The State of Apps

The most successful AI companies are achieving $100 million in revenue at a pace that dwarfs the historical growth of even the fastest SaaS companies, and they’re doing it while spending less on sales and marketing. This staggering velocity, fueled by overwhelming end-user demand, defines the current AI boom. According to venture capitalist David George, the top performers grew nearly 700% year-over-year, generating up to $1 million per employee. This isn’t just a story of efficiency; it’s a testament to product-led growth where demand so dramatically outstrips supply that every new GPU is instantly maxed out.

Beneath this breakneck growth, however, are significant shifts in how companies must operate. Legacy businesses face an “adapt or die” imperative, needing to reimagine their products and internal processes around AI natively—not just by slapping on a chatbot. The real bottleneck for enterprise adoption isn’t the technology’s readiness but the immense challenge of organizational change management. Meanwhile, the business model itself is evolving, potentially moving from subscription and consumption-based pricing toward outcome-based models, where companies get paid only when an AI successfully completes a specific task.

The financial landscape presents a complex picture. While public market valuations for leading tech companies are high, they are largely supported by earnings growth rather than speculative froth, with AI driving a significant portion of market returns. On the supply side, a historic capital expenditure cycle is underway, primarily financed by the robust cash flows of hyperscalers like Microsoft and Meta, though debt is beginning to enter the system. Despite the scale of investment, the immediate and total utilization of new computing capacity suggests the underlying demand is very real. The conclusion is that we are still at the very beginning of a 10-15 year product cycle, with the full disruptive impact and value creation yet to unfold.

Surprising Insights

• Low gross margins can be a positive signal: For AI-native companies, lower gross margins—often due to high inference costs—can indicate that customers are actively using the AI features, which is a badge of honor. The assumption is these costs will fall over time, while high initial margins might suggest the AI isn’t core to the product’s value.
• The “electricity vs. blood” mindset: A founder-driven philosophy is emerging where every new task is evaluated by asking, “Can I do this with electricity (AI), or do I need to do it with blood (human labor)?” This represents a fundamental shift in operational thinking.
• AI is increasing, not decreasing, some professional workloads: Contrary to the automation narrative, one anecdote noted that corporate lawyers are seeing increased workload because LLMs empower every client to draft legal documents, requiring more lawyer time for review and correction.
• Older computing hardware retains extreme value: Seven-to-eight-year-old specialized chips (TPUs) still maintain 100% utilization, and prices for previous-generation GPUs like the A100 remain strong on the secondary market, indicating sustained, insatiable demand for compute.
• Disruption is accelerating: The average tenure of a company on the S&P 500 has declined by 40% over the last 50 years, highlighting how technology-driven disruption is happening at an increasingly rapid pace.

Practical Takeaways

• For any company, AI adaptation must be foundational, not incremental: Success requires reimagining workflows and products with AI at the core, not making superficial additions. Leaders must aggressively push for this change in both product development and internal operations.
• Prioritize and measure deep product engagement: For sustainable growth, focus on metrics like user time-in-product and task completion rates, not just top-line revenue. High engagement indicates real value and predicts strong retention.
• Embrace AI-powered development tools now: The leap in productivity from tools like Cursor and advanced coding models is so significant (10-20x faster in some cases) that product and engineering organizations not fully adopting them risk falling catastrophically behind within 12 months.
• Evaluate new tasks through an “AI-first” lens: Institute a practice of questioning whether any new process or capability can be accomplished with AI agents (“electricity”) before defaulting to human labor (“blood”).
• Legacy companies must study the new benchmarks: The efficiency of leading AI companies, measured by revenue per employee, sets a new standard. Incumbents need to understand that competing will require achieving similar operational leverage through AI integration.