April 24, 2026
Share:
Musings of a Market Practitioner
The Mag 7 is Dead. Long Live...?
The Return of the Three Kings: Google, Apple & Amazon.
When I wrote the original essay in August 2025 it was still early in the AI paradigm shift, and since then, much has changed. Anthropic’s emergence as the first AI-native disrupter, Google flexing its core advantages, OpenAI’s stumbles, and Apple’s reminder of its relevancy are a few of the notable twists and turns so far.
In this version of Game of Thrones, the leaders have emerged, bringing with them an army of advantages that will be hard to circumvent.
My core assumption remains the same:
“LLMs will become commoditized. Scale is the moat — and scale requires capital. To earn a return, you must monetize the model. Value will accrue to those building compelling applications on top of LLMs and distributing them efficiently. As with the internet, adoption accelerated once real utility (e.g., e-commerce, logistics optimization) became available.”
Now, what has also become clear is how foundational the following three assets have become: hardware, application layer, and distribution – the Three Foundations.
I use the term hardware in broad terms, but it captures infrastructure, compute optionality, and distributed devices.
There is no AI without compute infrastructure, and the infrastructure does not work without power. Compute optionality means in-house silicon expertise to tailor the applications you build as efficiently as possible.
Finally, for AI to become useful, it has to be distributed as widely as possible, pushing more compute to the edge.
These foundations have helped define which winners emerged to dominate in past technology paradigms shifts, and I believe that will also be the case in the future.
This note outlines my revised thinking on each of the Mag 7, and how it has changed from my original views in August 2025.
Survive and Thrive
Amazon (AMZN)
Updated View – As an early investor and partner of Anthropic, Amazon chose wisely. As the main infrastructure partner, Amazon is riding the wave of disruption from Claude Code, which is not only enhancing its enterprise application layer but also accelerating AWS’s growth. Also, Amazon’s investment in its own silicon (Trainium/Graviton) and its logistics infrastructure (robotics) give it both margin and growth optionality.
Original View – Rather than building its own LLM, Amazon partnered with Anthropic—suggesting it sees more value in the application layer. By combining Anthropic’s LLMs with its own models, Amazon enhances AWS’s infrastructure and its utility to developers and enterprise users.
Apple (AAPL)
Updated View – Apple has the Three Foundations – hardware (custom silicon); application layer (not only iOS and the App store, but access to Gemini); Distribution.
The DoJ ruling in August 2025 was a gamechanger, allowing Apple the freedom to partner with long time search partner Google to integrate Gemini’s in-cloud LLM with Apple’s on-device personalized inferencing. If Apple can execute, they will likely have the first scaled consumer AI utility – your own personalized AI assistant.
Original View – While some claim Apple has fallen behind in AI, I disagree. Apple has structural advantages: no other company combines such global reach, compute density, and hardware desirability as the iPhone. For AI to become truly useful to consumers, iOS-based apps will need to merge LLM-scale knowledge with on-device inferencing to offer personalized intelligence. Apple has the platform and distribution—it only lacks an LLM partner. A pivotal catalyst could be the DOJ’s ruling in its antitrust case against Google. A favorable verdict would likely open the door for Apple to partner with a leading LLM, potentially vaulting it into AI leadership.
Alphabet (GOOGL)
Updated View – The DoJ ruling was indeed a clearing event for Google. Since then, Gemini has become a top 3 LLM and its AI partnership with Apple cements its place as a vital AI infrastructure partner, underlined further by its recent partnership with Anthropic.
Importantly, Google has the platforms to monetize this massive capex in AI, which will be vital for the market to continue to underwrite the spending. As an industry, we may be approaching peak capex (likely on a 2nd derivative basis), with future incremental investment going toward the application/agentic layer.
Original View – The risk: AI disrupts Google's core search business. The response: Gemini and AI Search aim to slow the decline. I’ve been bearish since ChatGPT’s 2022 launch but now believe Google has the right ingredients: Cloud infrastructure, a competitive LLM, deep data assets (e.g., YouTube), and monetization paths via Android, GCP, and iOS. The DOJ’s upcoming antitrust ruling could serve as a clearing event—especially if it separates traditional search from AI search.
Nvidia (NVDA)
Updated View – Since October 2025, Nvidia has traded in range for two reasons, in my view: investor concerns about a deceleration of growth, starting in CY2027 as the 2nd derivative of capex growth starts to slow; and potential competition in inferencing which is a bigger consumer of compute needs.
At 18x FY2028 earnings, I think most of the deceleration derating has happened, while revenues have the potential to be revised higher as compute demand continues to outstrip supply. Strategically, Nvidia is also ahead of the curve – it is no longer just about the GPU or CPU; it is the whole rack, server, and datacenter. Nvidia is working on smoothing out multiple points of friction, for example, in storage (use of SRAM for speed), and networking (copper to optical). Cost per token is declining, and Nvidia’s datacenter architecture changes will keep driving those costs lower, which is a critical variable as customers project an eventual RoI.
Original View – Model upgrades continue to validate AI scaling laws—ensuring demand for Nvidia’s latest chips. Beyond AI data centers, new buyers are emerging: sovereign nations, large enterprises, and more. Many overlook that GPUs also powers crypto and blockchain, both early in their adoption curves. When combining AI, crypto, and blockchain, the total addressable market for Nvidia may be far larger than most assume.
Survive? Yes. Thrive? It Depends.
Microsoft (MSFT)
Updated View – Microsoft’s strategic positioning has been negatively affected by three things, two of their own making, one that was outside of their control: 1) Their early lock-in with OpenAI proved to be a blind spot as they were late to an Anthropic deal, and OpenAI has made multiple missteps due to a lack of strategic focus; 2) Slowing down capex spending in 2025 proved to be a massive mistake as they were unable to participate in the accelerating growth of token consumption; and 3) the launches of Claude Code, Opus, and Cowork were a broadside hit to the software industry and raised questions about the terminal value of a broad range of software companies.
While Microsoft’s enterprise moat will be hard to displace, AI-native applications are forcing the company to re-engineer their own applications, which is coming at a cost.
Regarding AI capex spend, Microsoft has a decision to make: massively accelerate capex and temporarily crush free cash flow in order to accelerate Azure growth in the future; or try to manage capex spend that balances their customers’ demands as well as their own internal product development. In my view, for the stock to work in the future, they must rip the band aid off and follow Google, Amazon, and Meta with a massive acceleration in capex. This will accelerate Azure growth, but longer-term questions on software will remain.
Original View – Like Amazon, Microsoft has doubled down on its partnership with OpenAI. Its capital spending now emphasizes inference over training, with a clear goal: delivering real, user-facing utility. That said, Microsoft’s SMB franchise could face long-term pressure from AI-native operating systems adopted by startups and younger users.
Meta (META)
Updated View – Not much has changed from my original view. Meta’s new model placed them in 4th place, but the question remains: if used for internal purposes only, where will Meta’s operating margins end up? If Meta Compute is a bolder strategy to provide infrastructure to third parties, then margins are going lower.
Strategically, Meta is chasing the leaders and the cost to compete is too high for the existing business model to sustain. It will have to change. Longer term, this line is probably the biggest problem for Meta – “More troubling is the risk that users shift attention toward personalized AI assistants—undermining Meta’s engagement-driven ad model.”
Original View – Zuckerberg’s hiring of a “super team” of AI engineers raises a question: bold move or desperation? I lean toward the latter. Meta’s open-source LLM strategy has fallen flat. More troubling is the risk that users shift attention toward personalized AI assistants—undermining Meta’s engagement-driven ad model. While Meta owns valuable assets, its AI team must now execute—and fast—against four established LLM players.
Multiple Challenges Ahead
Tesla (TSLA)
Updated View – Musk is as creative a financier as he is an engineer. By merging xAI into SpaceX ahead of this IPO, Musk is using this sought-after strategic asset as a back-door way to fund the billions of dollars he wants to invest in xAI. You, the SpaceX shareholder, will be paying for this capex. Can he monetize this investment in the future, and which enterprise in their right mind would partner with him? But if it is for internal use (to fund his many projects from Optimus, Robotaxi, TerraFab), we may likely never see a return on investment. This is still not a constructive setup.
Original View - Though not directly threatened by AI, Tesla warrants comment. The company is trying to pivot from EVs to robotics/autonomy—a bold move complicated by CEO Elon Musk’s erratic leadership. Musk overpaid for Twitter, abandoned OpenAI (a strategic blunder), and tainted Tesla’s brand through political theatrics. His AI startup xAI has shipped fast, but burns $1B/month—raising doubts about its sustainability and appeal to enterprise buyers. Tesla shareholders must now contend with a distracted CEO leveraging TSLA equity to fund side projects. This is not a constructive setup.
Next time I update this view, there may be two new companies vying for leadership in an AI world.
These views are mine alone and may evolve as the technology landscape shifts.
Sincerely,
Sam Rahman
HGRO Portfolio Manager, Hedgeye Asset Management
ABOUT US
Hedgeye Asset Management, LLC (“HAM”) is an SEC-registered investment adviser that manages exchange traded funds (ETFs) and provides model portfolio and index services to other registered investment advisers. HAM is a subsidiary of Hedgeye Risk Management, LLC. For more information about the ETFs, strategies, research or HAM, please visit www.hedgeyeam.com or email info@hedgeyeam.com.
DISCLOSURE
This information is adviser marketing and market commentary, and neither tailored nor specific investment advice, nor an offer to sell, or a solicitation of an offer to buy any investment product, security, or services offered by Hedgeye Asset Management, LLC (“HAM”) or its affiliates. HAM and its affiliate, Hedgeye Risk Management, LLC (“Hedgeye”), issue market research which may differ from the strategies and products due to varying objectives and parameters. HAM does not provide legal or tax advice. Investors should make their own decisions regarding any investments mentioned, and their prospects, based on such investors’ own review of publicly available information, including fund prospectuses, and should not rely on the information contained herein. Investors should consider investment objectives, risks, fees and expenses carefully before investing. Certain information contained in this document may constitute “forward-looking statements.” Due to various risks and uncertainties, actual events or results may differ materially from those reflected or contemplated in such forward-looking statements.