Former Microsoft AI transformation director says demand for AI infrastructure remains strong, with enterprise on-premise and edge computing investments boosted by Agentic AI

Overview

This report is a summary from the Goldman Sachs Americas Technology Hardware Expert Network series based on an investor webcast with former Microsoft AI transformation director William Fong on July 10, 2026. The report focuses on trends in AI infrastructure demand, especially the impact of Agentic AI on enterprise compute on-premise, edge computing, cloud orchestration, training infrastructure, server refreshes, power, and memory supply chains.

Core views

The report argues that Agentic AI is prompting enterprises to adopt a hybrid compute strategy: individual AI agents are more likely to run on-premise or at the edge to achieve lower cost, lower latency, and stronger data governance, while complex multi-agent orchestration layers can still rely on cloud. Experts view AI infrastructure as not facing obvious overbuilding risk because inference demand from Agentic AI is still expanding, and enterprises still contain large amounts of data that have not yet been used for model training, which will continue to support additional training infrastructure demand.

Analysis framework

The report uses an expert network interview format and relies on the sector observations of a former Microsoft AI transformation director to judge enterprise AI infrastructure demand, budget sources, changes in deployment architecture, and supply-chain constraints. The analysis emphasis is not on company financial models or valuation, but on inferring the direction of hardware demand from technology architecture, enterprise procurement logic, and infrastructure bottlenecks.

Methodology notes

• Expert interviewGoldman Sachs Expert Network Series

  Obtaining frontline technology and procurement trend judgments through industry expert network briefings.

  The report discusses the impact of Agentic AI on cloud, on-premise, edge, and training infrastructure demand based on William Fong’s observations of enterprise AI transformation and infrastructure deployment.

• Factor framework disclosureGS Factor Profile

  Compares stocks versus the market and industry peers using standardized rankings of growth, financial returns, valuation multiples, and composite indicators.

  The disclosure section states that Goldman Sachs uses a standardized ranking across forward sales, EBITDA, EPS, ROE, ROCE, CROCI, P/E, P/B, EV/EBITDA, but this report does not provide MICROSOFT CORP’s specific factor results.

• M&A framework disclosureM&A Rank

  Evaluates the likelihood of a company becoming an M&A target on a scale of 1 to 3.

  The report explains the definition of Goldman Sachs’ M&A rank; however, it does not provide MICROSOFT CORP’s specific M&A rank.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• MICROSOFT CORP (MSFT.US)

  theme-related company

  Strengths

  Microsoft is highly relevant to enterprise AI, cloud orchestration, and AI transformation demand, and the report cites views from its former AI transformation director, showing that enterprise AI infrastructure demand remains robust.

  Weaknesses

  This report does not provide Microsoft-specific revenue forecasts, Azure growth, capex, or valuation calculations, so it cannot directly quantify the impact on Microsoft’s profitability and valuation.

  Comparison

  The report clearly notes that enterprise compute hardware vendors such as DELL and HPE may benefit more directly; Microsoft is more positioned in the context of cloud and AI ecosystem demand.

  Risks

  If enterprise AI ROI falls short of expectations, data-center power is constrained, memory supply tightens, or AI inference architecture changes, related infrastructure demand may not materialize as expected.

• Enterprise compute hardware suppliers

  potential beneficiary assets

  Strengths

  Rising demand for local deployment and edge computing is likely to drive updates to servers, networking, and efficient CPU platforms.

  Weaknesses

  The realization of hardware demand is constrained by enterprise budgets, supply chains, and power availability.

  Comparison

  Compared with pure cloud service providers, enterprise hardware vendors may benefit more directly from workload repatriation to on-premise and edge environments.

  Risks

  Memory bottlenecks, slower procurement cycles, and insufficient data-center power interconnection could limit shipments.

Key data

• Webinar date2026-07-10Goldman Sachs held an investor webcast with William Fong.
• Share of internet data in global dataabout 1/3The expert noted that internet data represents only part of global data, and enterprises still have large amounts of untrained data within internal systems.
• Next-generation server efficiency exampleabout 2x performance per rack, 50% lower power consumptionThe report gives examples indicating that new platforms such as Arm AGI CPUs can increase per-rack performance and reduce power usage.
• Goldman Sachs global equity coverage universe3,104 stocksAs of 2026-07-01, the number of global equities covered and rated by Goldman Sachs investment research.
• Rating distributionBuy 50%, Hold 34%, Sell 16%The disclosure table shows the rating distribution of Goldman Sachs global stock coverage.

Impact & implications

From an investment perspective, the shift of enterprise AI workloads from purely cloud to a hybrid of cloud, on-premise, and edge deployment may support demand for servers, networking, storage, memory, edge devices, and enterprise compute suppliers. DELL and HPE and other enterprise compute hardware vendors could benefit. For MICROSOFT CORP, the report mainly reflects its ecosystem and customer demand backdrop, rather than providing a new rating or target price directly. Power assurance and memory supply will affect the pace of AI cluster and edge platform rollout.

Risks

• AI infrastructure buildout is constrained by power interconnection and grid capacity, and some data-center projects may be canceled or delayed if guaranteed power cannot be secured.
• Memory supply chain constraints could limit shipments of edge platforms and AI devices.
• If enterprise AI use cases do not continue to generate sufficient ROI, hardware reinvestment budgets may weaken.
• The report is primarily based on expert interviews and does not provide a complete financial model, company rating changes, or target-price estimates.
• Goldman Sachs disclosed that it may have investment-banking or other business relationships with covered companies, so investors should consider disclosed information to assess potential conflicts of interest.

What to watch

• Whether enterprises continue to migrate AI agent workloads to on-premise and edge deployment.
• Whether the division of labor between cloud multi-agent orchestration and local inference becomes clearer.
• The progress of power assurance, grid interconnection, and high-power AI cluster build capability in data-center projects.
• Whether memory supply improves and how it affects edge AI platform shipments.
• Whether enterprises continue to reinvest in AI infrastructure through server refreshes, power savings, and productivity gains from staff efficiency.
• How Microsoft, DELL, HPE and other related companies subsequently describe demand for AI infrastructure, capex, and order visibility.