Key Takeaways
- • Chip spending represents only 25% of AI infrastructure investment, with the remaining 75% directed toward power management, cooling systems, and data center buildout—signaling a fundamental market rotation away from semiconductor-focused trades.
- • Vertiv Holdings posted Q1 2026 revenue of $2.65 billion (30% YoY growth) with adjusted EPS surging 83%, while maintaining a $15+ billion backlog extending visibility into 2027 and beyond; management raised full-year guidance to approximately $13.75 billion.
- • Bloom Energy achieved 130% YoY revenue growth to $751 million and crossed into profitability with $71 million in net income, while expanding its Oracle partnership to 2.8 GW capacity and securing a $25 billion financing arrangement with Brookfield.
- • Crypto mining operators are pivoting into AI data center hosting, with IREN securing a $3.4 billion Nvidia GPU computing deal and announcing a 5 GW infrastructure partnership, leveraging existing high-power facilities and energy provider relationships.
- • Power capacity (measured in gigawatts) and backlog growth are replacing chip sales as the critical performance metrics for infrastructure investors evaluating AI-related opportunities.
Summary
The artificial intelligence infrastructure investment thesis is undergoing a significant reorientation. Rather than concentrating on semiconductor manufacturing—historically the perceived bottleneck in AI deployment—institutional capital is increasingly flowing toward the foundational systems that support computational density at scale. Analysts now estimate that chips account for only 25% of total AI infrastructure spending, with the remaining 75% allocated to power management, thermal management systems, and data center construction. This architectural shift in investor focus is reflected in the recent quarterly performance of two companies positioned squarely within the non-chip infrastructure segment: Vertiv Holdings and Bloom Energy.
Vertiv, a manufacturer of power and thermal management systems for data centers, reported Q1 2026 revenue of $2.65 billion, representing 30% year-over-year growth with organic growth of 23%. The company's adjusted earnings per share surged 83%, demonstrating substantial operational leverage as AI infrastructure demand scales. Most significantly, Vertiv's backlog now exceeds $15 billion—a metric that provides exceptional visibility into future revenue streams extending well into 2027 and beyond. Management capitalized on this momentum by raising full-year sales guidance to approximately $13.75 billion.
Bloom Energy's financial results were even more pronounced. The company reported revenue of $751 million, a 130% year-over-year increase, and achieved a milestone by crossing into profitability with net income of approximately $71 million. Beyond headline growth, Bloom secured strategic partnerships that underscore market confidence in its positioning. The company expanded its existing partnership with Oracle to encompass up to 2.8 gigawatts of capacity and secured a $25 billion financing arrangement with Brookfield, indicating substantial capital availability for infrastructure expansion.
A noteworthy secondary trend within this infrastructure shift involves former Bitcoin mining operations. Companies with established expertise in high-power facility management and relationships with energy providers are repositioning themselves as AI data center hosts. IREN exemplifies this transition, having secured a $3.4 billion deal with Nvidia for GPU cloud computing services alongside a 5 GW infrastructure partnership announcement. This convergence of crypto mining infrastructure with AI compute demand represents a capital-efficient use of stranded assets and demonstrates the fungibility of power-intensive infrastructure across applications.
For institutional investors evaluating exposure to the AI infrastructure cycle, the relevant performance indicators have fundamentally shifted. Traditional chip sales metrics are no longer the primary focus; instead, power capacity (denominated in gigawatts) and backlog visibility serve as the principal valuation drivers. The $15+ billion backlogs observed at companies like Vertiv, combined with expanding partnership announcements and substantial financing arrangements at operators like Bloom Energy, suggest that infrastructure capacity constraints—rather than computational chip availability—have emerged as the binding constraint on AI deployment velocity.
Vertiv, a manufacturer of power and thermal management systems for data centers, reported Q1 2026 revenue of $2.65 billion, representing 30% year-over-year growth with organic growth of 23%. The company's adjusted earnings per share surged 83%, demonstrating substantial operational leverage as AI infrastructure demand scales. Most significantly, Vertiv's backlog now exceeds $15 billion—a metric that provides exceptional visibility into future revenue streams extending well into 2027 and beyond. Management capitalized on this momentum by raising full-year sales guidance to approximately $13.75 billion.
Bloom Energy's financial results were even more pronounced. The company reported revenue of $751 million, a 130% year-over-year increase, and achieved a milestone by crossing into profitability with net income of approximately $71 million. Beyond headline growth, Bloom secured strategic partnerships that underscore market confidence in its positioning. The company expanded its existing partnership with Oracle to encompass up to 2.8 gigawatts of capacity and secured a $25 billion financing arrangement with Brookfield, indicating substantial capital availability for infrastructure expansion.
A noteworthy secondary trend within this infrastructure shift involves former Bitcoin mining operations. Companies with established expertise in high-power facility management and relationships with energy providers are repositioning themselves as AI data center hosts. IREN exemplifies this transition, having secured a $3.4 billion deal with Nvidia for GPU cloud computing services alongside a 5 GW infrastructure partnership announcement. This convergence of crypto mining infrastructure with AI compute demand represents a capital-efficient use of stranded assets and demonstrates the fungibility of power-intensive infrastructure across applications.
For institutional investors evaluating exposure to the AI infrastructure cycle, the relevant performance indicators have fundamentally shifted. Traditional chip sales metrics are no longer the primary focus; instead, power capacity (denominated in gigawatts) and backlog visibility serve as the principal valuation drivers. The $15+ billion backlogs observed at companies like Vertiv, combined with expanding partnership announcements and substantial financing arrangements at operators like Bloom Energy, suggest that infrastructure capacity constraints—rather than computational chip availability—have emerged as the binding constraint on AI deployment velocity.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.