This article first appeared on GuruFocus.
Software is eating the world was the meme for the first 20 years of this century. Now it seems that AI has started to eat Software.
I have noticed that over the last year, many large software company stocks have not being doing as well as before. Since the advent of AI, the PE Ratios of these companies have started to compress suggesting a cooling of investor enthusiasm.
Technology has underperformed the overall market and within the Technology sector, Software has been the worst performing industry.
Is AI now threatening Software Companies? Is AI now threatening Software Companies?
Fears of AI disruption may have contributed to PE ratio compression for software companies like Adobe (NASDAQ:ADBE), ServiceNow (NYSE:NOW), Salesforce (NYSE:CRM), and Workday (NASDAQ:WDAY) over the last five years, particularly intensifying since 2023, as AI-native competitors emerged. Current PE ratios (as of early 2026) stand notably below their 2021 peaks and multi-year averages: ADBE at ~18x versus 61x in 2021 and a 5-year average of 42x; NOW at ~77-105x versus 560x in 2021 and a 5-year average over 300x; CRM at ~30x versus historical peaks over 200x; WDAY at ~79-88x versus a 10-year average of 167x.
The chart below shows PE Ratio’s without Non-Recurring Items (PE w/o NRI) for selected business oriented Software as a Service Vendors. Average drawdown from Peak, over 5 years is more than 60%.
Is AI now threatening Software Companies?
A similar dynamic is observed with Price to Sales (PS) Ratio.
Is AI now threatening Software Companies?
Further, observations show that the reduction in PE & PS Ratio is mostly due to investor sentiment as growth in revenue and operating earnings remains relatively strong.
ADBE
CRM
NOW
WDAY
TSX:CSU
PE Ratio Current
14.13
19.71
38.3
21.53
60.75
PE Ratio 2020
53.67
45.85
109.65
77.66
120.06
PE Compression
-39.54
-26.14
-71.35
-56.13
-59.31
1-Year Operating Income Growth Rate (Per Share)
18.5
22.5
38.2
102.2
33.2
3-Year Operating Income Growth Rate (Per Share)
16.3
140.9
73
0
25.1
5-Year Operating Income Growth Rate (Per Share)
16.7
88.1
84.2
0
24.4
10-Year Operating Income Growth Rate (Per Share)
25.6
0
0
0
22.7
1-Year Total Revenue Growth Rate
10.5
8.4
21.1
13.2
19
3-Year Total Revenue Growth Rate
10.5
12.7
23
18
29.9
5-Year Total Revenue Growth Rate
12.4
17.5
25.8
18.6
27.1
10-Year Total Revenue Growth Rate
17.6
22.9
31.8
26.1
21.2
Current PE ratios for ADBE, CRM, NOW, WDAY and CSU are significantly lower than their 2020 levels, signaling compressed valuations across these SaaS enterprise software firms amid market pressures. Growth metrics are healthy over 10,5, 3 and 1 year periods with CRM and NOW showing explosive operating income surges over 3 and 5 years, while WDAY lags at zero growth in those periods but has shown explosive growth over 1 year. Revenue growth remains positive for all over 3-10 years.
Large Enterprise software companies like MSFT and SAP are not yet much affected by PE compression but that could change. MSFT is of course a big player in the AI revolution, of course.
MSFT
SAP
PE Ratio Current
31.6
31.67
PE Ratio 2020
33.99
22.62
PE Compression
-2.39
9.05
1-Year Operating Income Growth Rate (Per Share)
20.3
30.6
3-Year Operating Income Growth Rate (Per Share)
15.9
6.5
5-Year Operating Income Growth Rate (Per Share)
18.9
4
10-Year Operating Income Growth Rate (Per Share)
19.6
4.3
1-Year Total Revenue Growth Rate
15.6
9.7
3-Year Total Revenue Growth Rate
12.4
8.2
5-Year Total Revenue Growth Rate
14
4.6
10-Year Total Revenue Growth Rate
12.8
5.8
My overall impression from the above analysis is that business growth of the software houses remains healthy so the compression in the PE Ratio is likely due to other reasons, such as investor sentiment. PE ratios have dropped sharply since 2020: ADBE from 53.67 to 14.13 (74% decline), CRM from 45.85 to 19.71 (57% decline), NOW from 109.65 to 38.3 (65% decline), WDAY from 77.66 to 21.53 (72% decline). This suggests improved earnings relative to prices or sector-wide derating, making most appear cheaper on a forward basis.
Investor concerns center on AI eroding seat-based pricing models, boosting productivity to reduce software license needs, and enabling rivals like OpenAI, Anthropic, and startups to disrupt incumbents. Stocks like ADBE and CRM shed over 20-25% in value in the past year alone amid events like Anthropic’s “Cowork” launch and OpenAI’s enterprise push, with analysts citing AI competition and slow adoption of incumbents’ AI features. Goldman Sachs and others, note fears of margin compression and survival risks for SaaS giants.
Seat-based pricing in SaaS, which charges per user or license, is eroding due to AI-driven shifts toward usage-based, outcome-based, and hybrid models that better align costs with actual value delivered. This transition reflects maturing SaaS markets, cost-conscious buyers, and AI’s ability to automate tasks across fewer seats
AI agents and tools boost employee productivity, reducing the need for multiple per-user licenses as one human oversees automated workflows. For instance, AI handles routine tasks in CRM or project management, allowing firms to consolidate vendors and seats while questioning upgrades for marginal features. One human worker using AI Agents can over-see a larger part of a work-flow reducing the number of seats required.
Dynamic modelsbundling AI credits into base plans, usage tiers, or per-outcome feesemerged dominant in 2025-2026, with over 1,800 pricing changes signaling seats’ decline. Hybrid seat-plus-usage structures persist but face pressure as AI enables outcome-focused billing
AI programming tools like GitHub Copilot, Cursor AI, and Claude’s code interpreter are fueling a surge in software startups by slashing development time, costs, and barriers to entry. Domain experts and Non-technical founders can now prototype Minimum Viable Product (MVP) in days using no-code/low-code AI platforms like Lovable, Bolt.new, or Bubble with AI integrations, cutting engineering needs by 40-60% and enabling rapid iteration without hiring full dev teams. AI handles code generation, debugging, testing, and even full-stack apps, allowing solo founders or tiny teams to build complex software at 10x speed. Startups report scaling with half the team size, focusing humans on strategy while AI manages routine coding. Industry reports highlight AI coding adoption at 84% among developers in 2025, driving more launches in AI-powered niches like automation, e-commerce tools, and custom apps, as no-code AI has collapsed idea-to-MVP timelines to hours, boosting entrepreneurship amid maturing SaaS markets where incumbents face AI disruption.
Is AI now threatening Software Companies?
Adobe (NASDAQ:ADBE), ServiceNow (NYSE:NOW), and Salesforce (NYSE:CRM) all exhibit slowing growth in ARR/revenue metrics that proxy for seat expansion, though ServiceNow remains the strongest of the three, with high-teens to low-20s% growth. Adobe’s Digital Media ARR decelerated to 11.7% YoY in Q3 FY2025 from higher prior rates, with FY2026 guidance implying ~10% amid scale effects and AI delays, alongside subscriber growth to 35-40M but compressed ARPU signaling optimization rather than aggressive adds. ServiceNow’s subscription revenue rose 21% YoY to $2.87B in Q4 2024 with 19% cRPO, though FY2025 guidance of 19.5-20% trails some expectations and renewals dipped to 97%, supported by strong customer metrics like 12% growth in $1M+ ACV accounts. Salesforce saw FY2025 revenue up ~8.7% to $37.9B with workforce cuts of 7-9% annually, as Agentforce ARR hit $1.4B (114% YoY) but offset only modest broader subscription growth including 1% in Marketing/Commerce segments. These patterns reflect enterprise SaaS optimization trendsfewer incremental seats per value dollar via AIwithout explicit seat count disclosures.
There is only emerging factual evidence of slowing and optimizing seat-based licensing that vendors and buyers are shifting toward usage/hybrid models, and not yet a clean time series of seat licenses down x% across SaaS globally. Investors are not taking any chances and reducing their exposure to software in favor of other sectors and industries. They are now valuing the terminal value of these enterprises less richly and increasing the discount value of their DCF models to account for uncertainty.
AI is already eroding the traditional per-seat SaaS business model and will likely cause structural disruption, though it won’t eliminate seats entirelyinstead, it will evolve them into hybrid structures combining seats with usage – or outcome-based pricing. However this will take time as firms adjust to the new reality and there will be winners and losers.
Several professions have faded away due to dramatic productivity gains from technological advancements, which automated repetitive or manual tasks far more efficiently than human labor ever could. Switchboard operators, who once manually connected phone calls by plugging cables into jacks, became obsolete with automated dialing systems in the mid-20th century. Elevator operators, responsible for controlling lifts and assisting passengers, disappeared by the 1950s as self-service buttons and safety mechanisms took over. Human computers, who performed complex calculations by hand for science and engineering, were replaced starting in the 1940s by electronic calculators and digital machines. In manufacturing, pin setters manually reset bowling pins before automatic machines arrived in the 1950s, while assembly line workers for toys and vehicles declined with the rise of industrial robots that worked faster and more precisely. Slubber doffers, often children swapping bobbins in textile mills, vanished entirely with automated machinery. Service roles like typewriter repair technicians faded with personal computers, and film projectionists were supplanted by digital projectors. Now the same thing is starting to happen with taxi and truck drivers as self-driving vehicle technology mature. It is increasingly clear AI will affect almost every industry and software is top on the list. How these firms adapt to this change remains to be seen. Some will recover and some won’t. The market is signaling its uncertainty.