The reasons behind a downgrade for Microsoft and Amazon

Key Concepts

• Hyperscalers: Large cloud computing providers like Microsoft and Amazon.
• Cloud 1.0: The traditional cloud computing model characterized by high capital deployment and value creation.
• Generative AI: A type of artificial intelligence that can create new content, such as text, images, or code.
• NBV (Net Book Value): The value of an asset after accounting for depreciation.
• NPV (Net Present Value): The difference between the present value of cash inflows and the present value of cash outflows over a period of time.
• CAPEX (Capital Expenditure): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, buildings, and equipment.
• IRR (Internal Rate of Return): A discount rate that makes the net present value (NPV) of all cash flows from a particular project equal to zero.

Downgrade of Microsoft and Amazon to Neutral

Alex Heisel, an analyst at Rothschild & Co. Redburn, has downgraded Microsoft and Amazon, two of the "Magnificent Seven" companies, from "Buy" to "Neutral." This decision stems from a more cautious stance on hyperscalers in light of generative AI's impact on their business models.

Reasons for the Downgrade

Heisel's primary reasoning is that the market is currently pricing in an overly optimistic scenario regarding the value generative AI will create for both Microsoft and Amazon. The current market perception still relies on the "traditional cloud playbook" where hyperscalers were dominant in "Cloud 1.0."

Key Points:

• Capital Deployment and Value Creation Mismatch: In Cloud 1.0, hyperscalers converted $1 of CAPEX into $1.4 of NBV. However, with generative AI, this ratio is projected to be $1 of CAPEX for only $0.20 of NPV. This significant mismatch indicates a less efficient value creation process.
• Lack of Clear Path to Revert to Cloud 1.0 Economics: Heisel does not see a clear pathway for Microsoft and Amazon to return to the highly profitable economics of Cloud 1.0.
• Increased Competition from New Platforms: The emergence of new platforms like Anthropic and OpenAI fundamentally changes the competitive landscape. These are not just consumer-facing applications but act as a "platform layer," where significant value is generated. This increased competition risks pushing hyperscalers to operate more in the "lower part of the stack."

The Role of Anthropic and OpenAI

The transcript addresses Microsoft's $35 billion investment in Anthropic. Heisel clarifies that while Anthropic will utilize Microsoft's infrastructure (similar to AWS and GCP), this primarily translates to "server rental" and some auxiliary services for the hyperscalers. The substantial value accrual, according to Heisel, will be for companies like Anthropic and OpenAI, which are building products on top of these services. He emphasizes the need to differentiate between "server rental" and the "product on top of this service" to capture true value.

Desired Metrics for Reassessment

To feel more comfortable with Microsoft and Amazon, Heisel would like to see:

• Sustained Re-acceleration of Growth: Continued strong revenue growth.
• CAPEX Coming Down: A reduction in capital expenditure.

He explains that without this combination, the conclusion is that the "cost to grow is really, really high." If CAPEX were to decrease while revenues accelerate, it would signify an increase in revenues per unit of compute, which is what is needed to reassess his view.

Thoughts on Quantum Computing

Heisel explicitly states that he is "probably not the right person" to provide insights on quantum computing, as his focus is strictly on hyperscalers.

Impact of NVIDIA's Earnings

Heisel anticipates a "quite limited" impact from NVIDIA's earnings on Microsoft and Amazon. He notes that while the market has historically rewarded growth and CAPEX, recent earnings from Amazon and Microsoft showed decent growth but the market is now looking beyond top-line figures and IPO numbers. Investors are increasingly focused on understanding the underlying value creation, making the market more nuanced. Therefore, he doesn't expect any significant impact from NVIDIA's earnings on these two companies.

Confidence in Current Analysis

Heisel's confidence in his analysis stems from his deep focus on hyperscalers like Amazon, Microsoft, and Oracle. He highlights that since launching his coverage in June 2022, his team has been "defending these companies" with an "above consensus view in terms of re-acceleration." He acknowledges that it was a challenging period, requiring thorough research, and that these changes would not have been made without being "very sore in the work" they have done.

Ability to Adapt and Change

Heisel is "sure" that Microsoft and Amazon have the ability to adapt and change. He clarifies that the returns on generative AI deployments (e.g., GPUs, AWS training tools) are still "okay," with an expected IRR of 15-16%. However, the core issue is that the market is still valuing these companies as if they were operating under the Cloud 1.0 model, with an expectation of reverting to past performance. He reiterates that there is "no clear path towards this."

Conclusion

Alex Heisel's downgrade of Microsoft and Amazon to Neutral reflects a critical assessment of how generative AI is reshaping the economics of cloud computing. He argues that the market is overestimating the immediate value creation from generative AI for hyperscalers, failing to account for increased competition from platform-layer AI companies and the shift from high-margin cloud services to lower-margin infrastructure rentals. He emphasizes the need for a clear demonstration of accelerating revenue growth coupled with declining CAPEX to justify a more optimistic outlook.