Summary

Broadcom's "Private Cloud Outlook 2026" report forecasts a turning point in enterprise-wide AI workloads. 56 percent of surveyed companies now use private clouds for productive AI inference, while the public cloud share fell by 15 percentage points to 41 percent. Costs, complexity, and desire for control are driving this shift. Half of all companies have already moved workloads back to private clouds; 33 percent are considering this step. The survey captured 1,800 IT decision-makers from companies with at least 1,000 employees in eight countries between February and March 2026.

People

  • Moritz Förster (Author)

Topics

  • Cloud Infrastructure
  • Artificial Intelligence
  • Private Cloud
  • Public Cloud
  • IT Cost Management
  • Data Sovereignty

Clarus Lead

The study marks a politically and economically significant turning point: Cost efficiency displaces security concerns for the first time as the primary driver for infrastructure decisions (31% vs. 26% previous year). At the same time, data sovereignty gains geopolitical weight – 54 percent of IT decision-makers prioritize storage location requirements over traditional compliance. For companies in regulated industries (financial sector, healthcare, public sector), this intensifies pressure to operate sensitive AI systems under direct control rather than relying on external cloud providers.

Detailed Summary

The cost argument dominates the shift back: 97 percent of IT decision-makers acknowledge waste in their public cloud budgets; over half estimate the share at more than 25 percent of the total budget. Particularly for AI inference workloads (productive model execution), continuous operational costs weigh heavily – unlike compute-intensive, one-time training, for which public clouds with elastic GPU resources remain attractive. This economic boundary explains the asymmetrical shift.

A second, strategic motive emerges from the rise of autonomous AI agents that directly access sensitive enterprise data. Their high requirements for security, governance, and latency make private clouds the preferred operating environment for many organizations. Geopolitical factors reinforce this trend: data sovereignty and regulatory storage location requirements have overtaken traditional compliance considerations.

Critical Contextualization: Broadcom's study positions private cloud as the solution – while the company itself markets VMware Cloud Foundation, a market-leading private cloud product. Open-source projects and hyperscaler portfolios in the hybrid and private cloud segment have simultaneously grown; the market situation is less monopolistic than the report suggests.

Key Statements

  • 56% of companies rely on private clouds for AI inference; public cloud share fell by 15 points to 41%
  • Cost pressure (31%) overtakes security concerns (26%) for the first time as the main driver for infrastructure decisions
  • 97% of IT decision-makers report budget waste in public clouds; 52% estimate losses at over 25% of the budget
  • Data sovereignty (54%) displaces traditional compliance (51%) as the primary factor for storage location decisions in regulated industries
  • 50% of companies have already repatriated workloads; 33% are considering this step

Critical Questions

  1. Evidence & Source Validity: How representative is a sample of 1,800 IT decision-makers for global infrastructure markets? Are methodological details (sampling procedure, response rate, weighting by industry/region) disclosed?

  2. Conflicts of Interest: How independent is the analysis when Broadcom directly benefits from private cloud migration through VMware Cloud Foundation? Were external validators involved?

  3. Causality: Does the measured workload shift truly correlate with cost pressure, or does it reflect maturity of AI models (less experimental, more productive)? What counter-narratives explain public cloud retention?

  4. Implementation Risks: How realistic is private cloud operation for companies without deep IT expertise? What hidden costs arise from internal personnel resources, hardware amortization, and scaling constraints?

  5. Geopolitical Incentives: Are national data protection laws (GDPR, China localization) driving the sovereignty metric, or are genuine business control motives evident?

  6. Methodological Bias: Do respondents consciously distinguish between training and inference, or do they simplify complex hybrid scenarios into binary answers?

Reference List

Primary Source: Broadcom – "Private Cloud Outlook 2026" – https://www.heise.de/news/KI-Workloads-Trendwende-zur-Private-Cloud-meint-Broadcom-11334253.html

Verification Status: ✓ February–March 2026 (Survey Period)

This text was created with the support of an AI model. Editorial Responsibility: clarus.news | Fact-checking: 2026