iCapital Market Pulse: AI’s Price Tag Is Rising

Key Takeaways:

• Hyperscalers are boosting capex spend and therefore are tapping capital markets at a greater rate given how much this spending is draining operating cash flow.
• Software companies are facing increasing dispersion as investors raise the bar for results, given concerns around AI disruption and changing models of pricing given token economics.
• It is still early to see AI’s impact on jobs and executive leadership, but workplace shifts mark evidence of companies beginning to make tradeoffs between capex spend and headcount.
• We remain bullish on prospects for AI adoption and productivity gains caused by such uptake, though we believe there will be wider dispersion across single name stocks and private funds invested across the sector.

The AI trade has been met with tremendous uncertainty, with split performance across the hyperscalers and diverging sentiment across hardware and software. Yet as the exponential rate of AI adoption becomes apparent, the question becomes: Who wins in a world where AI begins to sweep across corporate America?

Business spend on AI services is rising, with over half of businesses in the U.S. paying for AI subscriptions and pure dollars committed to the AI trade rising through the first quarter.¹ Unit economics are surging given token pricing and an uptick in token usage due to increased adoption of agentic AI, with Alphabet announcing a 60% increase in token spend over last quarter and Amazon reporting a 170% increase.² With the changing nature of how AI software is sold, based on usage rather than licenses, the opportunity may exponentially increase to an estimated $5.5 trillion across the market.³

We believe adoption and pricing trends will be increasingly important indicators that companies can continue securing the dollars needed to stay competitive in a fast-changing industry. Given the rising spend required, we continue to focus on names with fortress balance sheets (those that prioritize high liquidity, low leverage ratios, and robust capital reserves) and the ability to control their supply chains.

Hyperscalers

Hyperscaler capex spend keeps rising, consistently defying projections. Amazon, Microsoft, Alphabet, and Meta collectively spent roughly $400–$450 billion in 2025 and are projected to exceed $600 billion in 2026, with an estimated 70–75% of that spend now tied directly to AI infrastructure.⁴ Amazon, which spent more than expected, and Alphabet exceeded top line estimates for the first quarter, proof that prior capex investments are beginning to accelerate growth.⁵ However, unlike prior cycles, spend is being front loaded to secure scarce inputs such as compute, power, and real estate reflecting a shift from demand matching to platform control.

The cash flow impact of this shift is increasingly visible. In 2025, hyperscalers directed close to 70% of operating cash flow into capex, up from roughly 40% in 2023, materially compressing free cash flow flexibility even as absolute cash generation remains strong (Exhibit 1). This limits flexibility and could restrict the ability of select hyperscalers to conduct share buybacks—historically a key facet in supporting their stocks’ performance.

Concerns around sustainability in capex spend are also rising. Hyperscalers have supplemented internal cash generation with debt financing, marking a subtle but important evolution in capital structure strategy. Consensus forecasts call for even higher capex beyond 2026, raising questions around marginal returns if revenue growth or margin leverage does not materialize in parallel.

While equity market performance has remained mixed among hyperscalers given such concerns around ROI, debt investor appetite has remained robust for new deals (Exhibit 2). Meta, for example, raised its 2026 capex projections, and sought to tap debt investors for more than $20 billion to finance the spend—though equity investors did not take this move kindly.⁶ The stock fell more than 10% at the open after Meta reported results, the largest drop since investors began to grow more concerned about mounting capex in October of 2025.⁷

While elevated capex remains a demand signal, rising investment alone is no longer sufficient to support valuation expansion. Today, sustained capex spending without clear evidence of operating leverage risks being viewed as the cost of staying competitive without a path towards “winning”—or emerging as a key beneficiary of expanding AI adoption. Against this backdrop, hyperscalers that can articulate a credible path from infrastructure dominance to durable cash flow generation are likely to maintain investor confidence.

LLM providers

Even large language model (LLM) providers are not immune to scrutiny in this environment. On April 28, after the Wall Street Journal reported that OpenAI missed an internal goal of reaching one billion weekly active users by end of 2025, the Nasdaq Composite Index declined over 1%.^8,9 Within a day, OpenAI partners like Nvidia, Oracle, and SoftBank Group experienced drops in stock price up to 10%.¹⁰ OpenAI also faces key investor concerns as it looks to compete in business adoption. Only 81% of organizations who have a generative AI vendor use OpenAI, down from 94% last April.¹¹

Anthropic has taken advantage of this decline, increasing its adoption rate from 10% to 63%.¹²  Google Gemini’s uptake by businesses remains modest, with similar growth rate in market share as OpenAI over the last six months.¹³

Meanwhile, newer generations of models are becoming more expensive. The most recent widely available Claude model, Opus 4.7, uses up to 1.35x as many tokens when processing text compared to previous models.¹⁴ According to Coatue Management, the addressable market for SaaS (software as a service) expands by 25x beyond the existing software industry today with this shift towards a “services-as-software” paradigm.¹⁵

Software

The outlook for software market remains cloudy with the uncertainties that AI adoption brings—$2 trillion in market cap has been lost from publicly-traded software companies in 2026.¹⁶ We previously wrote about our views on AI disruption risk, noting that the application layer is more vulnerable given the lack of proprietary data and potentially less sticky enterprise customer relationships. As this cycle matures, software performance is likely to bifurcate based less on headline growth rates and more on competitive positioning in an AI first operating environment.

Recent earnings reactions underscore this dynamic. In the first quarter of 2026, shares of both ServiceNow and IBM fell sharply—down 18% and 8%, respectively—immediately following earnings, despite delivering results broadly in line with consensus expectations.¹⁷ Despite stable near‑term demand and solid execution metrics, both companies failed to sufficiently reassure markets that AI adoption would translate into accelerating revenue growth rather than defensive repositioning. In both cases, the market response was less about reported results and more about the absence of clear evidence that AI is enhancing pricing power, stickiness, or unit economics in a durable way.

The implications extend beyond public markets, as direct loans to vintage-risk software companies face downward valuation pressure. Valuation compression in listed SaaS names is increasingly filtering into private markets, where a significant proportion of enterprise and vertical‑specific software companies were capitalized during an era of high growth expectations and limited competition from foundation models. As public comparables re‑rate downward, private software valuations face increased scrutiny on growth durability, customer retention, and defensibility. This dynamic is particularly acute for businesses with limited data moats or those reliant on seat‑based pricing structures that may be disrupted by AI‑driven workflow reconfiguration. The path to liquidity may now require clearer differentiation in an AI‑native landscape.

Importantly, we do not view the recent sell‑off across software as a uniform signal of impairment. Rather, it reflects an ongoing repricing process as investors attempt to separate AI beneficiaries from AI‑exposed incumbents. Companies that own mission‑critical workflows, control high‑value proprietary datasets, or are embedded deeply within enterprise decision‑making are better positioned to absorb and monetize AI capabilities over time.

Impact on jobs and leadership

As capex spend continues to rise, labor has increasingly become a key margin-adjustment lever. Meta plans to cut about 10% of its workforce, roughly 8,000 employees, and eliminate around 6,000 open roles.¹⁸ Microsoft has offered voluntary buyouts to up to 7% of its U.S. workforce, and Amazon has reduced headcount by roughly 30,000 roles since late 2025, largely outside core AWS build out functions.^19,20 Importantly, these reductions are occurring in parallel with continued hiring in a narrow band of AI-related roles, underscoring the bifurcation between replaceable operational functions and scarce technical talent.

Beyond workforce levels, leadership structures are evolving in response to the AI transition. The most consequential leadership change announced so far in 2026 came from Apple, where Tim Cook will transition to Executive Chairman and John Ternus will assume the CEO role effective September 1, 2026, marking the company’s first CEO succession in 15 years.²¹ The appointment of a hardware-engineering-centric CEO is widely viewed as a signal of increased emphasis on silicon integration, systems efficiency, and internally developed AI capabilities.

While it remains too early to quantify the full labor impact of AI, evidence across hyperscaler investment, software repricing, and organizational change suggests the transition is largely proactive rather than reactive. As infrastructure spending is front loaded and software markets refocus on competitive positioning and product stickiness, technology companies are realigning labor and leadership to support higher capital intensity and more efficient growth. For investors, AI productivity gains are increasingly visible not just in product roadmaps, but in operating models, headcount strategy, and executive succession. Rather than signaling retrenchment, these shifts point to leaner, more automated platforms and leadership profiles increasingly oriented toward technical depth, infrastructure fluency, and long term value creation.