Robert Smith’s Blueprint for the $1 Trillion Software Boom
Five Filters to Separate Software Winners from Losers After the $830 Billion Crash
TL;DR
- The software sector lost $830 billion in six trading days after Anthropic launched new AI tools that spooked investors about the future of paid software.
- Robert Smith, CEO of Vista Equity Partners, manages $100 billion in enterprise software and sees this crash as a shakeout — not an extinction event.
- Smith’s five-pillar framework identifies which software companies are built to survive the AI era and which ones face real danger.
- Companies that own proprietary data, deliver deterministic results, and capture services revenue could see a 10x jump in revenue per customer.
- Vista coined the shift from the “Rule of 40” to the “Rule of 70” — a new performance bar where growth plus EBITDA margin hits 70% for top-tier agentic companies.
David L. Berkowitz, Chief Investment Officer and Financial Advisor
Nearly 40 years of experience — from trading and research at a $250 million hedge fund in the early 1990s, to two decades as a portfolio manager, to teaching thousands of executives and employees how to create shareholder value through EVA and value-based management. Now helping individuals and families become shareholders through disciplined investing, concentrated portfolios, and direct stock ownership.
What Happened: $830 Billion Gone in Six Days
The software sector just had one of its worst weeks in memory. In six straight trading sessions, software and services stocks dropped roughly $830 billion in total market value. That number isn’t a typo. It’s nearly a trillion dollars.
Microsoft fell 10% in a single session despite reporting $81 billion in quarterly revenue. ServiceNow dropped 11% on the day it announced earnings that actually beat Wall Street’s expectations. Good numbers. Stock still crashed.
The spark was a product launch from Anthropic — the company behind the Claude AI model. Its new Cowork tool showed that AI could handle legal, marketing, and data tasks that companies currently pay software vendors to do. Investors panicked. If AI can do the work, why pay for expensive software subscriptions?
That fear makes sense on the surface. But it misses half the picture.
Robert Smith runs Vista Equity Partners. His firm manages over $100 billion and owns more than 90 enterprise software companies. He has spent three decades buying, building, and running software businesses. He is not panicking. He sees something most investors are missing.
Pillar 1: Data Is the Moat
Why does proprietary data matter more than AI models?
Smith’s first principle is simple. The AI model is not the real advantage. The data is.
Think about what ChatGPT, Claude, and Gemini were trained on. Social media posts. Reddit comments. Wikipedia pages. News articles. That’s consumer data. It’s public. Anyone can access it.
Now, think about what they were not trained on. Thirty years of insurance underwriting rules. Supply chain logic for a Fortune 500 manufacturer. Compliance frameworks that govern trillions in financial transactions. That data sits inside the software platforms these companies already use. No startup can copy it overnight.
Smith says the companies that keep this data on their own terms — that bring the AI model to the data instead of shipping the data out — are the ones with a lasting edge. He calls AI startups without this deep history “probabilistic guessers dropped into a vacuum.”
At VAP, we saw this play out with a client’s holding. The company had decades of proprietary workflow data baked into its platform. No competitor, no matter how good their AI, could match that overnight. The data itself was the barrier to entry.
Pillar 2: Beyond “Mostly Right”
Why can’t enterprises rely on AI that’s right 93% of the time?
Public AI models guess. They give you the most likely answer. For writing an email or summarizing a meeting, 93% accuracy works fine.
For a company wiring $14 million in payroll? 93% means someone doesn’t get paid. For an insurance underwriter pricing a policy in wildfire territory? “Mostly right” is a lawsuit waiting to happen.
Smith draws a hard line between two types of systems. Probabilistic means “best guess.” Deterministic means “correct, every single time.” Enterprise software demands deterministic results — especially in banking, insurance, and healthcare.
| Probabilistic AI (Public LLMs) | Deterministic Enterprise Systems |
|---|---|
| Gives the most likely answer | Gives the verified, correct answer |
| 93% accuracy — fine for email drafts | 100% accuracy — required for payroll, compliance |
| Trained on public internet data | Built on decades of proprietary business rules |
The winning companies are building an architecture that takes AI’s speed and forces it through a system that guarantees the answer is correct. They get both efficiency and accuracy. That’s the combination the market should be pricing in.
At VAP, this maps to how we pick companies. We look for management teams that understand the difference between a tech trend and a durable business process. The companies adding AI to mission-critical workflows with accuracy guarantees are compounding value. The ones tacking AI onto a pitch deck to impress analysts? Different story entirely.
Pillar 3: The 10x Revenue Multiplier
How does AI turn a $250,000 customer into a $2.5 million customer?
The panic story is straightforward: AI kills software revenue by replacing human workers. Fewer workers means fewer software licenses. Fewer licenses mean less revenue. For basic, commodity tools, that logic may hold.
But Smith sees a counter-move most investors ignore.
Here’s the old way: A company pays a software vendor $250,000 for seat licenses. Then it pays a separate services firm $5 million for the humans who actually run that software. The vendor only captures a small slice of the total spend.
Now, picture the agentic model. The software vendor doesn’t just sell the tool. It deploys autonomous agents that do the work the 100-person outsourced team used to handle. Revenue jumps from $250,000 to $2.5 million or more. Same customer. Ten times the revenue.
| Old SaaS Model | Agentic Model | |
|---|---|---|
| Software Revenue | $250,000 | $250,000 |
| Services/Agent Revenue | $0 (goes to outsourcer) | $2.25M+ (captured by vendor) |
| Total Vendor Revenue | $250,000 | $2.5M–$5M+ |
There’s also a demographic reality that most people miss. The federal workforce shrank by over 300,000 employees in 2025, and more than 100,000 federal workers retired that same year. Across government, experienced workers are leaving faster than they can be replaced. Agentic systems aren’t competing with workers in many sectors. They’re the only way to keep things running.
Smith calls this a shift from the old SaaS “Rule of 40” to a new “Rule of 70” — where growth plus EBITDA margin should hit 70% for the best companies. That’s a higher bar. And a much bigger reward.
We walked a client through this exact framework last quarter. They were nervous about their software holdings after the selloff. When we showed them which companies were positioned to capture services revenue versus which ones just sold seats, their view of the correction changed completely. Share prices dropped. Business value, for the right companies, did not.
Pillar 4: The 20/80 Rule
If AI can write code, why aren’t software engineers finished?
The media story about “vibe coding” goes like this: AI writes code now. Software engineers are done. Developer tools are dead.
Smith has real data on this, from more than 90 portfolio companies in Vista’s network. AI tools have delivered 30% to 50% productivity gains in writing code. That part is real.
But writing code accounts for only about 20% of the total work. The other 80% is maintenance. Connecting with existing systems. Customer-specific changes. Regulatory compliance testing. Moving data off old platforms.
A chatbot that writes Python faster doesn’t solve any of that.
This is Smith’s 20/80 Rule. The 20% that AI disrupts is the most visible part. The 80% that stays is the part that actually keeps enterprise software running. Deep, vertical platforms that own integration, maintenance, and customer-specific logic sit firmly in that 80%. Horizontal tools — the commodity “shovels” — are the ones at risk.
Pillar 5: Engineer for Constant Evolution
What separates software companies that survive from those that don’t?
Smith’s last principle is a filter for management quality. He says companies with generic workflows that any AI can copy “have no right to exist.” That’s a direct quote.
The survivors treat their technology like a living system. Smith calls it the “Agentic Factory” — a model that pushes new capabilities through the business nonstop. Vista built this as a collaboration with Microsoft, Google, AWS, and Anthropic, and more than 30 of its companies are already generating revenue from agentic AI.
Smith breaks agentic systems into two categories that matter:
Always-On Monitoring Agents. High-volume, low-complexity systems that watch for specific triggers around the clock. Fraud detection. Compliance monitoring. Supply chain alerts. Tasks where the speed and scale make human oversight impossible.
Highly Orchestrated Agents. High-complexity systems where multiple agents work together to meet strict regulatory and government reporting requirements. These demand deep domain knowledge baked into the architecture.
Both types reward companies with proprietary data and institutional expertise. Both types punish companies that rely on generic, copyable features.
This connects directly to something we talk about at VAP: the agency problem. When management teams invest in building these deep, proprietary systems, they compound business value. When they chase AI headlines without a real structural advantage, they burn capital. Investors who know the difference between the two can make better decisions about what to hold and what to sell.
The Five Filters for Your Portfolio
The $830 billion selloff is real. The fear is understandable. But fear and value destruction are not the same thing.
Smith’s framework gives you five questions to ask about every software company you own:
- Does it own proprietary data — or rent commodity features?
- Can it deliver deterministic (100% accurate) results — or just probabilistic guesses?
- Is it positioned to capture service revenue—or stuck selling seats?
- Does it control the 80% of work that AI can’t replace?
- Is management building a living system — or standing still?
Those five questions apply to every software company in your portfolio right now. The answers tell you which positions to hold through the noise and which ones need a harder look.
If this framework is useful to you, take the next step. We built a full breakdown of the three types of companies we look for at VAP — Ownership Companies, Monopolies, and Cannibal Companies — and how each compounds wealth in different ways over time.
IMPORTANT DISCLAIMER — NOT FINANCIAL ADVICE
This article presents the general views of ValueAligned Partners and is provided for educational and informational purposes only. Nothing here should be regarded as personalized investment advice or a recommendation. Before making any investment decision, consult with qualified professionals — a fiduciary financial adviser, CPA, and attorney. All investing involves risk, including the potential loss of principal. Past performance is not a guarantee of future returns. ValueAligned Partners LLC is a Registered Investment Advisor. For additional information, refer to Form ADV at www.adviserinfo.sec.gov.
Endnotes
-
Yahoo Finance / Reuters — Selloff coverage reporting $830 billion wiped from software sector in six sessions. Source
-
CNBC — Vista Equity ‘Agentic Factory’ — Robert Smith interview on Vista’s AI transformation. Source
-
Private Markets Insights — Detailed analysis of Smith’s framework: Rule of 70, data sovereignty, agent workforce shift. Source
-
AI2Work — SaaS apocalypse analysis, Microsoft 10% drop, $285B single-day wipeout from Claude Cowork launch. Source
-
Salesforce Ben — ServiceNow stock drops 11% on earnings day despite beating expectations. Source
-
CNBC — Software Selloff — Coverage of Anthropic Claude Cowork triggering software stock declines. Source
-
StartupHub AI — Smith on deterministic vs. probabilistic systems and 30–50% productivity gains in code. Source
-
Federal News Network — Workforce Cuts — Federal workforce shrinkage of 317,000 employees in 2025. Source
-
Federal News Network — Retirements — 105,858 federal employees retired in 2025. Source
-
Vista Equity Partners — Beyond the AI Bubble — Official perspective on agentic AI and Rule of 60/70 metrics. Source
-
ValueAligned Partners — Ownership, Moats, and Cannibals: A Framework for Value-Aligned Companies. Source
-
Morningstar — Software Selloff — Analyst Dan Romanoff’s assessment of software fundamentals. Source
-
Fortune — $2 Trillion AI Wipeout — J.P. Morgan data on $2 trillion wiped from software market caps. Source
-
S&P Global Market Intelligence — Analysis of SaaS growth deceleration and AI headwinds. Source
-
Vista Equity Partners — Software’s Next Chapter — Smith on market expansion for agentic models. Source
