Kicking off our hunt for the next AI winners
Markets: Where to invest and where not to invest in AI now
Everyone and their cousin are throwing out estimates of how AI is going to have a huge positive impact on the global economy. Just this week, we got IDC telling us that through 2030, AI will boost the economy by $20T, which equates to a 460% ROI. McKinsey expects AI to deliver a $25.6T benefit to global GDP over the next 20 years, including with a $7.9T annual benefit in peak years. UBS predicts that the AI industry will grow to annual revenues of about $450B by 2027, a 72% CAGR from 2022.
During the recent Q2 earnings season, more than 40% of companies mentioned AI as a possible game changer for their business. CEOs from ServiceNow (NOW), Salesforce (CRM), AirBNB (ABNB)…to name a few…have been talking their book…discussing how AI is the next best thing to sliced bread and how they are taking market share and squashing the competition.
In the Beachman community, we prefer to look past such hype and big talk. We prefer to look at actual numbers and trends. We pay more attention to what exactly is happening in the AI world, who is spending on AI, who is using AI, who is making money on AI.
Earlier this year, we launched our AI tracker which monitors about 25-30 stocks in terms of their AI-specific products, revenues, forward estimates and competition. We continue to update our tracker with new information from earnings reports, company updates and industry developments.
It is time to look past NVDA’s GPUs to find the next set of investment opportunities that could work in 2025. Now might be the time to buy those stocks.
Follow Beachman
On the daily, you can find Beachman in two places…
Beachman’s substack chat line here
Beachman’s Threads feed here
Please check out the must-reads listed on the About page and the Roadmap page. Also, take note that I have stopped using my Twitter account.
Table of contents
Today, we will kickoff with a few important baseline topics:
What is AI and how does it work?
Key questions in the AI space
High level timeline
As is my objective with all my writing, these posts will be actionable. I don’t write long, boring white papers that will put you to sleep. I keep it short and sweet, super focused on what you need to know as an investor. The goal of these writings is to boil it down to important data points, emerging theses, investing opportunities and in many cases risks and potholes to avoid.
I will highlight and call out the actionable take-aways as we go along.
Now, let’s kick off our quest for the emerging AI winners…
What is AI and how does it work?
For a couple of summers now, I have published a very simple, easy-to-understand primer on artificial intelligence and how it works. Here is a link to the most recent post.
If you have not read it, please take a few minutes to do so.
I promise you that it is a quick read and it will enhance your understanding of the topic at hand.
As an investor, this is important to not get intimidated or confused or overwhelmed with all the jargon and highty-flighty commentary.
The post linked above will help you understand how AI works. It includes several everyday examples of AI in our world today and it provides a simple stratification of the types of technologies used to implement AI solutions.
Recently, Jensen Huang, CEO of Nvidia (NVDA) told us about two major trends that are shaping the future of computing:
A shift from CPU-based data centers to faster, GPU-accelerated computing.
Generative AI, which enables computers to learn from data instead of being programmed.
“The days of every line of code being written by software engineers are completely over,” Huang said. For example, Tesla’s self driving software was not coded by software programmers. It is an AI model that processes millions of images from cameras mounted on their cars and then makes real time driving decisions based on what it has learned from past human drivers’ behaviors.
Key questions in the AI space
As we sift through the mountains of information and opinions, I find it particularly useful to jot down important questions that I need to investigate and answer before I can decide where to invest.
When I think about AI, the key questions that come up over and over again are:
Who is actually making money in AI now or in the near future?
When will the current AI capex boom turn lower?
Who will be the next beneficiary of hyperscaler and enterprise AI spend?
Now these questions, in turn, raise further, more granular questions that we will explore in this series:
Are business using AI today? To do what? How much benefit are they getting?
Are consumers using AI today? To do what? Are they willing to pay for it?
Which solution providers are seeing new, incremental business due AI?
This last question is particularly important. We want to distinguish between a company that is launching new, AI-specific products versus those who are just redesigning their existing products to use AI…something they would have to do anyways in order to compete in the marketplace.
Take a company like SnowFlake (SNOW). So far, AI has not brought in any meaningful, new revenues. They continue to provide data management solutions to thousands of customers and they are supporting various AI pilots with their current platform. SNOW is also integrating AI into their tools, which they would have to do anyways in order to stay competitive. It remains to be seen if SNOW will become a true AI winner or stay as a data solutions company. The same applies to Confluent (CFLT), Elastic (ESTC), Pure Storage (PSTG), Cloudflare (NET) and many others.
Moving on, here are some more granular questions to explore:
Is AI going to disrupt traditional software platforms and applications?
How will data be used, stored, managed and shared in the future?
Can companies (eg. Reddit, New York Times) make money selling data to AI LLMs?
Will AI data centers need a lot more energy and power in the future?
Will AI robots take over the world?
OK, that last one was in jest…partially. We will spend some time talking about AI and robotics and how they are likely to come together in the future. However, just in case you were wondering…no, I do not think that robots will take over the world. No Matrix like world in our future.
High level timeline
The AI adoption curve is playing out in phases and in layers across multiple use cases and verticals. Many of these are progressing in tandem with each other…in parallel.
As an investor, this timeline is particularly useful to map out where and when to invest in certain aspects of AI. Currently, I am overweight in the semiconductor space, however, at some point, I will transition those funds to other AI verticals that are about to ramp up.
Infrastructure buildout
This is a massive effort currently in progress, led by the largest mega cap companies, the hyperscalers, the cloud service providers - AMZN, GOOG, MSFT, META, ORCL and others. Collectively, in 2024 they are spending more than $210B +45% yoy on upgrading their data centers to the latest and greatest AI infrastructure - GPUs, servers, networking, storage and power.
There could be a slowdown in the infrastructure buildout in 2025 as AI goes through the trough of disillusionment.
Investors are getting impatient and asking more questions about the ROI on these expensive upgrades. Nevertheless, we should remember that these mega caps deferred their CPU-based data center upgrades in 2023 in anticipation of directly moving onto the next gen AI GPUs instead. So the upgrades will continue, even if the pace drops a few notches.
In 2025, AI stocks could take breather if this happens.
R&D, proofs of concept (POC), trial runs
Every business leader is asking themselves and their teams about whether AI can improve their operations, reduce their expenses and boost their sales. You have to if you want to compete in your market, because you can bet that your competitors are doing the same.
As stated above, in the recent Q2 earnings season, more than 40% of companies mentioned AI as a possible game changer for their business. They have created special project teams that are fervently investigating AI solutions that might make sense for their business. POCs are happening all over the place, likely in most departments. 90% of these efforts will fail, however the 10% that succeed could add to the top line and the bottom line.
CFOs typically afford such strategic projects a 2-year window to deliver an ROI, to show concrete results. 2024 is the first such year.
Teams will be given 2025 as the 2nd year of latitude.
Then hard decisions will be made about whether to cut bait or invest more in AI.
We will dive deeper into this work in a subsequent post.
Consumer AI ramps up
Consumers like you and me will get our first taste of everyday AI when Apple (AAPL) launches their Apple Intelligence features in a few weeks.
Apple Intelligence is going to test AI from several angles - consumer adoption and preferences, AI on the edge, privacy, multiple AI providers on the same device, small language models (SLMs), voice-driven AI, multi-modal AI etc.
We will explore consumer AI in more detail in a subsequent post. Lots to watch closely and maybe even get excited...
AI on the edge
Imo, consumer AI is the most important type of AI on the edge. We use multiple, smart devices on a daily basis. As consumer AI takes off, these devices will have to learn to be generative-AI-ready even without a constant, high speed data connection or large amounts of on-device storage and memory.
Case in point, self driving vehicles. They will have to use AI proficiently and autonomously to take us from point A to point B in a safe and legal manner. Most drivers will not have a high speed internet connection in the car. Therefore, the self-driving software will have to complete its inference-based, data processing on the fly, on the edge using the onboard GPUs, memory, storage and its prior AI training.
Industrial AI and robotics
Robots have been used in the workplace for decades. They were first used in automobile manufacturing in the 1960s and in surgery in the 1980s. For our purposes here, we are referring to the application of artificial intelligence to industrial processes - autonomous machines used to perform tasks that typically require human intelligence, such as learning, reasoning, and problem-solving, within an industrial context. Industrial AI is ramping up in manufacturing, supply chain management, energy, maintenance, and quality control. By leveraging AI, industries are looking to automate tasks, improve efficiency, reduce costs and gain a competitive edge.
While software, data and GPUs typically represent the knowledge aspect of AI, industrial AI refers to the big, bulky, machines that make and move stuff in the physical world. This is a $30-40T portion of the global economy, growing about 15% yoy. Suffice to say, it’s no chump change.
Sovereign AI
Of all the AI promises, I am most skeptical of this one…for several reasons:
Governments make decisions based on politics and local cultural norms. AI is one of the biggest anti-theses to these considerations.
Administrations move slowly, they take time to make decisions and allocate budgets, especially to cutting edge projects.
Governments are more stringent in their approvals, pricing, acceptance criteria and payouts leading to elongated go-live timelines.
Granted, there will be a few forward-thinking and even autocratic countries that will lead the way in sovereign AI. Singapore, Japan and Saudi Arabia are stepping up as early adopters. But the vast majority of nations will tip toe into this space, likely taking a couple of decades for broader adoption.
Conclusion
Now that we have a baseline understanding of AI and its high level timeline, we will start exploring the underlying areas in the AI arena. How are businesses using AI? Where is AI impacting consumers’ lives? AI GPUs, LLMs, software, data management, storage, networking, power, AI on the edge and robotics.
Stay tuned…
"But the vast majority of nations will tip toe into this space, likely taking a couple of decades for broader adoption."
I think governments may behave more like multinationals on this point, and we may see a flurry of soveriegn investment in the next couple of years -- not a couple of decades, so as not to miss out and get 'hopelessly behind' (FOMO).