Machines will soon perform the bulk of digital work, assume the control of vehicles and perhaps even gain sentience. There will be robots of all shapes and sizes to help with daily life, and we can realistically hope for a step-change acceleration in the pace of scientific discovery. Along the way, we might experience periods of huge job displacement, consequential accidents caused by AI mistakes, or even deliberate harm caused by misaligned agents. We expect the next several years to be among the most exciting (and dangerous) that humanity has ever experienced, and the implications for financial markets to be unprecedented.
Gradient Ascent's central premise is that AI will transform the global economy, but we do not make any claim to know precisely how and when this will happen. When looking to the future it is crucial to differentiate between progress within a given paradigm, which can be predicted with a high degree of confidence, and trying to predict the timing of the next paradigm shift, which is incredibly difficult.
We believe speculation beyond the foreseeable future is unnecessary for successful investing, because the technology is sufficiently advanced to transform the global economy within the current paradigm. In software engineering, the best models are already capable of competently and consistently completing tasks of several hours, making its use invaluable. This level of competency will quickly spread into many fields of task-oriented knowledge work. In short, the goal of this essay is not to convince the reader that the Age of AI is coming, but to discuss the implications for investing, given it's already here.
Industry Insights
Our value-add is to translate industry insights from the frontier of AI development into actionable opinions about financial assets. We combine these insights with our knowledge of financial markets and select affected companies for deep-dive fundamental analysis, to isolate the best investment opportunities.
The first step in this process is to make sure we are extremely well informed about the current frontier, as well as promising developments in research. For the frontier, we start with the simple things – making sure we are using the models ourselves, following releases and the public reactions to them, as well as comparing against the benchmarks. We then form our own views on the technology and make inferences about economic implications of the latest developments.
This approach can get you a lot of the way, there is a lot of great information publicly available or behind relatively modest paywalls. However, we are going a step further. We want to have our theories probed and questioned by industry experts, and so, we are forming a Council of Experts composed of the Gradient Ascent team and AI researchers who are active in the field. The Council will meet regularly to discuss the latest developments at the frontier and to get their feedback on our interpretation of the news. Meanwhile we will provide them with a detailed market update and how asset prices have reacted to the latest developments as they became public.
In a world with so much information available, it is the synthesis of this information that really differentiates. What we are trying to create is a forum for discussion and good faith argument, between individuals with common interests but different backgrounds and expertise.
Identifying Tailwinds
Once our views on the technology have been formed and vetted by this panel of experts, we turn our attention to implications for asset pricing. As the global economy is transformed, every stock in the world will need to be repriced. Many existing stocks will benefit from extreme tailwinds, new sectors will be born while others will be faced with extinction. Recognising these impacts and their severity will be a huge part of investing well in the age of AI. We have already started to see the broadening impacts of AI disruption on a diverse array of sectors, such as SaaS, computer game development and insurance. To date, the incumbent companies in affected sectors have been negatively impacted. This trend will continue as the disruption widens, but I expect there to be some sectors where AI will actually be beneficial for incumbents. Those sectors with high moats, such as regulatory capture, and high white collar wage bills could gain enormously, achieving higher productivity for much lower costs, without necessarily having to pass all efficiency on to their customers.
The sectors which have benefitted most significantly so far have been those associated with the massive infrastructure required by LLMs' calculation-heavy approach. In contrast to the software boom, which has dominated investing for the previous two decades, the AI boom is capital intensive. In the age of software, finding and refining a truly disruptive or innovative solution was almost everything; once you had a concept that people liked, scaling globally was almost the easy part because dating apps, and even streaming services, required less physical infrastructure than the solutions they replaced. This will not be true of the AI era. LLMs benefit from scale, making them computationally heavy to use. This means that we need datacentres, lots of datacentres filled with very expensive custom equipment, and the electrical infrastructure required to power them.
Up to and including 2026 the bulk of this very expensive build-out has been funded by several of the massive tech companies (Google, Amazon, Meta and Microsoft). They are uniquely placed, being both close to the technology and having access to hundreds of billions of dollars in profits from their existing activities. Since 2023 they have been investing an increasing amount of these profits into AI datacentres. However, their capacity to keep increasing investment in this way will slow down significantly in 2027 and stop altogether in 2028. Even at the current rate of growth, we can't build datacentres fast enough to keep up with demand. See this blogpost for a more detailed discussion, where we conclude that capital will become the bottleneck unless the sentiment gap between AI insiders and the investment community closes significantly over the coming year or two.
The market seems to already be pricing this in to some extent, many of the stocks associated with the infra build-out have stalled in the last 6-12 months, while their earnings have continued to experience rapid growth; this is the market's way of telling us that it is concerned about future earnings. Nvidia is the prime example, both in terms of its size and the magnitude of the discount currently being applied, it ideally illustrates the point. Based on some reasonable assumptions the market is implying much lower growth than the company is guiding or analysts predict for 2027 (2026 earnings are mostly calculable to a tight range given their products are ordered months in advance). We've built a simple pricing tool to show this effect, and would encourage you to play around with it.
If this funding shortage materialises as we suspect, it's likely that investing in the semi-conductor sector might not be the sure winner it has been to this point. Margins always flow to the bottleneck, and where the build-out will be bottlenecked will evolve constantly over the coming years. Our job will be to stay current with developments in both markets and on the technological frontier, to identify these bottlenecks and infer the sectors with the best tailwinds as a result.
Market Intelligence
Once we've identified the sectors with the most prominent tailwinds, the next step is to find the very best opportunities within them. This is the area where our many years of experience in markets will be most valuable. We aren't going to provide an exhaustive and detailed list of ways we can find edge over our competitors, but we are happy to share some examples to give you a sense of our investment process:
- Focus on companies with strong leaders, who understand opportunity early and act decisively, but know the limits of their comparative advantage. It is possible to squander almost any opportunity with poor execution. We are looking for serious leaders who are excited without being exuberant, and—most importantly—who know their place in the market. These leaders concentrate on delivering value to their customers without losing focus and trying to do everything. The market is great at pricing what it can measure, and much less good at qualitative analysis, so good leadership is consistently undervalued. Acquiring this skill does not require psychic abilities, only that you actively try to differentiate companies by their leadership and then apply some reasonable effort and common sense to the task. An easy example is that there are some CEOs who will court the market, trying to do things the market will react positively to, and others who are focused on adding value to the company they are running.
- Unusual market dynamics. While fundamentals certainly drive markets in the long-term, short-term there is often a lot more going on. For long-term investors who understand these dynamics, they can use them to gain a tactical advantage, while for those less experienced, the short-term price movements can cause undue stress or even lead to panicked reactions, severely damaging their long-term outcomes. There are many versions of this but one recent example can be found by looking at the moves in SaaS names during the Iran war. There were several days where heavily-shorted stocks not only outperformed on days where the war worsened, but actually rallied outright. This is not because these stocks suddenly developed an inverse relationship to the global economy, but rather because on these sorts of days, hedge funds are forced to derisk by reducing sizes in all their positions, including in stocks where they have taken short positions. Understanding when you are trading against someone who is being forced to reduce is very different to doing so against someone who is looking at the same information as you and calmly concluding you are wrong. Knowing which of these situations you are facing at a given time is extremely powerful.
- The market is more consistent than it is right. We often refer to "putting the market on a hand," which is a poker analogy; if you can interpret the way the market price is reacting to news such that you understand how it views a stock, or an industry, you will be better able to identify your own mistakes in some circumstances and predict the market's future errors in others. For example, we have a strongly held view that the market is still mispricing memory stocks. We could just leave it there, and not be curious as to why this misprice exists, but that isn't good enough. We need to see if we can form (and seek evidence for) a reasonable thesis which explains their behaviour. In this case, we believe it is because people who have been trading these stocks for a long time (and, indeed, the management of the companies themselves) have seen many cycles where the market switches violently between undersupplied and oversupplied. The market for memory was oversupplied as recently as 2022, leading to losses for many of these companies that year. Given this experience, it isn't unreasonable for them to view the AI demand spike as just another cycle. Not unreasonable, but, at least from our point of view, wrong.
We believe that the AI demand is different, and it is this view—because it varies from the default—that needs justification. We agree with the market's assumption that the current acute memory shortage will not last, and that supply and demand will eventually meet once prices are raised enough to destroy a meaningful amount of demand. At this point, we suspect the market will assume that (as has always been the case before) this will be the top of a cycle that will be followed by the inevitable period of oversupply, whereas we believe that the demand from AI will be robust and ongoing.
We're not sure when this price surge will end, so for us it makes sense to own some memory stocks now, but we are also mindful that we are not yet at the point of maximum disagreement. This will come when prices stabilize and the first small price reductions are recorded. At that point, we expect these stocks to get crushed; assuming they do, there will be an incredible opportunity to buy them. Until then, we will keep following the stocks, the analyst commentary surrounding them, and developments in the technology, constantly questioning our view, actively seeking disconfirming evidence in case we are wrong, and all the while waiting for the right moment to act. - Overlooked or misunderstood stocks. There are >10,000 listed stocks, and there are good reasons why you've never heard of many of them. Trading companies can't afford to do deep dives into all these companies, the volume for many of the smaller stocks just isn't worth it, so instead these stocks are priced off models. These models are pretty good in general, but aren't updated that frequently for smaller stocks; when these companies change, it can take a long time for the models to catch up. A good example of this is former crypto mining stocks, which have moved away from Bitcoin mining to focus their datacentres on high performance computing (HPC) for AI compute. Despite this, many of these stocks fell dramatically in line with the Bitcoin sell off over the last several months.
There are many other ways in which experienced market participants can unearth value. In fact, achieving sufficient scale to allow capacity to do this research is one of the key reasons why investing collectively via a company like Gradient Ascent makes sense.
Principles for Investing
Even when we have identified the right sectors, and found the best opportunities within them, the work still isn't done. We have seen many investors lose money while investing on opinions that were fundamentally correct. This can happen for a variety of reasons including overcommitting too early leading to unsustainable losses, or simply losing confidence in their view when the market went against them, or worst of all, they stayed in a position when circumstances changed, altering the original thesis which was previously true. To ensure that we avoid these kinds of mistakes as much as possible, we will apply the following principles to our investment approach:
Curiosity is at the core of our approach. It's exciting to peel back the curtain and see what's next for the technology. We like to think of this as the basic research part of the job, where you are discussing topics and learning regardless of whether there's an apparent link between the information and future profitability. A broad appetite for information is critical: it is so hard to predict where the next opportunity will emerge from, and in GA we really subscribe to Pasteur's view that "chance favours the prepared mind." We are always looking for more topics to explore; and Jamie has catalogued some of his scattered topics of interest here.
We believe there will be ample opportunity to fill a portfolio with outstanding investments over the coming years. But to find these opportunities, we will need to look deeply into many attractive opportunities which ultimately don't meet the mark. Betting big on the very best positions is how true fortunes are made, but you can only do this if you are confident you understand all aspects of the business, its staff, and its place in the market. While certainty is never achievable, there's a much bigger information set available than most investors are willing to access, simply because deep understanding is a lot of work. In practical terms, this means you've got to be willing to spend hundreds of hours investigating a stock that you might ultimately pass on.
Good investing requires careful examination of 2nd degree knowledge. It is important to spend time and energy assessing our confidence level in our views. How strong is the theory underlying each opinion? What evidence do we have to support it? Do we understand the market's view and why it diverges from our own? All of these questions are important in determining how strongly we hold a given view, and therefore, how much money to put behind it. Even when all these things line up, it is important to keep a sense of epistemic humility. It is great to have strongly held opinions, but it is dangerous to confuse them for certainty. We are not in the business of making specific and grandiose claims about the future: our goal is to find opportunities about which we can express our views for maximum return with minimal risk.
Once we have positions, the related topic of truth-seeking becomes crucial. Opinions lead to positions, and positions intrinsically tie us to our existing opinions. We have to fight the urge to close our minds to information which might cause us to change our views about existing positions. This becomes especially true if a position has gone against us, because, as humans, we naturally try to protect our self-image and play mental gymnastics to avoid admitting past mistakes. We need to be able to see past these biases, recognise our own mistakes and act decisively to reverse positions when necessary.
Applying these principles on a daily basis and staying constantly mindful of the mission of delivering long-term returns for our investors will be key to successful execution of our goals in an exciting but extremely volatile period for markets.
Investors
We want to stress the importance of the role that we see for investors in this venture. The relationship that we imagine is one where investors are part of the team, rather than clients. Like the owners of a professional sports team, their overall support combined with the occasional constructive criticism can be fantastically helpful to the overall success of the project. There are few things more rewarding than productive collaboration amongst a group who share a strongly-held belief, particularly if that belief goes against popular sentiment.
If you have found this essay compelling, and agree with its premise on the future of AI, please follow the link below to get involved. There is much work to be done, and a lot of money to be made.