Meta Platforms' ChatGPT-3 Moment - Buy or Sell META Stocks?

Summary:

Meta Platforms recently published its latest work on the Segment Anything Model (“SAM”).
Nvidia AI specialist exclaimed that this marks the GPT-3 moment in the field of computer vision.
This article analyzes the immediate catalysts that Meta can benefit from such AI breakthroughs without waiting for more futuristic VR products.

<figure class=”getty-figure” data-type=”getty-image”><picture><img src=”https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w750″ alt=”Artificial Intelligence. Concept” data-id=”1364859722″ data-type=”getty-image” width=”1536px” height=”1024px” srcset=”https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w1536 1536w, https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w1280 1280w, https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w1080 1080w, https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w750 750w, https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w640 640w, https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w480 480w, https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w320 320w, https://static.seekingalpha.com/cdn/s3/uploads/getty_images/1364859722/image_1364859722.jpg?io=getty-c-w240 240w” sizes=”(max-width: 768px) calc(100vw – 36px), (max-width: 1024px) calc(100vw – 132px), (max-width: 1200px) calc(66.6vw – 72px), 600px” loading=”lazy”></picture><figcaption>BlackJack3D</figcaption></figure> <h2>Thesis</h2> Meta Platforms (NASDAQ:<a href=”https://seekingalpha.com/symbol/META” title=”Meta Platforms, Inc.”>META</a>) recently published their latest work on the Segment Anything Model (“SAM”). For investors who are interested in general AI and computer vision, you can find the full paper at this <a href=”https://ai.facebook.com/research/publications/segment-anything/” rel=”nofollow” title=”https://ai.facebook.com/research/publications/segment-anything/” target=”_blank”>Meta AI link</a>. For those who do not like reading technical papers, the Twitter thread by Nvidia AI scientist Dr. Jim Fan provides a more popularized summary of the gist, exclaiming that this marks the GPT-3 moment in the field of computer vision (“CV”). <a href=”https://twitter.com/DrJimFan/status/1643647849824161792″ rel=”nofollow” class=”paywall-full-content invisible”>Quote</a>: <blockquote class=”paywall-full-content invisible”>Reading @MetaAI’s Segment-Anything, and I believe today is one of the “GPT-3 moments” in computer vision. It has learned the *general* concept of what an “object” is, even for unknown objects, unfamiliar scenes (e.g., underwater & cell microscopy), and ambiguous cases.</blockquote> An example is shown below for further illustration. The top panel shows a picture of a kitchen. And the bottom panels show the segmented picture that SAM generated, diving the picture into different parts (knives, coffee mugs, fruits, et al). Readers new to CV may miss the key point. After all, tracing the bounds of a given object is the first thing that toddlers learn to do. Believe it or not, that is exactly the key point. The SAM model is kind of doing – or more precisely, learning to do – things like a toddler. It had no idea what a coffee mug is (just like a toddler who has never drunk coffee yet), but it can correctly identify a coffee mug from a picture containing it. As Dr. Fan exclaimed, this implies that SAM has mastered the concept of “what is an object” and can generate masks for any object in any image or video, even if it has not seen that object during training. <figure class=”regular-img-figure” contenteditable=”false”><a href=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533065545337_origin.png” rel=”lightbox” data-width=”1776″ data-height=”999″ data-og-image-twitter_small_card=”true” data-og-image-twitter_large_card=”true” data-og-image-twitter_image_post=”true” data-og-image-msn=”true” data-og-image-facebook=”true” data-og-image-google_news=”true” data-og-image-google_plus=”true” data-og-image-linkdin=”true”><img src=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533065545337.png” alt=”META” contenteditable=”true” loading=”lazy”></a><figcaption>Source: Twitter.com</figcaption></figure><figure class=”regular-img-figure” contenteditable=”false”><a href=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533066097107_origin.png” rel=”lightbox” data-width=”1772″ data-height=”1003″ data-og-image-twitter_small_card=”true” data-og-image-twitter_large_card=”true” data-og-image-twitter_image_post=”true” data-og-image-msn=”true” data-og-image-facebook=”true” data-og-image-google_news=”true” data-og-image-google_plus=”true” data-og-image-linkdin=”true”><img src=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533066097107.png” alt=”META” contenteditable=”true” loading=”lazy”></a><figcaption>Source: Twitter.com</figcaption></figure> Furthermore, also like a toddler, SAM has demonstrated sufficient generality to make it suitable for various use cases without the need for additional training. Once a toddler learned how to identify the bounds of an object from a picture (say a bird), he/she does not need to relearn the same skill for another object, in CV/AI lingo, the so-called zero-shot transfer learning capability. And SAM has demonstrated such capability, making it ready to use out of the box for new image domains. And next, we will see the wilder implications of such AI and the impacts on META’s future growth and investment returns. <h2 class=”paywall-full-content invisible no-summary-bullets”>META’s AI catalysts</h2> Investors have been frustrated (and many probably are still being frustrated) by META’s initiatives in the AI and VR space because these things seem too remote and futuristic. While many people are still under the impression that AI and VR are sci-fi fantasies and we won’t be able to benefit from them in any foreseeable future, <a href=”https://futureoflife.org/” rel=”nofollow” title=”https://futureoflife.org/” target=”_blank”>leading AI and VR experts</a> are worried about the exact opposite. They are worried that the singularity point (the point that they begin to improve themselves) is approaching faster than we can be ready. Besides the SAM example above from META, in 2014, a Google team led by <a href=”https://kimschmidtsbrain.com/2015/10/29/artificial-intelligence-neural-networks-and-deep-learning/” rel=”nofollow” title=”https://kimschmidtsbrain.com/2015/10/29/artificial-intelligence-neural-networks-and-deep-learning/” target=”_blank”>Ilya Sutskever</a> also demonstrated an algorithm that can recognize and label images. For example, the algorithm created the following labels for the three photos below, “a group of young people playing a game of Frisbee”, “a person riding a motorcycle on a dirt road”, and “a herd of elephants walking across a dry grass field”. It looks so unimpressive on the surface again – something a 4-year-old can do. But the scary or impressive part is that the algorithm had no concept of frisbee, young man, motorcycle, or elephants, whatsoever. The algorithm, purely based on the colors in a bunch of pixels, somehow “learned” these concepts and correctly labeled these images. <figure class=”regular-img-figure” contenteditable=”false”><a href=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533053641286_origin.png” rel=”lightbox” data-width=”699″ data-height=”195″ data-og-image-twitter_small_card=”false” data-og-image-twitter_large_card=”false” data-og-image-twitter_image_post=”false” data-og-image-msn=”false” data-og-image-facebook=”false” data-og-image-google_news=”false” data-og-image-google_plus=”false” data-og-image-linkdin=”false”><img src=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533053641286.png” alt=”Show, Attend, and Tell: Neural Image Caption Generation with Visual Attention” contenteditable=”true” loading=”lazy”></a><figcaption>Source: Google’s TensorFlow Update</figcaption></figure> As a leader in social media and digital ads, META can begin to benefit from such AI capabilities very soon the way I see things. Most social media users (you included I am presuming) post way more imaging content than texts and numbers (which are traditionally what computers are so good at analyzing). The ability for computers to analyze images will open up a whole new regime for platforms like META. And I view the following to be possibilities that META can immediately capitalize on. <ul class=”paywall-full-content invisible no-summary-bullets”> <li> User Behavior Analysis: AI such as SAM can analyze user behavior data via images – not only those that they’ve posted, but also those that they’ve stared at for the longest time – to understand user preferences, interests, and engagement patterns, which can inform product development, user interface design, and platform improvements to better meet user needs.</li> <li> Automated Content Creation: This is only a logical next step once AI has learned how to read and analyze images. AI can assist in generating content, such as automated video editing, image synthesis, and text generation, which can help streamline content creation processes on the Meta platform and increase content production efficiency. And since it understands your behavior, the contents will be so interesting and entertaining for you.</li> <li> Content Moderation: AI can help the various Meta platforms in moderating user-generated content by automatically identifying and removing inappropriate or harmful content, such as hate speech, fake news, and spam. This can help ensure a safe and positive user experience on the platform while reducing the need for manual content moderation.</li> <li> Enhanced User Experience: All the above would lead to personalized and enhanced user experience on the various META apps, thereby increasing user engagement and retention.</li> <li> Virtual Reality (“VR”) and Augmented Reality (“AR”): And eventually, META, being a pioneer in VR and AR technologies, can leverage AI to enhance the capabilities of its VR and AR applications. For example, AI can be used to improve the accuracy of hand tracking, facial recognition, and object recognition in VR and AR experiences, leading to more realistic and immersive interactions. But again, the key point is that META does not have to wait till the VR or AR stage to benefit from its current efforts.</li> </ul> <h2 class=”paywall-full-content invisible no-summary-bullets”>Play the long game with META</h2> META does face immediate issues (more on this in the next section), so I can only recommend it to investors with a longer horizon (at least 3~5 years). With a timeframe long enough, the rate of return (“ROI”) on any given business is determined by a few simple fundamental factors. And the most important factors are ROCE (return on capital employed) and reinvestment rate (“RR”) because these two factors determine the sustainable earning growth rate. The growth rate is simply the product of these two factors (i.e., Long term growth rate = ROCE * RR). The chart below displays my analysis of META’s ROCE performance since 2011. You can see that its ROCE peaked in 2017 at around 90%, and then started to decline from there. Thus, investors’ concerns in recent years are for good reasons. However, META has maintained an average ROCE of 61% since 2018, even after the decline started. And this level of ROCE is on par with the overachieving FAAMG stocks group. <figure class=”regular-img-figure” contenteditable=”false”><a href=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533052103004_origin.png” rel=”lightbox” data-width=”1771″ data-height=”920″ data-og-image-twitter_small_card=”true” data-og-image-twitter_large_card=”true” data-og-image-twitter_image_post=”true” data-og-image-msn=”true” data-og-image-facebook=”true” data-og-image-google_news=”true” data-og-image-google_plus=”true” data-og-image-linkdin=”true”><img src=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533052103004.png” alt=”Chart, bar chart Description automatically generated” contenteditable=”true” loading=”lazy”></a><figcaption>Source: author and Seeking Alpha.</figcaption></figure> In the meantime, thanks to its existing cash-cow segments, META enjoys great capital allocation flexibility and can afford to play the long game. With a platform business model, the capital allocation picture is simple. It generates ample cash flow from organic operations but requires minimal maintenance CAPEX spending (on average 15% of operating cash flow in recent years as shown in the chart below). Hence, most of its CAPEX investments are for growth CAPEX, not maintenance CAPEX. Readers interested in a more detailed analysis to delineate its growth and maintenance CAPEX can take a look at one of my <a href=”https://seekingalpha.com/article/4517004-meta-platforms-true-pe-only-11x” title=”https://seekingalpha.com/article/4517004-meta-platforms-true-pe-only-11x” target=”_blank”>earlier articles</a>. <figure class=”regular-img-figure” contenteditable=”false”><a href=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533053986409_origin.png” rel=”lightbox” data-width=”1780″ data-height=”959″ data-og-image-twitter_small_card=”true” data-og-image-twitter_large_card=”true” data-og-image-twitter_image_post=”true” data-og-image-msn=”true” data-og-image-facebook=”true” data-og-image-google_news=”true” data-og-image-google_plus=”true” data-og-image-linkdin=”true”><img src=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533053986409.png” alt=”Chart, waterfall chart Description automatically generated” contenteditable=”true” loading=”lazy”></a><figcaption>Source: author and Seeking Alpha data.</figcaption></figure> To wit, my estimate of its growth RR is in the order of 20%. With an average ROCE of ~60% and 20% RR, META can grow its earnings organically and sustainably by 12% per year (60% ROCE * 20% RR = 12% growth rates. And in the long term, if we think like business owners (and actually own the stock long enough), our ROI will be simply the sum of A) the owner’s earning yield and B) the 12% organic growth rate that we just estimated. At its current price levels, META is priced around P/E multiples of ~22x on FW and GAAP basis, translating into an earning yield of about 5%. However, its owners’ earnings yield is higher because, as just aforementioned, a large part of its CAPEX expenditure is for growth and should not be considered an expense (details are again in my earlier article). With the growth CAPEX adjusted, my estimated owners’ earnings yield is a little above 5% (5.26% to be precise, corresponding to a P/E of 19x on owners’ earnings). As such, my projection for its long-term return under current conditions would be an attractive ~17% per annum (12% growth plus about 5% owner’s earnings yield). Even in the case that the RR is halved (say due to a lack of high-margin areas to reinvest) the annual total return would still be in the mid-teens as shown by the star symbol in the chart below. The symbol illustrates a potential annual return of ~12%, consisting of 5% owners’ earnings yield and a 6% growth rate (60% ROCE * 10% RR). <figure class=”regular-img-figure” contenteditable=”false”><a href=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-1681653305367895_origin.png” rel=”lightbox” data-width=”1798″ data-height=”743″ data-og-image-twitter_small_card=”true” data-og-image-twitter_large_card=”true” data-og-image-twitter_image_post=”true” data-og-image-msn=”true” data-og-image-facebook=”true” data-og-image-google_news=”true” data-og-image-google_plus=”true” data-og-image-linkdin=”true”><img src=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-1681653305367895.png” alt=”Table Description automatically generated” contenteditable=”true” loading=”lazy”></a><figcaption>Source: Seeking Alpha data.</figcaption></figure><figure class=”regular-img-figure” contenteditable=”false”><a href=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533056340027_origin.png” rel=”lightbox” data-width=”1743″ data-height=”1098″ data-og-image-twitter_small_card=”true” data-og-image-twitter_large_card=”true” data-og-image-twitter_image_post=”true” data-og-image-msn=”true” data-og-image-facebook=”true” data-og-image-google_news=”true” data-og-image-google_plus=”true” data-og-image-linkdin=”true”><img src=”https://static.seekingalpha.com/uploads/2023/4/16/48844541-16816533056340027.png” alt=”Chart, line chart Description automatically generated” contenteditable=”true” loading=”lazy”></a><figcaption>Source: author and Seeking Alpha.</figcaption></figure> <h2 class=”paywall-full-content invisible no-summary-bullets”>Risks and winner-takes-all potential</h2> As mentioned earlier, META does face several immediate issues. META has been facing revenue headwinds lately due to a multitude of challenges. It has been experiencing slower user growth and fluctuating daily/monthly active users times due to competition from other social media platforms such as YouTube and TikTok. Operating expenses rose substantially (by almost 20% in 2022 compared to 2021) due to macroeconomic conditions (e.g., high inflation and labor shortage) and also due to increased spending on research and development. To make things worse, the whole industry sector is also experiencing a slowdown in digital advertising. As a result, META has to execute quite dramatic cost control efforts, scrutinizing closely its spending. <div class=”before_last_paragraph-piano-placeholder paywall-full-content invisible no-summary-bullets”></div> Although my view is that the above challenges are temporary and not structural, for investors who have a longer investment timeframe and can tolerate these risks, I view the odds to be favorable. After all, the AI potential is too attractive and substantial. Simply put, it can create a winner-take-all situation. In the old economy, we all routinely search for and settle with the 10th best accountant in town to file our taxes just because his/her office is closer or office hours more convenient for us. But in the age of TurboTax, most of us never even bother to look for ANY alternative anymore. The winner-take-all potential is exponentially more potent with the combination of AI and the vast amount of data that META has already amassed.

Analyst’s Disclosure: I/we have no stock, option or similar derivative position in any of the companies mentioned, and no plans to initiate any such positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article. Seeking Alpha’s Disclosure: Past performance is no guarantee of future results. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. Any views or opinions expressed above may not reflect those of Seeking Alpha as a whole. Seeking Alpha is not a licensed securities dealer, broker or US investment adviser or investment bank. Our analysts are third party authors that include both professional investors and individual investors who may not be licensed or certified by any institute or regulatory body.

<!–StartFragment–>As you can tell, our core style is to provide actionable and unambiguous ideas from our independent research. If your share this investment style, check out <a href=”https://seekingalpha.com/checkout?service_id=mp_1400″>Envision Early Retirement.</a> It provides at least 1x in-depth articles per week on such ideas.<!–EndFragment–>We have helped our members not only to beat S&P 500 but also avoid heavy drawdowns despite the extreme volatilities in BOTH the equity AND bond market.Join for a <a href=”https://email.seekingalpha.com/checkout/track?type=click&mailingid=email&messageid=marketplace_follow_up&serial=kygbnacx4h5iunw88xh06gyg&emailid=linmaa%40gmail.com&userid=&extra=&&&3000&&&https://seekingalpha.com/account/email_auth?auth_param=1eijnt:1gu6du6:da4a7685b918857ea382f1d385504139&ref=%2Fcheckout%3Fmail_subject%3D%26service_id%3Dmp_1400%26utm_medium%3Demail%26utm_source%3Dseeking_alpha?source=retargeting_follow_up”>100% Risk-Free</a> trial and see if our proven method can help you too.<img src=”https://static.seekingalpha.com/uploads/2022/11/12/48844541-16682787177402456.png”>

Tags: META

Meta Platforms’ ChatGPT-3 Moment

Leave a Reply Cancel reply