TMT Investment Banking Trend: Artificial IntelligencePublished:
Many joining banks that require them to go through a group placement process just select their preferred group based on generalized prestige, exit opportunities, etc.
There’s nothing wrong with that, and everyone does it to a certain extent. But in order to make your time in banking more tolerable (or even occasionally enjoyable!) it’s always a good idea to have some real interest in the trends informing your new group.
Practically, this means that if you’re interested in tech banking you shouldn’t be too pessimistic or cynical about whatever the latest hype cycle is. Because, in the end, deal flow is going to be heavily informed by whatever hype cycle we’re in (this is partly why some boutiques leaned so much into crypto the past few years, for example).
Today, the trend that is dominating all tech discussions is AI. And, relative to the hyped-up trends of the past five years, there’s much more merit to this trend (or, put another way, much more substance behind it).
Chances are you’ve heard about the recent release of Open AI’s ChatGPT-4. It’s currently only available to paying members ($20/mo) and you can only sign-up if you had access to previous ChatGPT iterations. But, if you haven’t tried it out, it’s really quite remarkable.
It’s perhaps the first time that we’ve personally looked at a technological advance and thought that this won’t just be marginally disruptive to some corner of the economy – the way some stellar vertical software could be, for example – but that this kind of AI advance seems destined to radically reorient many industries. (Along with disrupting many small, medium, and large cap tech companies who have carved out narrow niches that suddenly will be bowdlerized by what ChatGPT can do.)
In the end, this is why Microsoft likely made their $10bn investment in Open AI. Recognizing that it’s cash cow – the MS Office products – could be leapfrogged by a competitor, for the first time, if it somehow integrated new wave AI to make folks more efficient. So instead MS leaned in and not only integrated ChatGPT with Bing but also with the Office suite of products. And it did so with remarkable speed given that it’s barely been two months since their investment in Open AI.
Anyway, a few weeks ago Morgan Stanley published a phenomenal 100-page primer on the current AI landscape. Unlike many AI think pieces that are floating around, it illustrates that the evolution of AI has occurred over decades.
Further, it breaks down, in relatively layman terms, the vocabulary you should have (i.e., what transformers, multi-modal models, etc. are) and who among the major tech players are poised to lead the way on AI in the coming years.
The full report is linked in the members area. But in this post I’ll just take a stab at giving some high-level takeaways from the report, and also talk briefly about how this will likely inform deal activity over the next year. Because, as mentioned earlier, deal activity in tech banking is always going to be partially driven by whatever the prevailing trend is, and today there’s no doubt that it’s AI.
The Timeline and Adoption Rate
Depending on how exactly you want to define it, AI goes back decades and there has been, quite famously, a number of AI winters. However, Morgan Stanley decided to start their timeline in 1997.
But even within this narrower timeline, there was arguably at least one recent AI winter. Because after the publication of Superintelligence by Nick Bostrom in 2014, which helped spur a real resurgence of interest from those outside the field, there wasn’t that much that happened (at least on the surface) until Google launched BERT in 2018 and Open AI announced GPT-2 in 2019.
From a consumer perspective, it likely wasn’t until 2022 that there was meaningful broad-based interest in something that was a standalone AI product: ChatGPT-3. And it wasn’t the case that it only made waves in more tech-conscious consumer circles; it took just five days to reach a million users...
Part of why it took so long to get to a real consumer-facing application that could be both understood and found to be useful was that AI, as it currently stands, is built on several building blocks that have been continually improved over the past decade.
Indeed, these building blocks (most notably machine learning) have been a key driver over the past five years of both traditional tech companies and the strategic acquisitions that many have made (often acquiring companies more for the people than the underlying tech itself).
While the Morgan Stanley deck gets into the weeds on the tech and terminology behind AI more than it really makes sense to here, below is a great graphic showing the building blocks from bolts to bits...
How AI May Reshape the Broader Economy
As we’ve written about before, there’s a certain way in which tech deal flow is partly agnostic to the broader economic cycle. This seems a bit contrarian to say given that tech has benefited most from being in a low rates environment for the past decade and now that we’re in a higher rates environment deal flow has diminished significantly.
However, it’s still the case, as we’ve written many times before, that many of the larger tech companies are sitting on large piles of cash and are acutely aware of the need to avoid being “disrupted” partly because they became as large as they are through disrupting existing players.
So, for many, regardless of where we are in the economic cycle, they’ll be willing and able to deploy capital to ensure they stay on the bleeding edge if it’s deemed necessary; whether that’s through large-scale acquisitions or through acquihires of promising startups at eye-watering multiples.
It’s largely speculation at this point. But it also does seem that we’re approaching a tipping point, that may still be a year or two away, in which many companies will be radically reorientated due to the emergence of AI tools. This goes well beyond just tech-focused companies to those in the broader economy that could see their own approaches to, for example, medical imagining be suddenly one-upped by broad-based AI tools (i.e., tools that weren’t necessarily designed with medical imaging in mind but that can be used to do it better than existing solutions relying on different technology). Naturally, this could mean there’s suddenly more interest from strategics outside of the tech ecosystem in acquiring AI-focused tech companies.
If you’ve played around with ChatGPT-4, or have seen its demonstration, it’s difficult to look at many of the highlighted fields below and not see that disruption is imminent – and that companies who quickly look to acquire AI-focused companies will have a significant first mover advantage...
Note: Along this same line, Open AI just published a research paper discussing how disruptive they imagine GPT will be to the labor market moving forward. They break things down showing the hypothetical level of disruption to various occupations and it's a fascinating read.
The AI Enablers
As with any newly emergent technology, in order to have scale you need to not only have the technology, but you need to be able to deploy it. This is why MS spent a significant amount of time in their report discussing what they dubbed the “AI Enablers”: those that have the human capital to create and refine the models, and the resources and infrastructure to deploy them.
Perhaps it’s not surprising that it’s really the mega-cap tech names (plus, NVDA) that are highlighted...
But the inclusion of NVDA is just because they can enable growth in the broader AI ecosystem. When it comes to the main beneficiaries (today) from AI, it’ll be those with the largest accessible data sets to train their models on...
And while many point to just how cheap ChatGPT is and how its cost curve has been beaten down, there’s no doubt that capex spend has and will need to increase significantly in the ensuing arms race to have the most commercially-viable AI solution. Here’s one way to somewhat visualize the cost of AI to the AI Enablers...
In the wake of the release of Bing Chat – that, it turns out, was running Chat GPT-4 all along – many thought this was a core risk to Google’s search business. But MS is not only much more skeptical of this, they think Google is perhaps the best prepared of all the mega-cap tech names across the AI ecosystem when viewed in totality...
How AI Will Impact Deal Activity
In the beginning of this post, it was mentioned that tech investment banking – more so than with other coverage areas – tends to be influenced, from a deal flow perspective, by whatever the latest and greatest trend happens to be.
But it’s important not to overstate this. For example, given the still unfavorable funding environment, the vast majority of deals that sponsors are doing are going to involve regular-way tech companies that have relatively robust cash flows (i.e., deals that aren’t underpinned by large growth assumptions). This can already be seen with deals like Blackstone taking Cvent private.
But there are some recent sponsor deals that may not be pure-play AI deals, but that certainty have an AI flavor to them. For example, Vista doing a buyout of Duck Creek (the intelligence solutions provider in the property and casualty space). For Duck Creek, it’s quite obvious how AI – as it currently stands – can be leveraged, and that those in their industry that do so will probably have a significant advantage relative to those that don’t. But, at the same time, it’s not like Duck Creek is a pure-play AI company: they’re just partly leveraging AI to enhance their existing offering.
Note: BX did the Cvent take-private at a 7.3x EV/Rev multiple, and Vista did the buyout of Duck Creek at a 7.6x EV/Rev multiple. Qatalyst advised Cvent, and JPM advised Duck Creek.
Note: While technically done late last year, Thoma Bravo did a buyout of Coupa (all cash) at an eye-watering 10.2x EV/Rev multiple. Qatalyst advised Coupa while GS and Piper Sandler advised Thoma. Another late-2022 buyout by Thoma was ForgeRock at a 9.8x multiple and an over 50% premium to its closing share price (JPM advised ForgeRock).
For strategics, especially those sitting on significant cash, we’ll likely see some more pure-play AI acquisitions this year or those that further allow the “AI Enablers” to grow (i.e., acquisitions that facilitate data center growth).
Therefore, even though there’s no getting around the reality that tech deal activity has fallen sharply year-over-year, it’s likely going to be the case that we do see at least the same level of deal flow we’ve seen over the past few quarters or a slight uptick. Partially because some, like Thoma, just need to deploy capital from their freshly raised funds, and partially because strategics need to look through current macro conditions (especially if they have significant cash) to protect and grow their business (and the trend of AI may very well inform many of these more protective acquisitions).
In an interview, no one expects you to be an expert on any broad sector trend (especially one as technically complicated as AI). However, sometimes interviewees do bring up a sector trend in an interview but it quickly becomes pretty obvious that they don’t know much about it.
So hopefully this has been helpful in at least laying out a framework for thinking about AI from a slightly more banking perspective. If you haven’t already, make sure to check out the post covering TMT investment banking to get a better feel for sector breakdowns, and if you’re currently applying for tech-specific banking roles then take a look at the tech investment banking interview questions. Finally, keep in mind that even if you’re applying to a tech group, it’s great to show your interest in tech but you also need to make sure you have all your classic investment banking technicals down cold.