1 Theory on Why the Software Stock Sell-Off Could Get Even Worse

The software sector is taking an absolute beating this week. As of Tuesday, the broader software market is reeling from a massive wave of panic selling.

Shares of industry stalwarts like Salesforce (CRM 6.50%) and ServiceNow (NOW 5.93%) dropped about 6% on Tuesday alone, while Microsoft (MSFT 2.77%) declined about 3%.

The catalyst for this sudden drop appears to be a renewed wave of fear regarding artificial intelligence (AI). Specifically, investors are increasingly concerned that advanced artificial intelligence (AI) agents, capable of autonomously executing complex tasks across a browser or enterprise environment, will inhibit the traditional per-seat software licensing models that most software-as-a-service companies rely on.

While it is tempting to quickly conclude that the massive declines across many software stocks are overdone, I have one theory on why the situation could get even worse.

That theory is simple: As these companies transition their models to ones better built for an AI-agent world, demand for AI features could remain high while the costs of running them rise even higher.

Image source: Getty Images.

Big demand, but even bigger costs

To be fair to the bulls, the leading software giants are proving that enterprise customers want their new AI features.

Salesforce’s AI-driven Agentforce platform, for example, is seeing explosive adoption. The product’s annual recurring revenue skyrocketed 169% year over year in fiscal Q4 to $800 million.

And ServiceNow is enjoying similar momentum. The workflow automation company noted that its Now Assist net new annual contract value more than doubled year over year in its most recent quarter.

Further, this strong product adoption at ServiceNow is fueling a healthy backlog of contracted future revenue. ServiceNow’s current remaining performance obligations, which represent contract revenue expected to be recognized over the next 12 months, hit $12.85 billion in Q4 – up 25% from the year-ago period. This exceeds the company’s fourth-quarter revenue growth rate of 20.5%.

Microsoft CEO Satya Nadella said during the company’s most recent earnings call that daily users of Microsoft 365 Copilot were 10 times higher than in the year-ago quarter.

But here is the catch.

Despite these impressive absolute numbers, these agentic features still represent a very small proportion of overall revenue for these massive corporations.

Additionally, the infrastructure required to support these features is incredibly expensive.

During this awkward transition phase, when demand for AI features is soaring, the costs of delivering these advanced AI features could rise much faster than the incremental revenue they generate.

Consider Meta Platforms’ changing growth profile. Its fourth-quarter revenue rose 24% year over year as it continues to inject AI across its social media apps, but earnings per share rose only 11% year over year as the company’s operating margin narrowed from 48% in the year-ago quarter to 41%. And even though management is guiding for an acceleration in top-line growth in Q1, it only expects its total 2026 operating income to be “above” 2025 levels. With the midpoint of Meta’s guidance calling for an incredible 30% year-over-year top-line growth rate in Q1, an operating income outlook like this is pretty disappointing.

While Meta’s business is inherently different from many software-as-a-service companies, this deleveraging still underscores how challenging it will be to scale in an AI-first era. If even Meta, with its global scale, can’t find a way to roll out AI without its margins narrowing significantly, it wouldn’t be surprising to see similar narratives show up at much smaller software companies.

The integration curve

And here is the part of my personal theory that isn’t widely discussed: This margin pressure could be further exacerbated by the customer reality on the ground.

Users and organizations could face a prolonged learning, integration, and implementation curve.

Deploying an autonomous agent to handle your customer service or legal workflows sounds great in a marketing pitch. But practically implementing that technology across thousands of employees requires extensive training, rigid governance, and careful data orchestration.

Additionally, AI is still in its early innings – and may not be as capable as some investors hope.

When agentic systems are deployed at scale, they may cause more issues and headaches than software stock protagonists anticipate in these early years. If early enterprise adopters, for example, experience severe hallucinations or broken workflows, the sales cycle for these premium AI add-ons will inevitably slow down.

Overall, this creates a tough dynamic for software providers: to stay competitive, they are forced to spend aggressively on compute infrastructure today, but their customers may take years to fully digest, trust, and pay for these tools, let alone pay a high enough price for these software companies to maintain their lucrative profit margins.

Expand

NASDAQ: MSFT

Microsoft

Today’s Change

(-2.77%) $-10.63

Current Price

$372.37

Key Data Points

Market Cap

$2.8T

Day’s Range

$371.86 - $382.47

52wk Range

$344.79 - $555.45

Volume

12K

Avg Vol

35M

Gross Margin

68.59%

Dividend Yield

0.93%

A difficult market to navigate

Make no mistake, some software-as-a-service stocks will likely find surprising ways to benefit from this platform shift. The companies that successfully integrate these agents at scale – and do so profitably – will inevitably accelerate their businesses, generating value well in excess of their new AI costs.

But picking those winners out of the wreckage today may not be easy.

Ultimately, investors should tread carefully. For most software stocks, I believe the best move right now is to stay on the sidelines and wait for clear evidence that these software giants can monetize their AI agents faster than they spend to support them.