🔬 Case Study: Enterprise AI / Data Infrastructure

The Data Infrastructure Play: Government Contracts and Data Network Effects

📅 Original Analysis: Q1 2021 📊 Status: Ongoing (publicly traded) ⏱️ 14 min read
⚠️ Educational Analysis — Historical

This case study reflects my thinking at a specific point in time (Q1 2021). It is NOT a current recommendation. The company discussed may have materially changed since this analysis was conducted. This is presented for educational purposes only — to demonstrate how I evaluate enterprise software businesses with government exposure, not to suggest any investment action. Markets change; so do companies. Past performance doesn't predict future results.

The Setup

In early 2021, I was examining a data analytics company that had become synonymous with government intelligence work. The company had gone public via direct listing in late 2020, and the stock had run up significantly on AI hype. The question wasn't whether the technology was real — it clearly was. The question was whether the business model could translate into durable shareholder value.

The company occupied a unique position: providing data integration and analytics software to intelligence agencies, defense departments, and increasingly, large commercial enterprises. The numbers told a polarizing story:

$1.1B
Annual Revenue
47%
Revenue Growth
-$1.2B
Net Income
125%
Net Dollar Retention

The 125% net dollar retention was exceptional — existing customers were expanding rapidly. But the losses were staggering. Stock-based compensation alone exceeded revenue. This was either a generational opportunity or a value trap masquerading as a growth story.

The Central Question

Can a company built on bespoke government deployments — with heavy professional services and long implementation cycles — ever achieve software-like margins? Or is this fundamentally a consulting business with software economics priced in?

The Framework Applied

Moat Assessment

🔒

Switching Costs: Very Strong (Government), Moderate (Commercial)

For intelligence agencies, switching is nearly impossible. The software becomes embedded in critical workflows, handles classified data, and requires extensive security clearances for support. Years of institutional knowledge live in the system. For commercial customers, switching costs are high but not insurmountable — more like enterprise software generally.

🌐

Network Effects: Emerging

This is where it gets interesting. Each deployment generates data about what works — which integrations, which workflows, which analytics patterns. The company was building what I call a "deployment knowledge network." The more they deployed, the faster and better subsequent deployments became. Not a traditional network effect, but a compounding advantage.

🏛️

Regulatory/Security Moat: Strong

The security clearances, government certifications, and reputation for handling classified data create barriers competitors can't easily overcome. You can't just decide to compete for CIA contracts — you need decades of trust-building.

💰

Cost Advantages: Weak

No meaningful cost advantage. In fact, the heavy professional services component was a cost disadvantage compared to pure SaaS competitors.

The Bull vs. Bear Debate

🐂 The Bull Case

  • Government revenue is sticky and growing
  • Commercial segment accelerating (now 50%+ of revenue)
  • Platform approach means expanding use cases
  • Data integration is becoming more critical, not less
  • AI tailwinds — they were doing AI before it was cool
  • Once deployed, customers expand dramatically (125% NRR)

🐻 The Bear Case

  • Stock-based compensation is egregious
  • Profitability always "2 years away"
  • Heavy services mix means lower margins than SaaS
  • Customer concentration risk
  • Valuation assumes perfection
  • Insider selling was persistent

Key Questions I Needed to Answer

What Happened Since

The subsequent years validated some aspects of the thesis while challenging others:

The Outcome (Through 2024)

  • Revenue scaled significantly — from $1.1B to over $2.2B, validating the growth trajectory
  • Profitability arrived — the company achieved GAAP profitability, proving the model could work
  • Commercial became dominant — commercial revenue now exceeds government, diversifying the customer base
  • AI boom provided tailwinds — the company's AI/ML capabilities became more valued as the market embraced AI
  • Stock performance was volatile — investors who bought at the 2021 highs experienced painful drawdowns; those who bought during the 2022 lows saw significant gains
  • SBC moderated but remains elevated — better than feared, worse than hoped

The company proved it could grow and eventually profit. But the journey was far from smooth — the stock fell over 80% from peak to trough before recovering. Timing and valuation discipline mattered as much as thesis quality.

The Takeaway

🎯 Framework Lessons

  • Government contracts create real moats, but they're slow-moving. The security and switching cost advantages are genuine, but they don't prevent stock price volatility driven by sentiment shifts.
  • Professional services businesses can become software businesses, but it takes longer than bulls expect. The productization journey is measured in years, not quarters.
  • "AI company" isn't a moat. Having AI capabilities is table stakes. The moat comes from distribution, data, and customer lock-in — not the technology itself.
  • Stock-based compensation is a real cost. When SBC exceeds 50% of revenue, shareholders are being massively diluted. Management incentives may not align with shareholder outcomes.
  • High net dollar retention is necessary but not sufficient. 125% NRR is excellent, but if customer acquisition costs are too high or churn eventually appears, the math breaks.
  • Valuation provides margin of safety. The same business at $25B market cap was a reasonable investment; at $50B market cap, it was speculation. The business didn't change — the price did.

This case study illustrates how I evaluate enterprise software businesses with government exposure — not to suggest what you should buy or sell. Every investment decision depends on your circumstances, timeline, and risk tolerance. If you'd like to discuss how these frameworks apply to your portfolio, let's talk.

← All Case Studies Next: The Language Learning Compounder →