Conference Presentation, Interview, Webinar
AI, Infrastructure, and the Next Investment Cycle
- Market conditions following a 90% rise over four years (17% annualized) are expected to necessitate a correction, while global infrastructure investment needs are projected to reach $90 trillion by 2040 across power, water, roads, and transit.
- Hyperscaler capital expenditure is forecast to rise from $416 billion in 2025 to $780 billion in 2026, exceeding $1 trillion annually starting in 2027, with cumulative AI figures potentially reaching 20 times current levels within five to ten years.
- Free cash flow recovery for hyperscalers is anticipated beginning in 2028, despite supply chain delays for certain materials and products that may not be accessible until 2028.
- AI adoption remains in early stages with live deployments at 69% of S&P 500 companies and only 2% tracking the "ultimate barometer" metric, while current spending shows median users paying $200–$400 monthly compared to $7,500–$9,000 for power users.
- Enterprise budget allocation for AI tools ranges from approximately 1% for forward-leaning Fortune 500 companies to 10% for AI-native portfolio companies, with the top 1% of users spending roughly eight times the level of the top 10%.
- Specific company performance data indicates Chime achieved a compounded cost-to-serve reduction of nearly 50% over four years, Shopify saw an 8% growth in customers reaching five orders within 15 days after launching an AI sidekick, and ServiceNow reported over $1 billion in AI ACV alongside a 9x increase in agentic deployments.
- Cost efficiencies are projected across various AI applications, including a 35% lower cost for Databricks' smart router compared to the strongest individual model, a 60% reduction for Elise AI's fine-tuned models, and a 10x reduction in costs for financial chat workloads at Hebia.
- Consumer subscription penetration for AI remains low at just over 2% of U.S. households, contrasting significantly with Amazon Prime's 200 million and Netflix's 70 million subscribers, while platform strategies diverge regarding agent integration such as Amazon's acceptance of Instacart versus rejection of Muse.
- Advertising revenue risks are highlighted for Amazon if consumer search behaviors shift, and Google's monetization model faces potential disruption if AI enables autonomous user actions, contributing to a bifurcated software market where the index has returned to its year start.
- Profitability and growth metrics for public software show 75% of companies are profitable and 30% grow at 20% or more, a figure projected to drop below five if the growth threshold is raised to 30%, while investors target a 10% revenue growth acceleration achievable over the next 12 to 18 months.
- Cybersecurity and vertical software are expected to outperform horizontal applications as AI drives new security needs, with vertical AI companies like Harvey, Bridge, and A Lease growing faster than industry precedents.
- The combined valuation of the top six private AI companies stands at approximately $2.4 trillion, accompanied by 58% employee participation in Carta tenders and secondary market discounts near zero for fresh valuations.
- AI-related deal activity accounted for 86% of U.S. VC activity in a 2026 snapshot, up from 65% in 2025, with investment focus shifting toward robotics expected to expand three to five years earlier than LLMs.
- Autonomous driving networks are projected to expand by an order of magnitude, with 17 million new U.S. cars expected to become autonomous over the next decade, while significant progress in drug discovery and personal health is anticipated within ten years.
- Enterprise AI diffusion is expected to move deeply into workflows beyond coding, and investment in American dynamism, currently under 5% of overall spending, is set to grow dramatically as needs evolve.
- Residential electricity rates may decline by 40 basis points for every 10% increase in data center capacity, and Meta has confirmed working with communities to lower electricity costs at its Louisiana site.