人工智能资本支出加速_28页_1mb

Morgan Stanley Global Macro Forum Summary

Core Content Overview

This document provides an analysis of the AI capital expenditure (capex) acceleration, its implications for the tech sector, and its impact on credit markets, particularly focusing on Oracle Corporation (ORCL). The report is authored by multiple Morgan Stanley strategists and analysts, including Vishwanath Tirupattur, Brian Nowak, Keith Weiss, Stephen Byrd, Andrew Pauker, Vishwas Patkar, and Lindsay Tyler.

Main Points and Key Insights

AI Capex Acceleration

• Compute Supply-Demand Mismatch: The demand for compute is expected to significantly exceed supply, with Google execs forecasting compute needs to be ~3x the Morgan Stanley projected compute CAGR for NVIDIA.
• Token Usage Growth: Weekly token usage has increased by ~2,200% in one year.
• AI Task Complexity: The complexity of AI tasks is increasing at a non-linear rate.
• AI Adoption Expansion: AI is being adopted across a broader range of applications, both software-based and embodied in physical systems (e.g., robotics).
• AI Value Creation TAM: The total addressable market (TAM) for AI value creation is nearly evenly split between Agentic (software) and Embodied (robotics) AI, and it accounts for >25% of the S&P 500 2026e adjusted pre-tax income.
• ROIC and Margins: The return on invested capital (ROIC) for "neoclouds" (AI-focused cloud providers) is attractive. Margins for LLM developers selling model access are at ~60%.
• Cost of Intelligence: With rapid advancements in chip technology, the cost of intelligence is projected to decline rapidly, potentially by >70% as the industry transitions from Blackwell to Rubin GPUs.
• Regional Cost Impact: While global power cost differentials have a minor impact on the cost of intelligence, differences in active LLM parameters used for a query have a significant effect.

Equity Strategy

• Fundamental Tailwinds: Large-cap tech companies continue to benefit from strong fundamentals, with EPS revisions breadth improving in tech subgroups.
• Capex-to-Sales Factor: The high capex-to-sales factor remains a strong outperformer in both the tech sector and the overall market.
• Market Discipline: Although there are questions about whether the market is enforcing capex discipline more structurally, the high capex-to-sales factor still outperforms.

Credit Market Outlook

• IG Bond Issuance: Morgan Stanley forecasts that investment-grade (IG) bond issuance will reach a new record in 2026 at $2.25 trillion, driven by AI capex financing, M&A activity, and corporate optimism.
• Credit Spreads: IG credit spreads are expected to widen modestly by year-end due to accelerating issuance.
• Historical Comparison: The credit market scenario is similar to 1997/98 or 2005, with credit underperforming but not signaling the end of the cycle.
• ORCL Credit View: A cautious outlook is maintained for Oracle due to cash funding needs, lease growth, and counterparty risk. Morgan Stanley recommends buying 5Y CDS protection and selling specific bonds, with potential spread widening targets around 200bp.

Key Figures and Projections

AI Capex Estimates

• 2026 Hyperscaler Cash Capex: ~$740 billion
• 2027 Hyperscaler Cash Capex: Over $900 billion

ORCL Funding Plan

• CY26 Funding Plan: ~$50 billion
• Total Funding Needs Over 2.5 Years: >$150 billion, implying ~$100 billion for the company to address

Credit Issuance Projections

Credit Spread Projections

Appendix: ORCL Credit Trades

• Buy ORCLCP 5Y CDS: Entry Level +57bp, Current Level +156bp, Rationale: Credit profile deterioration, Risks: Basis selling, hedge unwinds, profit taking
• Sell ORCL 5.2 9/26/35 and ORCL 5.95 9/26/55: Entry Level +167bp and +208bp, Current Level +156bp and +200bp, Rationale: Credit profile deterioration, Risks: Different issuance cadence

Analyst Certification

• All analysts certify that their views are accurately expressed and that they do not receive direct or indirect compensation for specific recommendations in this report.

Conflict of Interest

• Morgan Stanley may have business relationships with companies covered in the report, including Oracle, which could affect objectivity. Investors should consider this report as only one factor in their investment decision.

Stock Rating Definitions

• Overweight: Expected total return to exceed the relevant MSCI Index or industry average.
• Equal-weight: Expected total return to be in line with the relevant MSCI Index or industry average.
• Underweight: Expected total return to be below the relevant MSCI Index or industry average.
• Morgan Stanley does not assign ratings of Buy, Hold, or Sell.

Regulatory Disclosures

• Morgan Stanley and its affiliates may have business relationships with the subject companies, including market making, liquidity provision, and investment banking services.
• Certain disclosures apply to non-US jurisdictions and are for compliance with applicable regulations.

Summary

The report emphasizes the significant acceleration in AI capex, the resulting compute demand-supply mismatch, and the implications for both equity and credit markets. It highlights the potential for high returns in the AI sector, the impact of rapid technological advancements on the cost of intelligence, and the cautious outlook for Oracle's credit profile due to its substantial funding needs and increasing lease liabilities. Morgan Stanley also outlines its stock rating methodology and provides important disclosures regarding potential conflicts of interest and business relationships with covered companies.