Blog Hero Image

Posted by Nicholas Mersch on Jun 12th, 2024

Is AI Eating Software?

"Water, water, every where,
And all the boards did shrink;
Water, water, every where,
Nor any drop to drink."

This passage from The Rime of the Ancient Mariner refers to sailors adrift at sea. They are blatantly aware of the irony that although they are surrounded by water, they are tormented by thirst because they cannot drink it.

Such is the plight of enterprise software in 2024.

Key takeaways:

  • AI Monetization: The recent revenue in AI is getting recognized solely around semiconductor companies and cloud providers, while the application layer is still in the early stages.
  • Software Struggles: Software companies are now facing several structural changes: 1) IT wallet share shifting from software to AI; 2) Margin compression for those who were late to AI adoption; 3) Lowering seat count hurts SaaS models; 4) More applications are being built in-house; 5) Barriers to entry are now lower as the marginal cost to develop software declines
  • Bright Spots: 1) Infrastructure Software - Data is the most important element in building GenAI systems. Data infrastructure platforms are critical for running AI effectively; 2) Software Security Companies - More data = more attack vectors. The need to protect this data persists.
  • Revenue Models: Ultimately, the "AI winners" bucket in software is bifurcated, favouring consumption revenue models over SaaS models.

AI opportunity seems to be everywhere, except for the very companies that have dominated the shift in digital transformation over the last several decades: software companies.

In 2011, all-star tech investor Marc Andreessen penned a famous article in the Wall Street Journal titled Why Software Is Eating the World. In it, he argued that software companies were poised to dominate the economy due to the transformative power of software across various industries. Spoiler alert: he was largely correct – and made a boatload of money betting on this.

But just like "software ate the world," we now move to the next important question: Is AI eating software?

Software companies have gotten off to a disastrous start in 2024. An ETF that tracks this performance is the iShares Expanded Tech-Software Sector ETF (IGV). IGV was down 1.1% in May, following -7.4% in April, bringing its YTD performance to -3.7% vs the NASDAQ +11.5% and the VanEck Semiconductor ETF (SMH) +37%.

We are at the point in the buildout cycle in AI where nearly all the value is accruing around the infrastructure/hardware companies. An important parallel to draw here is the monetization cycle that we saw around mobile/cloud and the companies that relatively outperformed the S&P:

The roadmap to monetizing cycles
Source: @carlquintanilla

The recent revenue in AI is getting recognized solely around semiconductor companies and hyperscalers (hyperscalers refer to the cloud service providers Amazon's AWS, Google's GCP, and Microsoft's Azure). The semiconductor revenue is getting recognized through selling GPUs to hyperscalers, who are then recognizing revenue from renting out those GPUs to AI startups that are initially training these models and subsequently calling these models (inference).

What we have yet to see materialize outside of a select few companies is the monetization of the application layer.

This became especially evident over the course of off-quarter earnings amongst software companies. Of the ten largest cloud software providers by annual revenue, eight have seen their stocks sell off by an average of 9% the day after their latest results. We got valuable readthroughs from this quarter, which I'll use to highlight the structural market dynamics that are unfolding:

IT wallet share is shifting from software to AI.
Throughout this earnings quarter, a common rhetoric we saw amongst executive teams was that IT departments are delaying major software decisions before they formulate an AI strategy. Some blame it on the macro, and while there is some truth to that, there is no denying that AI is a factor. After putting up a terrible quarterly growth number, Salesforce repeated the same three horsemen: elongated deal cycles, deal compression, and high levels of budget scrutiny.

Is "macro" just another way of saying, "AI is eating our lunch?"

Margin compression for those who were late to the AI party.
Companies that did not bring AI capabilities in-house early are starting to pay for it through margin compression. The most prevalent amongst these is Snowflake. Snowflake saw its stock fall 5% following its earnings, which included a sharp cut to its operating margin projection for the year because of its AI investments. The reality is that Snowflake cannot get on-demand GPU pricing right now with hyperscalers due to limited availability, which is resulting in Snowflake paying for GPUs, whether they use them or not in their own on-demand model with customers. The expectation is that hardware availability and hyperscaler pricing will improve over time and become a small tailwind to gross margins.

Lowering seat count hurts SaaS Models.
AI is the next revolution across productivity enhancement. By layering AI solutions across nearly every department, it is now possible for three people to do the job of 10. AI applications are essentially a substitute for an army of interns at a fraction of the cost. This lowers the TAM for all seat-based SaaS models significantly.

The first pocket of the market that this has come after is customer experience-based companies. You no longer need a large call center to handle customer inquiries when you can layer in a chatbot and reduce headcount. If you are a software solution provider selling a product in this area based on a per seat per month basis, you saw your revenue collapse. LivePerson (LPSN) is a company in this area and has seen its stock price implode from $70 to $0.67. But it isn't only customer experience. Workday (WDAY), an HR software company, cited "increased deal scrutiny," adding that customers were committing to "lower headcount levels" on deal renewals. In the creative content scenario, teams are paying OpenAI or MidJourney a fraction of the cost compared to an entire creative team, each with an Adobe (ADBE) subscription, which is now only used for final touch-ups.

More functions are being brought in-house.
With the ability to write code, generate images, and ingest massive amounts of information rapidly, companies no longer need to outsource certain elements of their workflow. By empowering in-house employees, they are optimizing internal workflows within their own walls while at the same time ensuring quality control. This is a cost-structure unlock at the company level at the expense of software revenue.

Barriers to entry are now lower.
It used to take a team of software engineers to build out version one of an application. However, as Jensen Huang, Nvidia CEO, said in a fireside chat, "The new coding language is English." When low/no-code solutions are prevalent across the space, this greatly reduces the technical frictions around barriers to entry for entrepreneurs with an idea but don't have the capabilities to hire software engineers. This effectively reduces the margin cost of creating code to near zero.

Now – there is definitely a high requirement for capital at the compute level to build out these datacentres and large language models, but the productivity unlock at the startup and operational levels in companies keeps getting better with every new iteration of frontier AI models.

But it's not all doom and gloom for software companies.
Despite these trends, I believe there are still going to be "AI winners" in the software space around two core components:

More attack vectors underscore the need for enhanced security.
Every organization has a vast amount of proprietary data that large language models do not have access to train upon originally. When unleashing AI within an organization, monitoring permissioning and access of this data is critical to ensure privacy and security. As data continually expands, so does the need to safeguard it. Winners in this space include Crowdstrike (CRWD), Zscaler (ZS), SentinelOne (S), Netskope (private) and Wiz (private). These companies have very specific specialized applications that will only become more valuable.

Consumption-based infrastructure companies will become more critical.
Again – AI leads to the increased need to leverage your data, and you need a way to manage this asset effectively. Data, whether structured or unstructured, is the most important element in building GenAI systems. There is significant value unlock in combining an LLM with a query engine or anything tied to data catalog or data orchestration workflows. Data infrastructure platforms are also critical to ensuring LLMs are trained on and inferenced against high-quality data.

Infrastructure software companies tend to "play nice" with the hyperscalers. On MongoDB's call, the CEO said, "Our relationship with the hyperscalers is actually very strong. We partner very closely with AWS, Azure and GCP in the field. And in fact, they're coming to us to partner on deals more frequently than we've seen in the past. And our win rates, frankly, are very high. So we don't see any issues where we're losing deals to any particular vendor whether they're the hyperscaler or a small independent company."

However, there is a caveat for this bull thesis to play out. We need to closely monitor this to ensure hyperscalers don't start "competing down" into smaller markets that these companies are operating in.

It appears there is still room (for now) for these players to carve out their respective niches.

So… is "AI eating software?"
Yes and no.

Ultimately, what this does is start to shrink the list of companies that are in the "AI winners" bucket. While software might have originally been thought of as the distribution layer of AI, it appears that there is much more bifurcation than expected.

Calls for the death of software are mostly sensationalized. There will need to be an adaptation of business models in the new normal. We went through this when we moved from mainframe to PC and again from PC to Mobile/Cloud and now from Cloud to AI. Revenue models shifted from license/on-prem to SaaS/Cloud.

Here comes the conclusion. I believe new revenue models focusing more on usage and consumption will prevail in this new paradigm of a shrinking seat-count environment.

The AI revolution is coming for nearly every knowledge-based role, and software companies will need to adapt or get left behind.

Strong convictions. Loosely held.

— Nick Mersch, CFA, Portfolio Manager

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.

Fund mentioned in this story

Nicholas Mersch, CFA

Nicholas Mersch has worked in the capital markets industry in several capacities over the past 10 years. Areas include private equity, infrastructure finance, venture capital and technology focused equity research. In his current capacity, he is an Associate Portfolio Manager at Purpose Investments focused on long/short equities.

Mr. Mersch graduated with a bachelors of management and organizational studies from Western University and is a CFA charterholder.