CALVERELL PERSPECTIVES

Diversifying exposures to the AI capital cycle

· · 5 min read

The rapid expansion of physical infrastructure by major technology companies alters their free cash flow profiles and shifts their valuation dependency toward long-duration capital investments. Aligning portfolios requires mapping these cash dependencies explicitly across vendors and capital structures rather than relying on varied sector labels.

Institutional asset allocation requires a precise examination of the evolving capital structure across the technology sector. The transition from historically asset-light software models to businesses heavily dependent on physical infrastructure fundamentally alters internal cash flows [104]. This transition operates against a specific economic backdrop characterized by moderating consumer prices. The Consumer Price Index for All Urban Consumers decreased 0.4 percent on a seasonally adjusted basis in June, marking the largest one-month decline since April 2020 [103]. The broad index rose 3.5 percent over the twelve months ending in June, while the index for all items less food and energy remained unchanged for the month and rose 2.6 percent over the prior year [103]. The energy index fell 5.7 percent in June [103]. Energy is excluded from core inflation measurements, and fluctuations in commodity inputs act independently from the structural capital commitments currently dominating technology balance sheets.

Cash dependency and capital commitments

The mechanics of the ongoing investment cycle reveal a heavy reliance on external financing to supplement internal cash generation. Market consensus anticipates that the leading hyperscalers will increase capital expenditures by approximately $534 billion by 2027 relative to 2025, which corresponds to approximately $1.57 of additional investment for every $1 of additional operating cash flow [104]. Microsoft reported $35.8 billion in operating cash flow during its fiscal second quarter while recording $37.5 billion in capital expenditures [104]. That capital expenditure figure includes finance leases [104]. Finance-lease-inclusive capital expenditure cannot be subtracted from cash operating cash flow as if it represented pure cash purchases, but the scale of these commitments indicates a significant extension of long-term liabilities to secure physical compute capacity.

These infrastructure demands regularly outpace organic cash creation, pushing firms to tap capital markets. Alphabet reported a 24 percent increase in consolidated revenues to $119.8 billion for the second quarter, along with $40.77 billion in consolidated operating income [106]. Its Google Cloud division experienced an 82 percent revenue increase to $24.8 billion [105][106]. Despite this robust operating profitability, Alphabet reported burning $5.9 billion in negative free cash flow during the quarter and increased its full-year capital expenditure forecast to a range between $195 billion and $205 billion [105]. To fund its capacity expansion, Alphabet raised $49.6 billion through a combination of Class A stock, Class C stock, and mandatory convertible preferred stock [106]. The company also issued senior unsecured notes yielding net proceeds of $20.3 billion [106]. While equity and debt issuance may be discretionary choices by management, the introduction of debt service obligations and mandatory convertible securities adds complexity to the capital structure. Furthermore, Alphabet reported a net gain of $98.0 billion in other income, primarily driven by unrealized gains on equity securities [106]. Accounting net gains are not automatically distributable cash, and investors must distinguish between paper profitability and the actual cash available for shareholder return programs.

Shared duration sensitivity

The simultaneous capital deployment across cloud service providers, semiconductor manufacturers, and energy infrastructure creates a shared duration sensitivity that transcends standard sector classifications. Cloud providers funnel capital into infrastructure while relying on enterprise customers to absorb the new capacity [106]. Capacity-intensive services depend on physical infrastructure, although not every software improvement requires a proportional increase in data-centre spending. This dynamic shifts the valuation framework. Equity valuations must incorporate the execution risk associated with translating illiquid physical infrastructure into future recurring revenue. Strong earnings do not automatically establish an adequate equity risk premium absent supportive price and valuation metrics.

Various technology platforms display differing intensities of capital consumption, which may offer varied trajectories for eventual cash recovery. Oracle’s capital expenditures rose from 47 percent of operating cash flow in fiscal 2022 to 174 percent for its fiscal 2026, totaling $55.7 billion in capital expenditures against $32 billion in operating cash flow [104]. Amazon saw its trailing twelve-month operating cash flow rise 30 percent to $148.5 billion, but its free cash flow fell to $1.2 billion [104]. A forward-looking revenue run rate is not realized annual revenue, though indicators like Microsoft’s artificial intelligence business surpassing a $37 billion annual revenue run rate show early integration [104]. The risk remains whether related cash generation falters while spending persists [104]. The shared sensitivity to this single investment theme embeds significant factor risk into market-capitalisation-weighted indices. Cap-weighted funds maintain constituent proportions mechanically; a higher weight following relative appreciation is not evidence of a guaranteed or mechanical purchase by every passive strategy. A single risk theme shared across instruments is not measured correlation, and concentration is not an automatic reversal signal.

Monitoring this capital cycle requires focusing on observable cash flow transitions rather than aggregate spending announcements. The thesis of shared duration sensitivity would be challenged by independent earnings and cash evidence emerging through idiosyncratic spending variations. Diverging capital expenditure trajectories among the hyperscalers alongside robust aggregate cloud revenue growth would demonstrate more meaningful fundamental diversification. Alphabet reported delays with its Gemini 3.5 Pro model and shifted training compute toward Gemini 4 [105]. Such product cycle variations highlight the execution risk inherent in these massive outlays. Companies have largely generated sufficient free cash flow to cover dividends and share repurchases [104], but these capital return programs could face pressure if monetization timelines extend [104]. Until a clear decoupling of operating cash flow generation from continuous infrastructure spending emerges, the technology supply chain presents a concentrated duration risk tied directly to the success of physical capital deployment.

Sources

[103] Consumer Price Index — June 2026 (14 July archived release)

[104] AI investment boom puts Big Tech’s free cash flow under pressure

[105] Google increases capex forecast again after cloud-driven quarterly beat

[106] Alphabet Announces Second Quarter 2026 Results

Originally published at Ocean Capital Insights.