","isIOS":false,"isAndroid":false,"isTiger":false,"isTHS":false,"isWeiXin":false,"isWeiXinMini":false,"isWeiBo":false,"isQQ":false,"isBaiduSwan":false,"isBaiduBox":false,"isDingTalk":false,"isToutiao":false,"isOnePlus":false,"isHuaWei":false,"isXiaomi":false,"isXiaomiWebView":false,"isOppo":false,"isVivo":false,"isSamsung":false,"isMobile":false},"href":"/m/news/2393541576"}