2024 wraps up another stepping stone year for AI. But it still feels like we were chasing fireflies in the dark. Each glimmer of insight appears fleeting, and catching one feels both elusive and rewarding.
Experimenting with early instances of AI feels indistinguishable from magic. While large model iterations push the limits of what we thought possible, it’s still in the “science project” phase of rollout.
IT departments have begun to implement AI solutions into their workflows, and we’ve figured out that the early markets for AI are customer support, writing code, digesting legal docs, and simulating drug discovery.
We’re moving the goalposts from chatbots to reasoning and now to agents. Later will come groundbreaking ideas that humans alone could not conjure, followed by AI that can function as an organization upon itself.
Will 2025 be the year that we catch lightning in a bottle?
2024 in Review
We had another gangbuster year, with major indices up another 20%+, marking strong back-to-back gains. Mag7 dominated both headlines and returns as they became adopters, adaptors, and enablers around the new regime shift in artificial intelligence.
The AI theme saw stocks rally throughout its vertically integrated value chain. Physical infrastructure players from semiconductors (poll position), racking systems, and cooling systems all benefitted as pick-and-shovel plays. Energy plays (particularly NatGas IPPs) saw a huge renaissance as previously predictable low growth now step-functions higher with the power-hungry data centre explosion.
Cloud computing finally saw an inflection in growth rates from the Covid hangover of 2023.
Software stocks have had a wild ride as VCs claim software dead and sell-side analysts tout system-of-record advantages, with the truth somewhere in between.
A stubbornly strong economy led the way to more growth despite inflation that seems to be reappearing, as well as a higher for longer rate environment as the 10yr creeps back up. 5% is the new 3%, and good news is bad news again, as the recent jobs report sent the market down, with some outliers claiming that we are now closer to a hike than a cut.
Throw in some volatility with the new administration’s opaque tariff posturing (and even antitrust), and we’ve got another interesting setup. I stopped watching House of Cards during Trump 1.0 because real life was even more interesting than the TV drama, and you can’t make some of this stuff up.
We’re in for another wild ride.
5 Predictions for 2025
Here, I will highlight five key themes through targeted predictions that I am focused on in 2025. The fifth theme will decide the fate of the performance of the overall market.
(See what I did there? Now you have to read it all).
1. Accelerating compute remains a key trend, and Nvidia will be the most valuable company in the world for the majority of 2025.
2024 was the year of Nvidia. The man in the leather jacket took the tech world by storm by enabling this revolution. Make no mistake, AI companies would still be treading water if they didn’t have parallel processing to do all this math on. Nvidia doesn’t sell just a chip; they sell an entire accelerated compute system. The unit of production is shifting from a single GPU to an entire data centre. The market sorted Semiconductors into two buckets: AI and non-AI. AI won, and everything else lost.
Although I believe we’re in the 6th or 7th inning of this CapEx buildout cycle, Nvidia is still a strong buy as long as they:
- Hit production targets with Blackwell,
- Continue to push the product cycle cadence with Rubin (~2yr per gen), and
- Protect pricing power.
If they defend these three components, Nvidia will easily be the most valuable company in the world for the majority of 2025.
We also saw the market wake up to the custom silicon solutions. These are chips that the hyperscalers (AWS/GCP/Azure + Meta) are building in-house that are enabled by select partners. Over 24H2, we saw Nvidia fatigue as it traded sideways while ASIC designers Marvell and Broadcom went on an absolute tear. While Broadcom and Marvell will benefit from hyperscalers trying to dampen Nvidia’s pricing power, there is more than enough room for both approaches to win. Long Nvidia, as well as Marvell and Broadcom. I also like Teradyne (testing), Micron (high-bandwidth memory), and TSMC (Wafer Fab).
2. Bottlenecks will move more evidently from chips to power by the end of 2025.
In the early stages of data centre buildout, the focus was on locations like Northern Virginia and Silicon Valley, optimizing for latency. Now, the trend is shifting toward areas with cheaper land and lower energy costs – like Texas. Texas took the bull by the horns (lol) earlier this year when it earmarked ~$5B of low-interest loans for ~10GW of new natural gas generation capacity to help meet expected power demand growth in the state. It seems they were the only ones to get the memo of the step function increase in demand:
In this space, I prefer IPPs like Vistra and Constellation, which focus on natural gas for the medium term, with a nuclear option over the longer term. Vistra has about 20GW of highly efficient natural gas capacity and recently did a 2GW expansion in Texas back in May. Constellation’s recent acquisition of Calpine is exactly in line with this thesis. The energy industry is gearing up for growth and cutting red tape, setting the stage for a strong nuclear resurgence. However, immediate energy needs still require natural gas as a critical backstop.
3. AI finally starts its move from the model layer to the application layer due to Agents. Early winners will be data infrastructure companies.
Everyone is tossing out analogies on how to think of AI distribution and productization (monetization). The two that I particularly enjoy are frontier LLMs as the operating system and LLMs as the CSP (cloud service provider). All these breakthroughs on the LLM front enable a score of other applications to be built on top of them (much like a cloud platform or an operating system). They do the hard technical stuff underneath, while companies can build tailored solutions to specific problems.
(Side note: people need to stop calling all AI startups “GPT wrappers” as an insult. That’s like saying Snowflake is just an SQL wrapper.)
Much like the cloud-era SaaS companies disrupted licenses and on-prem, we have the incumbents vs. the disruptors. The incumbents in this case are enterprise SaaS companies like Salesforce, while the disruptors are AI-native companies like Harvey (an AI company in the legal space that is powered by OpenAI). The incumbents have distribution, systems of records (customer data), established system integration, and familiarity with IT budgeting departments. The disruptors come at this with an AI-native architecture, rethinking what is and is not necessary. The only chance these incumbents have of surviving is implementing their own agents as the marginal cost of software moves to zero and business models shift from subscription to consumption (or value add).
You know who wins in either scenario? The data infrastructure plays. All of these solutions need a clean data estate in order to run smoothly. My top pick in this space, which will likely win in several different scenarios, is Snowflake. Other infrastructure plays I like include Datadog, Confluent, and Cloudflare. Additionally, DevOps companies like Gitlab and Atlassian can leverage AI capabilities instead of getting disrupted by them. I think a company like ServiceNow can properly implement agents, but I think Salesforce (despite the push on “Agentforce”) could be in trouble.
4. Cryptocurrency starts to matter again on the back of deregulation. This time, we get better utility.
Look, I’m not going to stand here and put a target price on Bitcoin, but you have to pay attention to what’s going on. Follow the flows. ETF inflows are showing massive inflection, as there is now well over $100B in AUM across the ETF product sleeve. This is causing everyone and their dad to “get off zero” and allocate 2% of their portfolio to the asset class.
The regulatory environment is shifting from litigating everything that has any crypto attached to it to regulating within a new framework. If you look at some of these payment systems, there are so many use cases where you can have positive utility behind what the technology is trying to do.
Yeah, there’s going to be wacky stuff like Microstrategy and all that, but this is getting institutionalized now to the point where you can’t really ignore it anymore. I’m not a maxi, but I’m not a Luddite, and I think you need to pay attention here. Keep it simple: own an ETF (Purpose Bitcoin ETF)… they’re safe and legal, and you don’t need to worry about forgetting a 13-word phrase as your password or losing your USB key by accident.
5. RO“AI” will be the most important thing in 2025 when it comes to fundamentals, and it’s not even close.
The Mag7 are firmly in the driver's seat. We know this. We’ve all seen those same old charts saying that the top companies make up more and more of the total share of the overall index with some sensationalized saying like, “CONCENTRATION LIKE THIS NOT SEEN SINCE THE DOTCOM CRISIS.” Fine. Still, the reason that they make up so many major indices is that they have the fundamentals to justify it. They are growing revenue and earnings faster than the rest and deserve what they get. Although these companies account for north of 30% of the market cap of the S&P, they are also responsible for more than 60% of the economic profit of the S&P. Seems pretty fair to me.
Now – since we know as goes the Mag7, so goes the market – let’s get to the most important thing for these companies: RO“AI,” or return on artificial intelligence. Megacaps are slated to spend north of $250B building these data centres out to do the fancy AI math. If AI revenue or cost savings don’t gain significant traction at least by the end of 2025, then there is a wrench in the plan. By recent estimates, all AI revenue is probably around $15-20B that we can definitively point to. This needs to be at least 5x by the end of next year for any of this investment to make sense.
However, an equal benefit that might not be as cut and dry as revenue is cost savings. Megacap companies are equal parts sellers and customers of their own AI productivity enhancements. They can experience significant operating leverage through hiring slowdown as an army of interns is now accessible in the form of AI agents.
The timing of CapEx revisions (still going up) vs. AI revenue (not enough yet) and cost savings (not as visible yet) will dictate forward EPS revisions. Earnings are the number one predictor of stock prices over the long run. Hence – As goes RO“AI,” so goes the market.
Thank you for coming to my TED talk.
Wrapping Up
I think we’re at an incredibly exciting part of the coming AI wave. There was a lot of fear earlier that large models are plateauing in terms of LLM benchmarks, and that may be the case. But we are now shifting from training scaling laws to post-training, reasoning and time inference scaling that is unlocking a whole other avenue.
Regardless, if you stopped all the advancement that we’re seeing in AI today and just looked at what the existing systems are capable of, I think you would have two decades of implementation that needs to be done. Just remember - people still send paper checks.
We have a boundless era ahead of us for anyone who can figure out how to harness artificial intelligence. This will enable the optimization of workflows, spreadsheets, operations, drug discovery, and physics breakthroughs, or even just as a companion when you’re listening to a podcast and want to understand something better.
Incredibly exciting time to be alive, and tech is at the center of it.
That’s where I like to be.
Strong convictions. Loosely held.
– Nick Mersch, CFA
Get the latest tech insights to your inbox every month.
The content of this document is for informational purposes only and is not being provided in the context of an offering of any securities described herein, nor is it a recommendation or solicitation to buy, hold or sell any security. The information is not investment advice, nor is it tailored to the needs or circumstances of any investor. Information contained in this document is not, and under no circumstances is it to be construed as, an offering memorandum, prospectus, advertisement or public offering of securities. No securities commission or similar regulatory authority has reviewed this document, and any representation to the contrary is an offence. Information contained in this document is believed to be accurate and reliable; however, we cannot guarantee that it is complete or current at all times. The information provided is subject to change without notice.
Commissions, trailing commissions, management fees and expenses all may be associated with investment funds. Please read the prospectus before investing. If the securities are purchased or sold on a stock exchange, you may pay more or receive less than the current net asset value. Investment funds are not guaranteed, their values change frequently, and past performance may not be repeated. Certain statements in this document are forward-looking.
Forward-looking statements (“FLS”) are statements that are predictive in nature, depend on or refer to future events or conditions, or that include words such as “may,” “will,” “should,” “could,” “expect,” “anticipate,” intend,” “plan,” “believe,” “estimate” or other similar expressions. Statements that look forward in time or include anything other than historical information are subject to risks and uncertainties, and actual results, actions or events could differ materially from those set forth in the FLS. FLS are not guarantees of future performance and are, by their nature, based on numerous assumptions. Although the FLS contained in this document are based upon what Purpose Investments and the portfolio manager believe to be reasonable assumptions, Purpose Investments and the portfolio manager cannot assure that actual results will be consistent with these FLS. The reader is cautioned to consider the FLS carefully and not to place undue reliance on the FLS. Unless required by applicable law, it is not undertaken, and specifically disclaimed, that there is any intention or obligation to update or revise FLS, whether as a result of new information, future events or otherwise.
