Artificial Intelligence: How We Got Here, Where We’re Going, and How to Invest

Introduction

“In the last 40 years, nothing has been this big. It’s bigger than PC, it’s bigger than mobile, and it’s gonna be bigger than the internet, by far.”¹

Nvidia CEO Jensen Huang had just taken the stage at Microsoft’s annual Ignite conference for developers and IT professionals. The ‘it’ he’s so clearly excited about is artificial intelligence (A.I.).

He’s not alone.

The potential for A.I. to transform our world captivates the imagination. And it has launched a spending spree of several billion dollars as the biggest technology companies invest in the A.I. revolution.

Investors have taken note. It would be difficult not to, with the parabolic stock price rises of many of the so-called Magnificent Seven stocks – Alphabet, Amazon, Apple, Meta Platforms, Microsoft, Nvidia and Tesla – in 2023 and 2024. All of the aforementioned stocks have some connection to A.I., whether through interactive chatbot A.I. interfaces, providing the infrastructure that enables A.I., or possessing the valuable data required to train large language models that power A.I. content generation, A.I. virtual assistants, and other A.I. fields.

What do we mean by A.I.? It is the ability of machines to perform tasks that would normally require human intelligence. This broad definition covers a lot of ground. A.I. has the potential to change how drugs are developed. How music is composed. How software is written. How cars drive (themselves).

The potential to disrupt industries and enhance productivity clearly has investment implications. But we are early. It’s difficult to know if A.I. will be the game-changer that many believe, and if it is, what exactly that looks like. Some early winners have emerged. It remains to be seen if their competitive advantages are sustainable.

At SAM, many of our clients and prospective clients have expressed an interest in getting exposure to A.I. in their investment portfolios. But only a few “pure play” A.I. stocks exist, and those that qualify tend to come with a hefty price tag.

In this paper we will explore how A.I. seemingly came out of nowhere to be the most talked about trend in the investment world. We will also share our thoughts on A.I. and how we have chosen to invest in this space.

An overnight success 70 years in the making

The rise of A.I. seemingly came from nowhere. Most people credit the A.I. boom to the launch of ChatGPT on November 30, 2022. This chatbot, developed by OpenAI (partially owned by Microsoft), brought A.I.-powered capabilities to the general public. With no particular skills prerequisite, users could begin to see the power of A.I. as they asked questions and generated content. It became the fastest-growing consumer application in history, reaching 100 million monthly active users by January 2023.²

What seemed to be an overnight sensation was in fact rooted in a paper written several decades earlier. Computer science pioneer Alan Turing introduced the concept of A.I. in 1950, when he wrote about the test of a machine’s ability to exhibit intelligent behavior. The actual term “artificial intelligence” was coined in 1956 by computer scientist John McCarthy.

In the 1950s, a foundational technique for achieving artificial intelligence emerged: machine learning. This technique involves training algorithms to parse, learn from, abstract and utilize data. Generally, the more data that is fed to the algorithms, the more accurate the inferences become.

In the decades that followed, while academics pushed forward with machine learning, A.I. was rarely on the consciousness of the public. But that began to change as A.I. found uses in real-world applications. In 1979, an A.I.-driven computer program beat the world champion at backgammon. This marked the first time a world champion in a recognized intellectual activity was defeated by a computer. It would not be the last.

In 1997, then world-chess champion Garry Kasparov was defeated by an IBM supercomputer called Deep Blue. And in 2011, another IBM computer system called Watson successfully competed on the Jeopardy! game show. Still, while deeply impressive, it wasn’t until the launch of ChatGPT that the public began to imagine ways that A.I. could impact their daily lives.

Why now?

With decades of academic research and impressive feats against human champions in skill-based competition, it may seem a wonder that A.I. did not take off sooner. It is due to the convergence of three interrelated trends that A.I. has recently propelled itself into the mainstream.

An explosion of available data

Massive amounts of data are required to train machine learning algorithms. Today such data is readily available thanks to our internet-connected devices. And the amount of it is steadily growing every year. In fact, estimates show a 13x increase in the amount of data created in 2023 (120 zettabytes) compared to a decade ago (9 zettabytes).³

Improvements in high-performance computing

Today’s powerful graphics processing unit (GPU) chips enable much faster computations than were possible in the past. GPUs can compute in parallel rather than sequentially, which makes them much more efficient for machine learning.

Breakthroughs in machine learning

Machine learning research has produced new algorithms that can process massive amounts of data and create remarkably accurate interpretations. A powerful machine learning technique called deep learning is particularly noteworthy. It relies on algorithms called deep neural networks which are loosely inspired by the circuitry of the human brain.

Economic Potential

Anyone who has tinkered with ChatGPT is likely to tell you that there is strong potential for productivity gains. This has been echoed by very different companies in very different industries. Coca-Cola (KO) is using A.I. to improve their marketing, supply chain, and even generate new soft drink flavors. Walmart (WMT) has a generative A.I. search in its app and is also employing A.I. to optimize inventories based on anticipated demand. Lockheed Martin (LMT) says it has incorporated A.I. into more than 1,000 projects and programs, in addition to using it to improve their own business operations efficiency.

Similar to the way the internet changed the world in the early 2000s, A.I. is expected to significantly disrupt the way that many industries operate. This disruption will come in many forms across the business landscape. For example, the applications for A.I. in healthcare are numerous. They include having an A.I. program that can read diagnostic imaging better than the average radiologist, or predictive analytics that can indicate correlations between diseases or treatments, and maybe most notably A.I. could be used to drastically speed up and increase the success rate for pharmaceutical development.

Similarly, the applications for A.I. in the world of finance are wide-ranging. A.I. could be used to take financial data and provide predictive analytics, leading to better lending practices and enhancing people’s ability to access financing. It can also be used by banks and other financial institutions to identify patterns in financial data that could be used to highlight and stop fraud or other nefarious financial activities.

Marketing is another area in which A.I. has broad applicability. Using A.I. models to help enhance targeted advertisements is something that’s already occurring. Generative A.I. tools are also already in use for creating graphic designs and marketing images without human involvement, drastically reducing the cost of creating content and speeding up the deployment of marketing campaigns.

Efficiency is great, but could A.I.’s disruption outweigh the benefits? IBM turned heads in 2023 when it announced it halted hiring for 7,800 jobs that could be replaced by A.I.⁴ There will undoubtedly be some jobs that are no longer needed due to increased efficiency from A.I. However, we believe the economic impact is likely to be more than offset by productivity gains.

Bad actor risk

A.I., as with any novel technology, comes with the threat of misuse by bad actors. And A.I. can and is being used for a wide array of nefarious purposes. The simplest way to disaggregate A.I. bad actors is to split them into two buckets: rogue groups or individuals, and state actors (or state-sponsored actors).

The former group is certainly less threatening on a global scale, but the potential for serious harm is very real. More innocuous examples include students using A.I. to cheat on school papers, and individuals using A.I. to create “deepfakes” (AI-generated content of a person such as their voice or image that is hard to distinguish from reality). In early 2024, a first-of-its-kind A.I. heist in Hong Kong was reported when an engineering firm suffered a loss of over $25 million after an unauthorized funds transfer.⁵ The scammer reportedly used A.I. tools to create convincing deepfake video and audio simulations in a multi-person video conference where all participants (except the victim) were fabricated images of real individuals.

Such applications of A.I. will not threaten global stability, but can certainly be extremely harmful to individuals and companies that are targeted.

The state actor group is a much more serious geopolitical threat. The use of A.I. in relation to cybersecurity and hacking is a new and profound risk to both nations and their populations. In the near-term, A.I. is already being used across the globe to spread misinformation and disinformation. Additionally, A.I. applications like ChatGPT can be tricked into providing inaccurate information in a way that dubiously makes it seem like an authoritative answer. But even more concerning are the ways in which foreign adversaries are using A.I. in an offensive manner in the context of hacking and cyber warfare.

OpenAI announced earlier this year that it had disrupted five state-affiliated actors using its services: two from China, and one each from Iran, North Korea, and Russia.⁶ In general, these groups were using OpenAI to refine phishing schemes for better effectiveness, pulling information on intelligence agencies as well as satellite and radar imaging, and in support of code development. And these are just the applications we know about because they used a domestic company’s services. Foreign adversaries are no doubt utilizing small armies of their homegrown computer science and A.I. experts to build out their offensive cyber capabilities using A.I. in ways that are far from the public’s vantage point. And as A.I. improves, it becomes exponentially more dangerous. A paper from Cornell found that “GPT-4 can hack 73% of the websites we constructed compared to 7% for GPT-3.5, and 0% for all open source models.⁷” The use of A.I. by adversarial actors is a threat that not only isn’t going to go away, but conversely will become more dangerous over time.

SAM’s stance on A.I.

Despite the potential employment impacts and bad actor risks, we at SAM are bullish on A.I. in the aggregate. The potential for productivity gains is well understood, and the potential for public companies to generate significant value from the development and/or use of A.I. is profound. Simply put, A.I. isn’t just coming, it’s already begun to arrive. And importantly for investors, we believe that profitable opportunities exist in this exciting new space.

How to invest

This leads us to the main question at hand: how and where should one be investing in A.I.? The answer, as usual when it comes to investing, is that it depends. This is because there are myriad ways to play the A.I. wave from an investing perspective. These include the major players, the ‘picks and shovels’ technology that facilitates A.I., tools for monitoring and managing A.I., owners of valuable data assets, and even the companies that will help fulfill the power needs that A.I. brings about.

Big Tech

As you might expect, Big Tech firms have been pouring virtually unending amounts of money and man hours into A.I. As exciting as A.I. is for these companies from a new business perspective, in many ways it’s also a major threat to their existing businesses.

Alphabet/Google (GOOGL, GOOG) is a multitrillion-dollar company and one of the most recognizable companies in the world. But Google’s marquee web search application is undeniably at risk of losing market share (and thus marketing dollars) to generative A.I. Many people may simply ask ChatGPT or other large language models (“LLMs”) to find the content they’re looking for, and even task the A.I. with summarizing the search results from multiple sites. In response, Google has invested significantly in developing its own generative A.I. to work in concert with its search engine. Google’s A.I. can also be used by businesses to build and deploy customer service assistants (sometimes called “chatbots”), reducing expenses for companies because they no longer need an army of customer service reps in a call center.

Not to be outdone, Microsoft (MSFT) invested $1 billion in 2019 into the company behind ChatGPT, OpenAI. Microsoft is a huge company, and so the 2019 investment – before A.I. was at the forefront of the market’s consciousness – flew somewhat under the radar. Microsoft’s follow-on investment of $10 billion in early 2023 did not, as it was one of the biggest stories in financial markets at the time. Lesser known is that Microsoft’s 2019 investment was allegedly a defensive response to Google’s A.I. ambitions⁸, as was revealed during the U.S. Department of Justice’s antitrust case against Google.

Microsoft CTO Kevin Scott sent an email to Microsoft CEO Satya Nadella and co-founder Bill Gates discussing OpenAI and Google, noting that Microsoft was “multiple years behind the competition in terms of machine learning scale.” Of course, in retrospect the OpenAI deal was an incredibly shrewd move. And now Microsoft has a diverse suite of A.I. offerings, including enterprise A.I. platform Azure OpenAI and multiple Microsoft Copilots that integrate into Microsoft applications like MS Office and GitHub.

Picks and Shovels

The saying often goes something like this: during the gold rush, it wasn’t the miners that got rich, but rather those who sold them the picks and shovels. What this means is that when there’s some new trend sweeping the markets, oftentimes the companies that provide the needed equipment or technology for the new trend are better bets than companies that are gambling on the new trend.

In A.I., that conversation has to start with Nvidia (NVDA). Nvidia builds graphic processing units, which are powerful semiconductors that facilitate the data centers that underpin A.I. In fact, Nvidia’s GPUs are so much better than its competition that Nvidia controls roughly 80% of the global GPU chip market⁹. Simply put, the future of A.I., at least at this moment in time, is directly reliant on Nvidia’s ability to keep producing its best-in-class GPUs. Over time, this should lead to soaring revenues and profits.

Proprietary Data Owners

When it comes to A.I., a model is only as good as it’s trained to be. This means that even if someone is the best A.I. scientist in the world, if they don’t have vast troves of valuable data to train their model, it has a very limited chance of commercial success. In practice, this means that over the last 10-15 years, a company’s proprietary data has actually become a significant and valuable asset.

For instance, Verisk (VRSK) is a world class software business in the insurance industry. Verisk, as an independent third party, collects data from the world’s largest insurance firms. It then analyzes all the data in the aggregate and sells mission-critical risk assessment data and analytics back to the insurers in what amounts to a “win-win” relationship. Verisk earns revenue via lucrative recurring payments from the insurers, and the insurers get better insights from their data because Verisk is able to aggregate all the insurers’ datasets. So instead of each insurance company only being able to analyze their own data, Verisk’s place in the ecosystem allows for the data analysis to be done across many insurers’ datasets, which provides better and more productive results.

Data and Software Management Tools

It may seem obvious, but with the explosion of the internet and now A.I. globally, companies have more software systems and deployments than ever before. And managing a company’s databases as well as their deployments of A.I. and LLMs is a crucial piece of the puzzle. That’s where companies like Snowflake (SNOW) and DataDog (DDOG) come into play.

Snowflake’s business model is centered around providing one platform for its customers to access, manage, analyze, and manipulate all of that customer’s data that they likely have siloed in dozens of databases, apps, files, and third-party warehouses. Snowflake calls this the “Data Cloud” and it’s incredibly powerful because Snowflake is constantly expanding its partner ecosystem to work with all the biggest data, cloud, and analytics providers to be a true one-stop shop to access and analyze nearly all your data, structured and unstructured. As noted above, your A.I. is only going to be as good as your data, so Snowflake’s value proposition is that it helps its customers get more out of their data and A.I.

DataDog has historically been the leading cloud-native software provider for monitoring networks, infrastructure, development and cloud-based applications. Think of DataDog as a real-time performance watchdog for all of a company’s computing units, from physical hardware and servers, to its cloud network, to its own proprietary software (and apps and containers) running across all sorts of devices. So in many cases, DataDog is already engrained in many major companies’ tech stacks. With the arrival of A.I. as a widespread phenomenon, DataDog is well positioned to now help companies manage their A.I. deployments the same way they help companies manage their data, hardware, and software. As you might expect, DataDog was one of the first companies to bring an A.I./LLM monitoring tool to market. This tool can help with things like tracking performance and speeding up troubleshooting. Additionally, DataDog was one of the first companies to begin to quantify the financial impact from “next-gen A.I. customers.” As of its most recent quarterly earnings, around 3.5% of DataDog’s business comes from these customers, and that figure should be expected to rise materially over time.¹⁰

Power Supply

A research paper from Goldman Sachs recently noted that on average “a ChatGPT query needs nearly 10 times as much electricity to process as a Google search.”¹¹ And as A.I. continues to boom, the data centers that run A.I. programs will themselves need significantly more electrical power than is currently needed today. In fact, the International Energy Agency expects roughly one-third of additional demand for electricity in the U.S. to come from data centers over the next three years.¹²

This power supply is likely going to require a mix of both fossil fuels and renewable sources to meet the demand. For the former, companies like the Tortoise Midstream Energy Fund (NTG) present an interesting way to play the A.I./data center power theme. NTG is a closed-end fund that generally invests in midstream companies that service the oil and gas industry. These pipelines get paid either based on the volume that they carry, or may simply receive a fee for use (like a toll bridge). With nearly 70% of NTG’s portfolio consisting of natural gas infrastructure, another 24% in natural gas liquids infrastructure, and natural gas being the largest source of U.S. electricity generation, NTG is heavily levered to the growth of U.S. power and should be poised to benefit from the A.I. and data center tailwinds for years to come.¹³ ¹⁴

Most companies will become “A.I. companies”

The dawn of the internet was a sea change in how data and content get distributed. For instance, the Wall Street Journal existed long before the internet did, and of course remains engrained in the world today. However, the internet allowed the articles printed in the WSJ to be accessed by anyone with a computer. The substance, the articles themselves, remain virtually the same today as they were in the 1980s.

A.I. represents a different sea change, one in which the production of content is being altered. In keeping with the same example, a WSJ article can now (theoretically) be written and produced by A.I. So, whereas the internet allowed for the same content to be distributed in ways the world had never seen, A.I. will allow for the same content to be produced in ways the world has never seen.

Just as most companies over the last ~25 years have evolved to being “internet companies” in ways they never would have labeled themselves previously, over the coming years companies will evolve into being “A.I. companies” in ways that would have seemed far-fetched not long ago. Whether that comes in the form of non-A.I.-related companies building their own A.I. (similar to how Amazon went from just selling books online to building itself into the multifaceted internet behemoth it is today), or in the form of companies harnessing A.I. to do things better or more efficiently (such as a pharmaceutical drug developer using a third-party AI model to pick and choose what drug compounds will work best in the clinic instead of having to test them all in practice), rest assured that A.I. will soon be a pervasive thread weaving throughout the entire global market.

How should you invest in A.I.?

It ultimately depends on your financial goals. That is, why are you investing in the first place? What are you ultimately trying to accomplish? At SAM, we have several strategies that are developed to meet different goals. We incorporate A.I. themes into many of these portfolios. Exactly how we do that – which A.I. stocks we own and to what degree – depends on how the investment fits into the larger picture of achieving your goals. Part of the process of becoming a SAM client is to work with a tenured Wealth Manager to identify your goals and which combination of strategies is optimal for you.