Goldman Sachs maintains its Buy rating on Mitsubishi Heavy Industries, viewing its NVIDIA AI data center collaboration as a potential re-rating catalyst

Overview

This report focuses on media coverage of Mitsubishi Heavy Industries' technology collaboration with NVIDIA on AI data centers. The reported areas of cooperation include high-efficiency cooling and air-conditioning systems, emergency power supply, and energy management systems, aimed at alleviating the enormous power consumption and heat dissipation burden caused by GPUs. Goldman Sachs believes that although the economic terms, implementation facilities, and timetable have not been disclosed and the near-term income statement impact will be limited, the collaboration strengthens MHI's strategic positioning in the AI infrastructure value chain.

Core views

Goldman Sachs' core views are as follows: first, the NVIDIA collaboration is strategically positive, but quantifiable information such as order value and implementation timing remains unavailable; second, MHI's data center-related business is still small and may account for less than 1% of company sales, limiting its near-term earnings contribution; third, as AI investment expands from GPUs to full-stack infrastructure including power, cooling, and energy management, the market may be undervaluing MHI's portfolio of high-efficiency power generation, cooling solutions, next-generation power distribution, nuclear power, and defense capabilities; fourth, the report maintains its Buy rating and 12-month target price of ¥6,000.

Analysis framework

The report combines an event-driven approach with a valuation framework. It first assesses the commercial maturity and level of information disclosure surrounding the NVIDIA collaboration, then compares the scale of the data center business with the company's total sales to evaluate the near-term earnings impact. It subsequently evaluates potential re-rating catalysts from the perspectives of the expanding AI infrastructure investment theme, the GTCC oligopoly involving three major players, nuclear power growth, defense positioning, and the extent to which these factors are reflected in the valuation. The target price is based on FY3/30E estimates and a relative EV/EBITDA premium multiple versus the Japanese aerospace and defense sub-industry.

Methodology notes

• Valuation methodsRelative EV/EBITDA valuation

  The target price uses FY3/30E estimates, sets target EV/EBITDA at a 40% premium to the Japanese aerospace and defense sub-industry average of 14x, and then discounts it to the midpoint of FY3/27E and FY3/28E using a 9% cost of capital.

  This method emphasizes MHI's potential for a valuation re-rating relative to defense and industrial peers rather than pricing the company solely based on the current scale of its data center business.

• Event analysisStrategic collaboration catalyst assessment

  Assess the impact of the reported NVIDIA collaboration on MHI's business positioning and market narrative.

  The report distinguishes between near-term earnings contribution and medium- to long-term strategic value. It believes that the absence of order value and implementation timing limits the earnings impact for now, while the AI infrastructure theme could improve valuation perceptions.

• Factor frameworkGS Factor Profile

  Goldman's factor profile compares growth, financial returns, valuation multiples, and composite percentiles.

  The disclosure section explains that the GS Factor Profile is based on analyst forecasts and constructs relative percentiles using metrics including sales, EBITDA, EPS, ROE, ROCE, CROCI, and valuation multiples.

• M&A frameworkM&A Rank

  Goldman Sachs uses a 1-to-3 M&A probability score to assess the likelihood that a company will become an acquisition target.

  This disclosure describes Goldman's general research framework; the report does not treat MHI's M&A probability as a primary investment thesis.

Asset mapping & comparison

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

• Mitsubishi Heavy Industries (7011.T)

  The core company covered by the report; the Buy rating is maintained.

  Strengths

  Offers multiple infrastructure capabilities spanning high-efficiency cooling, emergency power supply, energy management, GTCC, nuclear power, and defense. Its collaborations related to NVIDIA and SoftBank strengthen the AI data center narrative.

  Weaknesses

  The data center business remains small and may contribute less than 1% of total company sales; the NVIDIA collaboration's economic terms have not been disclosed.

  Comparison

  Goldman Sachs references the Japanese aerospace and defense sub-industry average EV/EBITDA of 14x for the target valuation and applies a 40% premium.

  Risks

  A stronger-than-expected yen, concentration in low-margin energy projects or one-off costs, and setbacks in portfolio reform could reduce overall returns.

• NVIDIA

  The AI data center technology collaboration partner, not the subject of this report's rating.

  Strengths

  Its AI factories are driving demand for GPU-related power, thermal management, and energy management infrastructure.

  Weaknesses

  The report does not disclose the order value, target facilities, or implementation timing of the NVIDIA collaboration.

  Comparison

  NVIDIA represents upstream GPU demand in the AI investment chain, while MHI corresponds to the power and cooling segments of AI infrastructure.

  Risks

  If the collaboration does not develop into specific projects, its actual contribution to MHI's earnings could remain limited.

Key data

• RatingBuyGoldman Sachs maintains its Buy rating on Mitsubishi Heavy Industries.
• 12-month target price¥6,000Based on FY3/30E estimates, the target EV/EBITDA multiple, and discounting at a 9% cost of capital.
• Current estimated scale of data center businessSeveral tens of billions of yen to less than ¥100 billionThe report states that this business may currently account for less than 1% of total company sales.
• 2024 MTBP growth area targetApproximately ¥100 billion in FY2026Growth areas include hydrogen/ammonia, CCUS, and electrification/data centers.
• NVIDIA report timing2026-07-14 16:00 JSTNikkei English reported that NVIDIA and MHI are collaborating on AI data center technologies.
• Target multiple assumptionA 40% premium to the Japanese aerospace and defense sub-industry average EV/EBITDA of 14xUsed to derive MHI's theoretical value.
• Cost of capital9%Used to discount the theoretical value to the midpoint of FY3/27E and FY3/28E.
• Key downside risksA stronger yen than assumed, declining profitability in the energy business, and setbacks in portfolio reformKey downside risks identified in the report.

Impact & implications

The direct earnings impact of the event is limited because no orders, facilities, or timetable have been disclosed and the data center business currently represents only a small share of the company. Its investment implication, however, is that the bottleneck in AI data center investment is extending beyond GPUs to power, cooling, and energy management. MHI's exposure to GTCC, nuclear power, cooling, power distribution, and defense could make it an undervalued AI infrastructure-related stock in the Japanese market. If the collaboration subsequently converts into actual orders or a clearer facility implementation plan, it could reinforce the market's re-rating of MHI's growth businesses.

Risks

• The yen appreciates more than Goldman Sachs assumes against major currencies, potentially reducing overseas business earnings or translated earnings.
• If the energy business becomes concentrated in low-margin projects or incurs significant one-off costs, profitability may fall below expectations.
• Setbacks in portfolio reform could reduce overall returns.
• The NVIDIA collaboration remains at the technology cooperation and adoption assessment stage. The absence of order value, target facilities, and implementation timing creates uncertainty regarding near-term earnings realization.
• Data center-related business currently represents a low proportion of company sales, so even growth may not materially change overall earnings in the short term.

What to watch

• Whether NVIDIA and MHI issue a formal announcement.
• Whether the collaboration discloses specific order values, target AI data center facilities, implementation timing, and business models.
• Whether MHI's data center-related sales expand from the tens-of-billions-of-yen range toward the hundreds-of-billions-of-yen range.
• Progress at MHI's strategic base in Dallas, post-acquisition integration of Concentric, and expansion of its North American data center business.
• Whether AI data center investment continues to expand from GPUs to power, cooling, power distribution, and energy management infrastructure.
• Whether Japan's defense, GTCC, and nuclear power businesses continue to drive a market re-rating of MHI.