The “picks and shovels” AI power play represents a fundamental shift in how investors and engineers are thinking about the artificial intelligence boom. Rather than betting directly on AI companies or model makers, this strategy focuses on the essential infrastructure providers—the companies that supply the components, systems, and foundational tools that make AI possible. Navitas Semiconductor (NVTS) exemplifies this approach, positioning itself as a critical supplier of power delivery technology for AI data centers rather than competing in the AI software or services space itself.
Think of it this way: during the Gold Rush, the most reliable profits often came not from miners themselves, but from those selling the picks, shovels, and other mining equipment. NVTS has gained significant market traction in this space by specializing in 800V to 6V GaN (gallium nitride) power delivery boards specifically designed for NVIDIA-based AI data centers. This isn’t a flashy product that captures headlines, but it’s absolutely essential—and that’s precisely what makes it valuable. The company’s recent surge in analyst attention and board-level changes reflect growing recognition that power management will be one of the defining constraints in how quickly and efficiently AI infrastructure scales.
Table of Contents
- What Makes the Picks and Shovels Strategy Different from Direct AI Investment
- Navitas Semiconductor’s Technical Position in 800V Power Architecture
- Power and Cooling as Bottlenecks in AI Infrastructure Expansion
- Recent Market Recognition and Analyst Validation
- Risks and Competitive Pressures in Power Delivery Infrastructure
- GaN Technology and the Semiconductor Manufacturing Challenge
- The Broader Infrastructure Buildout and Market Timing
- Conclusion
What Makes the Picks and Shovels Strategy Different from Direct AI Investment
The picks and shovels framework inverts the traditional investment narrative. While everyone knows about ChatGPT, Claude, or Gemini, fewer people recognize that these systems require massive, specialized hardware ecosystems to function. That’s where companies like Navitas come in. Instead of competing directly in a crowded AI software market, they supply the critical infrastructure components that every AI deployment requires.
This indirect approach carries a subtle but important advantage: the demand signal is distributed across multiple AI companies and data center operators, reducing dependency on any single player’s success. The infrastructure approach also insulates companies from some of the volatility plaguing pure AI software plays. When investors question whether a particular AI model will achieve product-market fit or whether a new startup’s capabilities justify its valuation, the picks and shovels suppliers benefit from the fact that *every* AI scenario—whether successful or failed—still requires power delivery, cooling, networking, and other foundational systems. The infrastructure gets built regardless of which AI company ultimately dominates. However, this model does create its own risk: if the pace of AI data center buildout slows unexpectedly, infrastructure suppliers face a demand cliff with limited ability to pivot.
Navitas Semiconductor’s Technical Position in 800V Power Architecture
Navitas has positioned itself at the intersection of a critical technical transition in data center power delivery. Traditional silicon-based power systems operate at lower voltages and generate more heat, creating thermal management problems at the scale of modern AI clusters. The shift toward 800V architecture addresses both efficiency and thermal challenges simultaneously. By stepping down voltage at the point of use (800V to 6V at the board level), Navitas’s GaN-based solutions reduce I²R losses—the energy wasted as heat during voltage conversion—compared to older approaches. This matters enormously in a data center environment where cooling costs can represent 20-40% of total operating expenses.
GaN technology specifically offers advantages over traditional silicon that go beyond simple efficiency gains. The wider bandgap of gallium nitride allows faster switching frequencies, which means smaller, lighter power delivery components. In densely packed AI server configurations where physical space constraints are becoming a real problem, this reduction in size translates to tangible operational benefits. The limitation worth noting here is that GaN components historically carried cost premiums over silicon equivalents. While that gap has narrowed as manufacturing scale increases, widespread adoption of 800V GaN architecture still represents a cost decision that data center operators must justify through efficiency gains and reduced cooling requirements.
Power and Cooling as Bottlenecks in AI Infrastructure Expansion
Power generation and cooling have emerged as the primary constraint limiting AI data center expansion, according to recent infrastructure analyses. This isn’t a theoretical concern—it’s a real, immediate problem. Major cloud providers are already reporting that available electrical capacity and cooling capability, not server availability or GPU supply, are what’s holding back their ability to deploy new AI clusters. This reality fundamentally changes which infrastructure suppliers matter most. A company that solves a bottleneck problem has pricing power and demand visibility that even a company with technically superior technology might lack if that superior technology addresses a non-constraint.
Navitas benefits from entering this market precisely when power efficiency is moving from a nice-to-have to a business-critical feature. Data center operators are actively seeking any technology that lets them deploy more AI compute without proportionally increasing their power draw or cooling requirements. Real-world deployments of 800V AI data center systems have reported 10-15% improvements in overall power efficiency compared to traditional 12V or 48V approaches. However, the transition to new power standards also creates adoption friction—existing server designs, power distribution architectures, and cooling systems were built around older voltage standards. Switching to 800V requires coordinated redesigns across multiple system layers, which slows initial penetration even when the technical benefits are clear.
Recent Market Recognition and Analyst Validation
In May 2026, Baird research more than doubled its price target for Navitas from $9 to $20 per share, issuing an Outperform rating explicitly focused on the company’s positioning in what they identified as “three waves” of secular growth in 800V AI data center power architectures. This analyst upgrade represents genuine validation of the picks and shovels thesis—an analyst comfortable enough with the company’s trajectory to make a public, quantified commitment. The timing and magnitude of the upgrade suggest that major institutional investors were underestimating both the pace of 800V adoption and Navitas’s market share potential. The month prior, in April 2026, Navitas appointed Gregory M. Fischer to its board of directors.
Fischer brings more than 40 years of semiconductor industry experience, including senior roles at Broadcom and active advisory positions with AI-focused technology firms. This board-level appointment signals that Navitas is thinking seriously about infrastructure and strategic positioning. A veteran semiconductor executive with deep connections to both data center and AI companies provides valuable guidance on market timing, technology roadmaps, and potential partnerships. The appointment also creates a signal to the market that sophisticated industry insiders believe in the company’s opportunity. That said, board appointments are often forward-looking—they represent what management wants the market to believe about future direction, not necessarily current operational performance.
Risks and Competitive Pressures in Power Delivery Infrastructure
The picks and shovels strategy, while conceptually sound, operates in markets where competition is asymmetrical and high. Navitas competes against established semiconductor giants—Intel, TI, and others—that have massive R&D budgets, existing customer relationships, and financial resources that dwarf a company of Navitas’s scale. Intel’s transition to foundry services and aggressive push into power management technology represents a direct competitive threat. These larger competitors can afford to operate at lower margins while still funding product roadmaps, a dynamic that pressures smaller, more specialized suppliers.
There’s also a structural risk in betting that 800V becomes the dominant standard. Technology adoption in data centers often involves multiple competing approaches coexisting for years or decades. Just as modern data centers still use multiple voltage standards across different applications, it’s entirely possible that 800V becomes one important standard among several rather than the dominant architecture. If the efficiency gains of 800V prove good-but-not-transformational, adoption could plateau short of the levels that would justify Navitas’s current valuation assumptions. Additionally, as 800V adoption increases, pressure on pricing is inevitable—the profit margins that make the business attractive today may compress as the market becomes larger and more competitive.
GaN Technology and the Semiconductor Manufacturing Challenge
Gallium nitride components require different manufacturing processes and expertise compared to traditional silicon power semiconductors. This creates both opportunity and vulnerability. For Navitas, it means they’ve chosen to compete in a domain where legacy semiconductor expertise matters less than deep knowledge of GaN-specific design and process optimization. However, manufacturing GaN at scale requires specialized fabrication capacity, and Navitas doesn’t own fabs—it relies on foundry partners.
Any constraint in GaN foundry capacity could limit Navitas’s ability to scale production with demand. This contrasts with vertically integrated competitors like Intel or Samsung, which control their own manufacturing capacity and can prioritize internal demand. The GaN manufacturing ecosystem is improving rapidly, with multiple foundries expanding capacity specifically for power semiconductor applications. This is good for the long-term health of the market but potentially reduces Navitas’s competitive moat. As more foundries become competent at GaN manufacturing, other companies can more easily enter the space or existing competitors can shift capacity into power delivery if strategic priorities change.
The Broader Infrastructure Buildout and Market Timing
The picks and shovels play on AI infrastructure is fundamentally a bet on the continued acceleration of data center buildout. If AI training and deployment continue on their current trajectory, data centers will require thousands of additional megawatts of power capacity over the next 3-5 years. Every watt of additional AI compute requires supporting power delivery, cooling, and networking infrastructure. This creates a multi-year tailwind for infrastructure suppliers regardless of which specific AI applications ultimately prove commercially successful.
The beauty of the infrastructure play is that it doesn’t require predicting which AI application wins—only that the overall adoption curve continues upward. Looking forward, the competitive landscape will likely consolidate around a few specialized suppliers that achieve both technical leadership and manufacturing scale. Navitas’s current position as a focused player in 800V GaN power delivery gives it credible positioning for that consolidation, but only if the company can maintain technical differentiation while scaling production. The analyst upgrade and board appointment suggest confidence in that trajectory, but execution risk remains real.
Conclusion
NVTS represents a classic picks and shovels play in the AI infrastructure buildout—a company solving a real bottleneck problem (power efficiency and thermal management) with specialized technical expertise rather than competing directly in AI software or services. The company’s 800V to 6V GaN power delivery technology addresses a genuine constraint facing data center operators, and recent analyst validation suggests the market is beginning to recognize this opportunity more broadly. The Baird price target doubling and high-profile board appointment indicate that sophisticated investors and industry insiders believe in the company’s long-term positioning.
However, the opportunity comes with real competitive and structural risks. Navitas operates in a space where larger competitors can sustain pressure, where technology standards are still partially unsettled, and where manufacturing capacity remains constrained. Success requires executing flawlessly on product development, managing foundry partnerships carefully, and maintaining technical differentiation as the market scales. For investors or engineers evaluating the picks and shovels thesis on AI infrastructure, Navitas is worth understanding—not as a guaranteed winner, but as a company positioned at a real bottleneck in an industry that’s doubling down on capital expenditure.
