● MANGOS AI Power Struggle

AI Hegemony Competition in the MANGOS Era: Microsoft, Anthropic, Nvidia, Google, OpenAI, SpaceX, and Even Amazon and Apple — Who Will Really Make the Money?

The core point of this article is not simply “which Big Tech stock will go up?”

We break down where the money in the AI industry comes from, where it flows, and what conditions the final winner must have into 7 items.

The terminology keeps changing from FAANG to Magnificent 7, and then to AI-centered MANGOS, but what really matters is not the name — it is the profit structure.

If you look at how AI data center investment, generative AI services, GPUs and the semiconductor supply chain, cloud infrastructure, and enterprise AI transformation are connected, the direction of Big Tech earnings also becomes much clearer.

In particular, the core takeaway that other news often misses is that a company that “controls data and workflow” can be stronger in the long run than a company that simply has a good model.

1. Why the term changed from FAANG to M7, and then to MANGOS

In the past, the term representing U.S. Big Tech was FAANG.

Facebook, Apple, Amazon, Netflix, and Google were at the center.

Later, as the market focus shifted to cloud, semiconductors, and platforms, the Magnificent 7 emerged.

This included Apple, Microsoft, Alphabet, Amazon, Nvidia, Meta, and Tesla.

But as generative AI became the true center of the industry, a new grouping called MANGOS emerged.

• M: Microsoft
• A: Anthropic
• N: Nvidia
• G: Google
• O: OpenAI
• S: SpaceX

MANGOS is less of an official index and more of a framework for explaining AI industrial dominance.

What is interesting is that Amazon and Apple are left out.

However, if you look at real AI infrastructure and user touchpoints, the analysis is incomplete without Amazon and Apple.

That is why, in this analysis, Amazon and Apple must be included alongside the six MANGOS companies.

2. The 7 key frameworks for evaluating AI companies

When looking at AI companies, you should not judge them only by chatbot performance.

The AI industry is a complex sector in which models, chips, data centers, agents, cloud, data, and workflows are all interconnected.

The core framework of this analysis consists of the following 7 items.

1. Can they directly make or secure AI semiconductors such as GPUs, NPUs, and TPUs?
2. Do they possess the best AI model?
3. How strongly are they building AI data centers?
4. Have they secured users in general generative AI services?
5. Are they generating revenue in the coding agent market?
6. Can they dominate the work agent market?
7. Is their existing cloud data center infrastructure strong?

Based on these 7 items, simply saying “the AI model is smart” is not enough to win.

Ultimately, what matters is who gets paid by users, and which infrastructure companies that money flows to.

3. The AI Big Tech competition structure by the 7 key items

Evaluation Item	Strong Companies	Core Point
AI semiconductors	Nvidia, Google, Amazon, Tesla·xAI ecosystem	Nvidia dominates GPUs, and Google TPU and Amazon Trainium are also important.
AI models	Anthropic, OpenAI, Google, xAI	It is a competition among Claude, GPT, Gemini, and Grok.
AI data centers	Microsoft, Amazon, Google, xAI	GPU availability, power, and cloud infrastructure are key.
General generative AI	OpenAI, Anthropic, Google	ChatGPT’s brand power and usage remain extremely strong.
Coding agents	Microsoft, Cursor-related companies, Anthropic, OpenAI	The competition among GitHub Copilot, Claude Code, and Codex is intensifying.
Work agents	Microsoft, Anthropic, OpenAI, Google	Work automation connected to documents, email, meetings, and spreadsheets is the key.
General cloud data centers	Amazon, Microsoft, Google	The existing cloud share of AWS, Azure, and Google Cloud is a major advantage in the AI era.

4. Looking at the gold medal 기준, Microsoft and Amazon are the strongest

The most impressive conclusion in the original text is the evaluation that “Microsoft has 3 gold medals, and Amazon has 2 gold medals.”

Here, a gold medal means being No. 1 or having effectively first-tier competitiveness in that field.

• Microsoft: Strong in AI data centers, coding agents, and work agents
• Amazon: Strong in general cloud data centers and AI data centers
• Nvidia: Overwhelming No. 1 in AI semiconductors
• Anthropic: A strong No. 1 candidate in AI model quality competition
• OpenAI: Overwhelming brand power and user base in general generative AI services
• Google: Strong across all areas, but its overwhelming No. 1 position is relatively ambiguous
• SpaceX·xAI·Tesla ecosystem: A hybrid player where the line between AI company, space, communications, and automotive company becomes blurred

From this perspective, Microsoft has the most balanced AI profit structure.

It has Azure-based AI data centers, has captured the developer market with GitHub Copilot, and has entered the enterprise workflow market with Microsoft 365 Copilot.

This is a structure in which enterprise software, cloud infrastructure, and AI agents are all connected at once.

Amazon was left out of MANGOS, but it is absolutely impossible to ignore.

AWS is still at the core of the global cloud market, and it also has enormous competitiveness in AI data center investment.

However, Amazon may have been excluded from MANGOS because it lacks a mass-market AI service like ChatGPT or a representative model brand like Claude.

5. Nvidia has only one gold medal, but it holds the most expensive one

Nvidia remains dominant in the GPU market.

As AI data center investment increases, the first company to make money is Nvidia.

Microsoft, Amazon, Google, OpenAI, Anthropic, and xAI all ultimately need large-scale GPUs or AI accelerators.

However, Nvidia’s next battleground is not just data center GPUs.

In the future, the markets for on-device AI, personal AI servers, and local LLM execution are likely to open up.

If the era comes when AI models are run directly on personal PCs or home devices, new demand will emerge from a security and cost perspective.

In other words, Nvidia is trying to expand from data center AI semiconductors into personal AI computing.

This flow is also connected to Apple, Tesla, AMD, and Qualcomm.

The AI semiconductor supply chain is likely to remain one of the most important variables in the global economic outlook.

6. Anthropic and OpenAI are central to the model war, but the risks are also large

Anthropic receives a very strong evaluation in high-performance AI model competition, centered on Claude.

In particular, among developers and enterprise users, Claude is highly rated for document understanding, coding ability, and long-context processing.

The fact that talented people are moving to Anthropic is also an important signal.

OpenAI has the overwhelming brand called ChatGPT.

It has captured the general user market to the point where people say, “I asked ChatGPT,” rather than “I asked AI.”

This brand power does not disappear easily.

However, both companies must be viewed coldly in the long run.

Being No. 1 in model performance does not last forever.

It is impossible to know whether Claude, GPT, or Gemini will be No. 1 a year from now.

When IPOs or large-scale investment valuations are being made, you should look not at “who is No. 1 now,” but at “whether this No. 1 is sustainable.”

Because AI company valuations reflect expectations heavily, they can be more volatile than companies with stable cash flow like Big Tech earnings firms.

7. Google is quiet, but it is in every battleground

Google is one of the few existing Big Tech companies that survived even in MANGOS.

Google has TPU, Gemini, Google Cloud, Workspace, Search, YouTube, and Android.

The question is how well all of these assets are connected into one strong AI profit structure.

Google has AI models, AI semiconductors, cloud, and a lot of user data.

But compared with Microsoft, its structure for naturally penetrating the enterprise workflow market with Copilot still seems relatively weaker.

If Google Docs, Sheets, Slides, Drive, Gmail, Meet, and Gemini are well connected, it could become an enormous work agent platform.

However, the key issue is how quickly the user experience of “handing over the entire workflow to Google AI” will spread.

8. SpaceX·xAI·Tesla is a highly unique AI ecosystem when viewed as one

It is difficult to regard SpaceX as an AI company in the traditional sense.

Its core businesses are space launch vehicles, satellite communications, and Starlink.

But if you look at Elon Musk’s ecosystem broadly, the story changes.

• Tesla has AI chips and autonomous driving data in its vehicles.
• xAI is pushing the Grok model and large-scale AI data centers.
• SpaceX has a global communications network called Starlink.
• Optimus robots could become physical-world AI agents.

One particularly interesting perspective is the idea that Tesla vehicles spread across the world could become a kind of distributed AI computing resource.

If the AI chips inside vehicles remain connected to the network even while parked, there is room for expansion into local AI computation or the concept of a distributed data center over the long term.

Of course, whether car owners would allow this, whether the power and compensation structure would fit, and whether security issues can be resolved are separate questions.

However, just in terms of direction, the structure in which Tesla vehicles, Optimus robots, xAI models, and the Starlink network come together is an AI strategy completely different from existing Big Tech.

9. Apple is not in MANGOS, but it may be the strongest candidate for personal AI

Apple is not included in MANGOS, but it should absolutely not be left out.

That is because the core touchpoint of the AI era is ultimately the device in the user’s hands.

iPhone, iPad, Mac, Apple Watch, and AirPods are devices where the most personal data accumulates.

Apple’s strategy is closer to a personal AI assistant than to a massive public chatbot.

The idea is to handle easy tasks with on-device AI and more difficult tasks with a dedicated cloud with enhanced security.

Collaboration with external giant models is possible, but Apple prefers a structure that directly controls and protects user data.

At this point, Apple’s strength is hardware design.

Apple Silicon has an integrated CPU, GPU, and memory structure, which makes it advantageous for running local LLMs.

In a normal PC, GPU video memory capacity can become a bottleneck, but Macs are relatively better suited to running larger models locally thanks to unified memory.

So when asking, “Who will be the 24-hour personal AI assistant of the future if it lives at home?”, Apple emerges as a strong candidate.

Devices like Mac mini or Mac Studio could also be used like personal AI servers.

If the on-device AI era becomes mainstream, Apple could strengthen its presence in AI hardware in a different way from Nvidia.

10. The real important change: the move from the model war to the data and workflow war

From 2023 to 2025, the core question in the AI market was “Who made the best model?”

But over time, the question is changing.

• Who made the best model?
• Who is generating the most tokens?
• Who secured the most GPUs and electricity?
• Who controls the most data and workflows?

The key question going forward is the last one.

AI can only work if it has data.

Work agents only function properly when internal corporate documents, meeting minutes, emails, customer service records, contracts, code, and spreadsheets are all connected.

This is also why Microsoft is strong.

Word, Excel, PowerPoint, Outlook, Teams, OneDrive, and SharePoint are already deeply embedded in enterprise workflows.

When Copilot is added on top, AI can work directly within the workflow without needing to fetch data separately.

Google has similar assets.

If Gmail, Docs, Sheets, Slides, Drive, Meet, and Gemini are connected, it can become a powerful enterprise AI transformation platform.

However, so far, many say Microsoft is ahead in enterprise AI.

11. Why Snowflake and Databricks are attracting attention

There are companies in the AI market that get overlooked if you only focus on model companies.

Representative examples are data platform companies such as Snowflake and Databricks.

These companies are already in positions where they store and manage a lot of enterprise data.

If OpenAI says, “Bring your data to our AI side,” then a data platform company can say, “The data is already here, so we just need to attach AI next to it.”

For companies, it may be safer and more realistic to attach AI within the existing data storage rather than moving sensitive data externally.

This flow is extremely important in enterprise AI transformation.

The winner of AI adoption may not be only the company that makes the best model, but the company that controls where enterprise data is already gathered.

12. Enterprise AX fails if you only spend money on it

The most realistic part of the original text is the discussion about enterprise AI transformation, or AX.

Many companies think that buying AI accounts for employees will automatically create transformation.

But most of the time, it fails.

Enterprise AI adoption requires stages.

1. Every-person-for-themselves type: Each employee uses AI on their own.
2. Cost-support type: The company only subsidizes AI subscription costs.
3. Training-support type: The company sets standard AI tools and provides training.
4. Management-and-operations type: Usage is measured, workflow application is managed, and continuous improvement is made.

Real transformation begins at stage 4.

You need to measure who is using it and how much, see which tasks it is effective for, and provide additional training to those who cannot use it well.

Just as buying cars does not create a driving culture, buying AI accounts does not create workflow innovation.

Especially if employees are worried that “if I use AI well, my job will disappear,” AX becomes even more difficult.

The company must persuade employees that AI adoption is not simply a workforce reduction tool, but a tool that enhances individual job competitiveness.

Without this persuasion and management, enterprise AI transformation is likely to end as a formal introduction only.

13. The most important points that other news often misses

First, the winner in AI may be the company that controls the workflow, not the company with the best model.

No matter how good Claude or GPT is, actual workflow automation is limited if it cannot access enterprise documents, email, meetings, approvals, and customer data.

That is why the positions of Microsoft, Google, Apple, Snowflake, and Databricks matter.

Second, the AI data center war will push up electricity and memory prices as well.

Buying GPUs is not the end of it.

Electricity, cooling, land, networks, HBM memory, server racks, and the power grid are all needed.

That is why AI data center investment is a giant economic variable that shakes both the semiconductor supply chain and energy infrastructure at the same time.

Third, when on-device AI opens up, the value of Apple and Tesla may be seen again.

Right now, data center AI is the center, but in the long run, AI running directly on personal devices could grow larger.

For companies and individuals where security matters, local AI becomes a very attractive option.

Fourth, enterprise AI transformation is a management issue, not a technology issue.

More important than handing out AI accounts are usage measurement, training, workflow redesign, and data cleanup.

If this is not done, the AI adoption budget ends as a cost and does not lead to productivity gains.

Fifth, valuation must consider both the number of gold medals and their durability.

Whether a company should receive a Microsoft-level valuation just because it is No. 1 in one area is a separate issue.

You need to examine whether that No. 1 position will last a year from now, whether it will connect to cash flow, and whether it can bear infrastructure costs.

14. Key investment points by company

• Microsoft: The key point is how quickly Azure AI data center investment and Copilot revenue connect.
• Amazon: You need to watch AI demand based on AWS and whether its own AI semiconductor, Trainium, expands.
• Nvidia: It is important whether data center GPU demand continues and whether it can expand into the on-device AI market.
• Anthropic: Claude’s model competitiveness and enterprise workflow penetration are key.
• OpenAI: The key issue is how much it can convert the ChatGPT user base into API, agent, and enterprise revenue.
• Google: It is important whether Gemini, TPU, Google Cloud, and Workspace can be tied into one AI ecosystem.
• SpaceX·xAI·Tesla: You need to check whether AI models, robots, vehicles, and satellite communications can be linked to an actual profit structure.
• Apple: The key point is whether Siri and Apple Intelligence can open the personal AI assistant market.

< Summary >

MANGOS is a new Big Tech framework for explaining dominance in the AI industry.

However, to assess real competitiveness, Amazon and Apple must also be considered.

Microsoft is the most balanced powerhouse in AI data centers, coding agents, and work agents.

Amazon is absent from MANGOS, but it is a core player in AWS and AI infrastructure.

Nvidia holds the most expensive gold medal: AI semiconductors.

Anthropic and OpenAI are strong in models and generative AI services, but they must prove a sustainable profit structure.

Google has all the assets, but execution synergy is the key issue.

Apple is a strong candidate in on-device AI and the personal AI assistant market.

The next AI competition is likely to be decided not just by model performance, but by who controls data, workflows, power, and cloud infrastructure.

[Related Articles…]

• AI Data Center Investment Cycle and Power Bottleneck Analysis
• AI Semiconductor Supply Chain and Big Tech Infrastructure Competition

*Source: [ 티타임즈TV ]

– 누가 제일 잘 나갈까? 마이크로소프트, 앤트로픽, 엔비디아, 구글, 오픈AI, 스페이스X, 아마존, 애플 (박종천 지란지교소프트 CAIO)

---