The AI investment narrative is rapidly pivoting from a singular focus on training hardware to the companies enabling the second wave of inference and agentic AI at scale.
The first wave of the artificial intelligence boom, dominated by Nvidia Corp.’s near-90 percent market share in training hardware, is giving way to a second, more structurally diverse phase focused on deployment. As capital pivots from training to inference and agentic AI, a new class of companies enabling real-world AI applications is moving into the spotlight, offering investors a different risk and reward profile beyond a handful of hardware giants. This new phase emphasizes not just the creation of AI models, but their practical, cost-effective application across industries.
"The next wave of AI will bring intelligence closer to the end user, moving from foundational models to inference to agentic," Intel CEO Lip-Bu Tan said on the company's recent earnings call. "This shift is significantly increasing the need for Intel's CPUs and wafer and advanced packaging offerings."
This pivot highlights the critical roles of specialized software, alternative silicon, and the foundational infrastructure required to power AI at a global scale. While the first wave was about the raw horsepower for training, the second is about the efficiency, cost, and specific application of that intelligence. Companies like Cellebrite DI in specialized software, Intel in inference hardware, and CleanSpark in data center infrastructure represent three distinct layers of this emerging ecosystem.
For investors, this second wave diversifies the AI landscape. It presents opportunities in companies that are not just building the AI engines, but are also crafting the applications, tuning the hardware for specific tasks, and laying the physical groundwork for the entire system, creating a broader and more complex investment thesis.
The Software Layer: Cellebrite's Vertical AI Play
Cellebrite DI (NASDAQ:CLBT) is demonstrating how to monetize the second wave by embedding sophisticated AI into vertical-specific software. The digital intelligence provider’s first-quarter 2026 results showed a 21 percent year-over-year increase in annual recurring revenue to $493 million, driven by a suite of new AI-powered products. The company’s new Genesis platform, an “agentic next-gen AI solution,” is designed to analyze complex digital evidence for law enforcement.
The product quickly attracted more than 500 registered users across 15 countries within eight weeks of its early access launch, according to CEO Tom Hogan, despite having no marketing or pricing attached. Hogan estimated the total addressable market for investigative AI at roughly $12.5 billion over the next four years, suggesting that if Cellebrite executes well, its AI revenue could approach the company's entire current revenue. This highlights a key aspect of the second wave: the market rewarding companies that can successfully apply AI to solve specific, high-value business problems.
The Silicon Layer: Intel's Inference Revival
Intel (NASDAQ:INTC) has seen its stock surge 116 percent in one month, fueled by a spike in demand for its central processing units (CPUs) for AI inference tasks. While Nvidia’s GPUs dominate the energy-intensive training phase, Intel’s CPUs are finding renewed purpose in the inference market, where models run in real-time to generate answers. This distinction is critical for the second wave, where the sheer volume of inference queries is expected to dwarf training workloads.
Reports of a preliminary chip manufacturing agreement with Apple have further bolstered confidence in Intel’s turnaround strategy. While the stock trades at a high 119x forward P/E ratio, the combination of a structural AI tailwind, government support via a $5.7 billion CHIPS Act grant, and strategic co-development partnerships with Google and Nvidia for its Xeon processors provides a compelling, albeit high-risk, narrative. Intel’s comeback is a direct bet on the growing importance of inference in the AI ecosystem.
The Foundation: CleanSpark's AI Data Factories
The AI boom is as much about energy and real estate as it is about silicon. CleanSpark, Inc. (NASDAQ:CLSK), a company with a heritage in bitcoin mining, is transforming itself into a digital infrastructure and data center developer to meet this demand. Management sees AI as a technology wave “denominated in compute,” which is fundamentally a function of access to energy and data centers.
CleanSpark holds 1.8 gigawatts of currently contracted power capacity and is actively marketing its 250-megawatt site in Sandersville, Georgia, for a full greenfield data center build. The company’s strategy is to use its profitable mining operations to fund the acquisition and development of these power-rich sites, then monetize them through long-duration leases to high-quality AI tenants. This “picks and shovels” approach offers a different way to invest in the AI buildout, focusing on the non-negotiable need for power and infrastructure that underpins the entire digital economy.
This article is for informational purposes only and does not constitute investment advice.
