Every week I’ll provide updates on the latest trends in cloud software companies. Follow along to stay up to date!
Authentication in the Age of AI Agents
I’ve been reflecting lately on how AI agents might evolve and how they’ll be built. Specifically, I’m interested in the distinctions between web-based and desktop-based AI agents. I don’t think these are the only two buckets of AI agents, but I think they cover a lot of the market.
Web-based agents are ones that can complete a task end-to-end in a browser session. Go to this website, grab this information, take that information and enter it into this website, etc. In the past, browser automations largely revolved around QA testing or scraping. Older tools like Playwright, Selenium, Puppeteer, etc enabled a lot of this. But they are fickle and brittle. They were extremely deterministic, and relied on defining a specific action related to a specific web page (or pages). If an html tag changed, the text on a button changed, then the script would break. LLMs completely changed the game, because they made these browser automations more flexible. Let’s say you wanted to log in to a website, and that website had a “log in” button. In the past, you’d tell the script to look for the button labeled “log in” and take an action. But what if that website changed their login button to “sign in” or “enter.” Or any other version of “log in.” Then the script would break. What’s great about a LLM is it can recognize the intent. If I told a LLM to look for the log in button, it won’t look for those specific words, but for the word(s) that most likely represent the “log in” feature. Expand this to all kinds of actions and you can see how web based automations can now become a lot more flexible (and less hard-coded and brittle). LLMs also enable chaining together much more complicated browser automations. At the same time, we have a budding ecosystem of new companies supporting browser based automations (like Browserbase, who’s creating cool virtual browser tools to enable browser based agents). But there’s an unsolved problem around authentication… If we think back to old school browser automations, a lot of them were “bad actors.” Scraping or bot attacks (or things like that). To combat this, the internet started to develop anti-bot technologies (captchas, two factor, etc). Basically a set of technologies who’s main goal is to recognize human visitors from non-human visitors, with the purpose of blocking the non-human visitors. But in a world of browser based agents – do you want to block these visitors? It’s a complicated answer. It may seem like an obvious “no of course not.” But what if an independent travel booking AI agent company was created. What if a Priceline, Booking, Kayak, etc wanted to create their own travel booking AI agent. Maybe they would only want to allow their own agent to interact with their website? Either way, authentication will be a big problem. The benefit of web based agents is they can string together many actions in a row. But if at some step along the way you need to log in to a website, how will the LLM effectively store and secure those credentials so they can enter them on the log in page? Or will we have a separate SSO style “agent login” button similar to a “log in with google” button? Who knows, only time will tell. But this appears to be one bottleneck of deploying web-based agents at scale. Another limitation of web based agents is they only can complete workflows that exist inside a browser..
Desktop-based agents are similar to web-based agents, but they can take actions on applications, the computer file system, etc. There are some applications (particularly old school ones) that don’t have a web portal, only a desktop app, that you want to access in a workflow. Or you want to download a file, and upload it somewhere else as part of a workflow (you need access to a file system to do this). Desktop-based agents (as I’m describing them) more closely mirror “computer use” agents we hear some of the large AI labs talk about. Historically, virtual desktops have been SLOW. I’m thinking about my days in investment banking when I had to VPN into my desktop computer from my laptop at home…not fun.. For these agents I think we’ll need a new set of virtual desktop infrastructure and tooling geared more specifically towards AI agents, just like Browserbase is doing for virtual desktops. And just like web-based agents, desktop-based agents will also need to solve the authentication problem (with much higher stakes, given they would have control of the entire desktop). We’ll need to rethink SSO to enable this future.
Authentication, virtual browsers, and virtual desktops are just a couple areas that I think will undergo creative destruction as we enter the age of AI agents.
February Inflation
February Inflation (CPI) Update:
-
Headline CPI: 2.8% YoY (2.9% consensus) and 0.2% MoM (0.3% consensus)
-
In January Headline CPI was 3.0% YoY and 0.5% MoM
-
-
Core CPI: 3.1% YoY (3.2% consensus) and 0.2% MoM (0.3% consensus)
-
In January Core CPI was 3.3% YoY and 0.4% MoM
-
Q4 Cloud Software Earnings
We’re now largely through Q4 earnings season. So far the results have been “meh.” Rubrik blew the top of earnings yesterday, beating Q4 estimates by 11%, guiding Q1 8% above consensus, and guiding full year 8% above consensus. But for the most part, Q4 earnings have been average at best. Coming into earnings season, I expected much stronger results. It seemed like the qualitative feedback I was getting heading into the quarter was better, and I also expected to see some budget flush in Q4. Instead, Q4 saw a decline in most reported metrics.
First – The median guide for Q1 was slightly below consensus estimates. This comes after 6 of the last 7 quarters saw the median forward guide come in above consensus. You can see the overall median guidance vs consensus graphed below.
Second – when we look at the number of companies who beat Q4 estimates that figure dipped back down to 91% after being nearly 100% in the prior quarter (graph below)
Third – forward guidance has not impressed. 50% of companies guided Q1 below consensus estimates. This is the lowest its been in the last 8 quarters. Are companies guiding conservatively given all the political and economic uncertainty? Time will tell. But for now, Q4 has not been an exciting or positive earnings season for cloud software companies.
Quarterly Reports Summary
Top 10 EV / NTM Revenue Multiples
Top 10 Weekly Share Price Movement
Update on Multiples
SaaS businesses are generally valued on a multiple of their revenue – in most cases the projected revenue for the next 12 months. Revenue multiples are a shorthand valuation framework. Given most software companies are not profitable, or not generating meaningful FCF, it’s the only metric to compare the entire industry against. Even a DCF is riddled with long term assumptions. The promise of SaaS is that growth in the early years leads to profits in the mature years. Multiples shown below are calculated by taking the Enterprise Value (market cap + debt – cash) / NTM revenue.
Overall Stats:
-
Overall Median: 5.2x
-
Top 5 Median: 16.5x
-
10Y: 4.3%
Bucketed by Growth. In the buckets below I consider high growth >27% projected NTM growth (I had to update this, as there’s only 1 company projected to grow >30% after this quarter’s earnings), mid growth 15%-27% and low growth <15%
-
High Growth Median: 9.6x
-
Mid Growth Median: 8.6x
-
Low Growth Median: 4.0x
EV / NTM Rev / NTM Growth
The below chart shows the EV / NTM revenue multiple divided by NTM consensus growth expectations. So a company trading at 20x NTM revenue that is projected to grow 100% would be trading at 0.2x. The goal of this graph is to show how relatively cheap / expensive each stock is relative to their growth expectations
EV / NTM FCF
The line chart shows the median of all companies with a FCF multiple >0x and <100x. I created this subset to show companies where FCF is a relevant valuation metric.
Companies with negative NTM FCF are not listed on the chart
Scatter Plot of EV / NTM Rev Multiple vs NTM Rev Growth
How correlated is growth to valuation multiple?
Operating Metrics
-
Median NTM growth rate: 11%
-
Median LTM growth rate: 15%
-
Median Gross Margin: 76%
-
Median Operating Margin (5%)
-
Median FCF Margin: 16%
-
Median Net Retention: 107%
-
Median CAC Payback: 44 months
-
Median S&M % Revenue: 38%
-
Median R&D % Revenue: 25%
-
Median G&A % Revenue: 16%
Comps Output
Rule of 40 shows rev growth + FCF margin (both LTM and NTM for growth + margins). FCF calculated as Cash Flow from Operations – Capital Expenditures
GM Adjusted Payback is calculated as: (Previous Q S&M) / (Net New ARR in Q x Gross Margin) x 12 . It shows the number of months it takes for a SaaS business to payback their fully burdened CAC on a gross profit basis. Most public companies don’t report net new ARR, so I’m taking an implied ARR metric (quarterly subscription revenue x 4). Net new ARR is simply the ARR of the current quarter, minus the ARR of the previous quarter. Companies that do not disclose subscription rev have been left out of the analysis and are listed as NA.
Sources used in this post include Bloomberg, Pitchbook and company filings
The information presented in this newsletter is the opinion of the author and does not necessarily reflect the view of any other person or entity, including Altimeter Capital Management, LP (“Altimeter”). The information provided is believed to be from reliable sources but no liability is accepted for any inaccuracies. This is for information purposes and should not be construed as an investment recommendation. Past performance is no guarantee of future performance. Altimeter is an investment adviser registered with the U.S. Securities and Exchange Commission. Registration does not imply a certain level of skill or training.
This post and the information presented are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future.