Five major hyperscalers face scrutiny over AI infrastructure funding sustainability as corporate bond issuance reached $194.1 billion in July, according to data from Bloomberg and Samsung Securities. KB Securities and Samsung Securities analysts project these companies retain a minimum two-to-four-year funding runway despite the shift from cash reserves to debt-based capital raising. The analysis emerges as semiconductor stocks including Samsung Electronics and SK Hynix experienced sharp declines in July amid concerns over the duration of AI capital expenditure cycles.
KB Securities Projects CAPEX Expansion Through 2027
Five hyperscalers — Alphabet, Microsoft, Amazon, Meta, and Oracle — are projected to reach $725.1 billion in combined capital expenditures for 2026, representing a 4.9-fold increase from 2023's $148.1 billion, according to KB Securities. The securities firm attributes the expansion to agentic AI adoption driving inference demand and memory bottleneck conditions raising unit costs. Microsoft's GPU costs account for 60–70% of CAPEX, with memory costs comprising 13% of total capital expenditure, the report states.
KB Securities analyst Kim Se-hwan projects the CAPEX-to-revenue ratio will peak at approximately 38% in 2027 before declining to 34% in 2028. "AI infrastructure investment occurs 6–24 months ahead of revenue generation, so CAPEX burden will gradually normalize as advance investments convert to sales," Kim stated in the research note.
Corporate Bond Issuance Surges to $194.1 Billion in July
Hyperscaler corporate bond issuance reached $194.1 billion as of July, compared to $20.1 billion in 2024 and $108.4 billion in 2025, according to Bloomberg and Samsung Securities data. Oracle's bond issuance totaled $25 billion in July, pushing its total debt-to-EBITDA ratio to 5.6x. In contrast, Microsoft, Alphabet, Meta, and Amazon maintain ratios of 0.6x, 0.6x, 0.8x, and 1.3x respectively as of March 2026, Samsung Securities reported.
The rapid increase in Oracle's debt metrics triggered a spike in its credit default swap premiums, raising questions about capital adequacy across the hyperscaler sector. However, Samsung Securities researcher Kwon Beom-seok characterized credit concerns for the four other hyperscalers as premature given their current debt levels.
Samsung Securities Estimates Two-to-Four-Year Funding Runway
Samsung Securities projects hyperscalers retain between 1.9 and 4.9 years of funding capacity before debt levels reach thresholds that burden credit ratings, assuming current bond issuance rates continue. The firm's company-specific estimates are: Microsoft 4.0 years, Alphabet 4.9 years, Meta 3.9 years, Amazon 3.1 years, and Oracle 1.9 years. These calculations incorporate each company's cash generation capacity (EBITDA), existing debt, and annual bond issuance volume.
"Credit risk materialization for hyperscalers requires a minimum of two years based on current trajectories," Kwon stated. "With AI-related revenue maintaining 30%+ growth rates, profitability improvements may gradually reduce CAPEX burden over time." The 10-year US Treasury yield approached prior highs above 4.6% amid US-Iran tensions, though the July US Consumer Price Index came in slightly below expectations. CME Group's FedWatch tool showed an 83.4% probability of the Federal Reserve holding rates steady in July.
FAQ
What is the projected 2026 CAPEX for the five major hyperscalers?
KB Securities projects combined capital expenditures of $725.1 billion for 2026 across Alphabet, Microsoft, Amazon, Meta, and Oracle, representing a 4.9-fold increase from 2023's $148.1 billion.
How much corporate bond issuance did hyperscalers complete in July?
Hyperscaler corporate bond issuance reached $194.1 billion as of July, compared to $20.1 billion in 2024 and $108.4 billion in 2025, according to Bloomberg and Samsung Securities data.
What funding runway do analysts estimate for hyperscaler stocks?
Samsung Securities estimates funding capacity ranging from 1.9 years (Oracle) to 4.9 years (Alphabet) before debt levels reach credit rating thresholds, with Microsoft at 4.0 years, Meta at 3.9 years, and Amazon at 3.1 years.
