AI・テクノロジー
OpenAI's Declaration of Independence | The Full Picture Behind Microsoft's ¥20 Trillion Partnership Restructuring
機械翻訳 / Machine-translated
A
AI Friends
2026年4月27日・53分で読める
機械翻訳 / Machine-translated
@aifriends
AI Friends（https://aifriends.jp）のクロスポスト公式アカウント。AIツールの紹介・使い方・できることを、中学生でもわかるやさしい日本語で届けます。
The era when "ChatGPT could only run on Azure" has finally ended. On April 27, 2026, OpenAI and Microsoft announced a major overhaul of their partnership agreement, ending Azure's exclusive rights.
Microsoft remains a 27% shareholder (worth approximately ¥20 trillion) and retains IP usage rights through 2032, while OpenAI is now free for the first time to offer its products on Google Cloud, AWS, and Oracle.
Here's a plain-language breakdown of what changed and how it affects Japanese companies.
Let's start by organizing the basics of the news into three stages.
The main players are OpenAI, the maker of ChatGPT, and Microsoft, known for Windows.
On April 27, 2026, both companies announced a major revision to their partnership agreement, published simultaneously on the Microsoft official blog and the OpenAI website.
The biggest development was the "end of Azure exclusivity." Until now, OpenAI's products could only be delivered through Microsoft's Azure.
Under the new agreement, OpenAI gained the freedom to "offer all products across all clouds."
It's a transformation on the scale of "a manufacturer that could only wholesale to one exclusive retailer finally being allowed to supply other stores."
In addition, a "cap on revenue sharing" was established. OpenAI's 20% revenue share paid to Microsoft continues through 2030 but with a total ceiling.
Conversely, Microsoft's payments to OpenAI (such as cloud rebates) have been eliminated.
The direction is one of "putting a little more distance between each other's wallets."
Major media outlets including Bloomberg and CNBC reported simultaneously, making it the top news story in the AI industry.
It is an event symbolizing the tectonic shift in the AI industry in 2026.
This agreement didn't come out of nowhere — it's a continuation of the major restructuring that took place in October 2025.
On October 28, 2025, OpenAI completed its reorganization into a "Public Benefit Corporation (PBC)."
A "PBC" is a special type of corporation that writes into its articles of incorporation the pursuit of a "social purpose" in addition to shareholder profits.
The structure is: OpenAI Foundation (non-profit) as the controlling shareholder, employees/investors at 47%, and Microsoft at 27%.
Microsoft's holdings amount to approximately $135 billion (roughly ¥20 trillion), with OpenAI's valuation at $500 billion (roughly ¥75 trillion).
The impact is roughly equivalent to "a student council that incorporated into a company, with its biggest supporter ending up as a major shareholder holding 30% of the stock."
The agreement at the time also confirmed "an additional Azure order worth $250 billion (approximately ¥37.5 trillion)," with OpenAI continuing to pay massive cloud fees to Microsoft.
However, "Microsoft's right of first refusal" was abolished, making OpenAI free to enter contracts with other clouds.
That October agreement led to the fine-tuning that produced the current April update.
Over half a year, both companies settled on their "new sense of distance."
With the end of Azure exclusivity, OpenAI rapidly built a multi-cloud structure.
In November 2025, OpenAI signed a long-term computing resource contract with AWS worth $38 billion (approximately ¥5.7 trillion).
Next came a massive cloud contract with Oracle worth $300 billion (approximately ¥45 trillion) — the largest cloud deal in history.
OpenAI also partnered with Google Cloud for access to TPUs (Google's proprietary AI chips).
Azure $250 billion + Oracle $300 billion + AWS $38 billion — a total of approximately $600 billion (roughly ¥90 trillion) in computing resources secured.
It's the kind of risk diversification equivalent to "a ramen shop that was sourcing from one flour supplier switching to sourcing from four."
The reason: "computing resource shortages are the greatest business risk" — multiple sources reduce that risk.
Microsoft remains as the "primary cloud partner," and OpenAI products are deployed on Azure first.
However, if Microsoft cannot support a feature, OpenAI is allowed to launch it on other clouds first.
The AI industry of 2026 has fully shifted from "dominant dependence on one company" to "multipolar collaboration."
For customer companies, it's a favorable environment where options have exploded.
Here's an explanation of why they're loosening things "despite being such close partners," from three angles.
OpenAI's greatest motivation is "securing computing resources" and "establishing independence ahead of a future IPO (stock listing)."
In the AI industry of 2025–2026, "computing resources (GPUs/TPUs)" are the most precious commodity — Azure dependence creates supply shortage risk.
CEO Sam Altman reportedly said internally, "If we don't diversify risk with multiple clouds, we'll lose to Anthropic."
It's a fundamental strategic difference akin to "buying electricity from one utility company versus buying from several."
Furthermore, for the OpenAI IPO expected in 2026–2027, excessive dependence on a specific company is unattractive to investors.
The appearance of being "an independent company not bound to any single firm" raises valuation at listing time.
With this agreement, Microsoft's right of first refusal is gone, making OpenAI completely free.
Moreover, the revenue share cap protects future profits from the major shareholder.
OpenAI's valuation stands at $500 billion (approximately ¥75 trillion) as of April 2026, with $1 trillion (approximately ¥150 trillion) potentially in sight at IPO time.
It reflects the startup philosophy that "independence is the greatest asset."
The successful precedent of Anthropic operating on a non-exclusive basis is also part of the calculation.
Surprisingly, Microsoft also has "benefits from loosening the exclusivity."
Overdependence on a single AI company (OpenAI) means that OpenAI's troubles become Microsoft's troubles.
In 2024–2025, executive turmoil at OpenAI (the Sam Altman firing incident) caused Microsoft's stock price to fluctuate sharply in tandem — a painful lesson in risk.
It's the business maxim of "don't put all your eggs in one basket" put into practice.
It's a decision similar to "a baseball team owner diversifying away from relying solely on an ace pitcher by nurturing multiple young players."
In February 2026, Microsoft began offering Anthropic's Claude on Azure as well, promoting AI model multi-vendor adoption.
With this agreement, "Microsoft's payments to OpenAI have been eliminated," lightening Microsoft's financial burden.
IP usage rights are extended through 2032, securing practical long-term benefits.
It's a smart move of "giving up exclusivity while strengthening its position as a long-term partner."
Microsoft's stock price remained stable even after the April announcement — the market approves.
A strategy that is quintessentially Microsoft: "diversify and extend."
Another important change concerns "the method for determining AGI (Artificial General Intelligence)."
"AGI" refers to AI capable of handling a wide range of tasks at or above the level of humans — something that would transform society if achieved.
Under the previous contract, the rule was: "If the OpenAI board declares AGI has been reached, Microsoft's usage rights terminate."
In other words, OpenAI could unilaterally end the partnership simply by saying "this is AGI."
It was a structure as disadvantageous to Microsoft as "a rental contract where the property owner can tell you to vacate immediately just by saying 'I'm done renting.'"
The revised agreement changes this to: "AGI determination will be made by an independent panel of experts."
OpenAI alone cannot make the determination — third-party verification is required.
Furthermore, "Microsoft retains IP usage rights through 2032 even after AGI is reached" has been explicitly written in.
This gives Microsoft the assurance that "even in a post-AGI world, Microsoft can still access the technology."
OpenAI's revenue share (20%) continues through 2032, but with a total cap.
An ingenious design in which both companies mutually hedge long-term risks.
The relationship feels like "a couple who won't divorce but have arranged to live separately for a time."
It is also a judgment that contributes to long-term stability in the AI industry.
Here's a breakdown of the 2026 AI industry power map in three frameworks.
Until 2025, the AI industry had a two-camp structure: "Microsoft × OpenAI" and "Google × Anthropic."
"Microsoft × OpenAI" was exclusive (Azure only); "Google × Anthropic" was non-exclusive from the start (both AWS and Google Cloud).
With the April 2026 revision, Microsoft × OpenAI also became non-exclusive — both camps are now unified as "open."
It's an industry-wide tectonic shift comparable to "two major chain store groups, each of which had long dealt exclusively with separate suppliers, both simultaneously switching to an open approach."
As a result, "AI model × cloud" combinations are now freely available: ChatGPT × Azure/AWS/Google/Oracle, Claude × AWS/Google/Azure, and so on.
Customer companies can choose the "optimal combination," freely comparing on price, performance, and region.
The moment the industry standard shifted from "exclusivity" to "multi-vendor."
The Microsoft × OpenAI move is driving the entire industry forward.
2026 is the inaugural year of "AI democratization."
For users, it's a favorable environment with abundant choices and falling prices.
The competitive landscape among cloud providers has also changed dramatically.
AWS has invested a cumulative $150 billion (approximately ¥22.5 trillion) in Anthropic, deploying the Claude Platform exclusively.
In April 2026, Amazon announced an additional $25 billion (approximately ¥4 trillion) investment in Anthropic — a deal totaling roughly ¥15 trillion.
Meanwhile, Google Cloud has also invested up to $40 billion (approximately ¥6.4 trillion) in Anthropic, offering Claude Opus 4.7 via Vertex AI.
ChatGPT/OpenAI is now offered across all clouds: Azure, AWS, Oracle, and Google Cloud.
It's the kind of freedom where "all the major players in the industry simultaneously play on multiple teams."
As a result, competition intensifies among the three major cloud providers (AWS, Google, Microsoft) and rising challenger Oracle.
Oracle has set a target of "over $100 billion in AI revenue by 2027" and is growing rapidly.
The overall cloud market pie is expanding, creating opportunities for every player.
2026 marks the full onset of the "cloud warring states era."
Customer companies gain stronger negotiating leverage in cloud contracts.
It's a perfect time to optimize IT budgets.
The three-way structure of "two major camps + independent forces" remains very much alive.
Meta publishes "Llama" as open source and integrates it into its own services (Instagram, WhatsApp).
Elon Musk's xAI is independently developing "Grok," announcing in-house GPU production at SpaceX and a ¥260 trillion IPO plan.
China's DeepSeek released "R2," bringing price disruption to reasoning-type AI.
It's the kind of diversity seen in "the consumer electronics industry, where Sony and Panasonic collaborate while independent makers compete on their own terms."
The 2026 AI industry has a four-way structure: "OpenAI (multi-cloud)," "Anthropic (AWS/Google)," "Gemini (Google)," and "Meta/xAI/DeepSeek (independent)."
Running in parallel is the "AI arms race" as each company competes to reach AGI.
With OpenAI going multi-cloud, competition becomes more fair and transparent.
A healthy competitive environment where companies compete fiercely on technology and price.
The power map is expected to shift further in the second half of 2026 through 2027.
Which AI will emerge as the "winner" is still difficult to predict.
It is the biggest spectacle in the AI industry.
Here are three angles on how this relates to Japanese businesses.
The biggest good news for Japanese companies is that "ChatGPT Enterprise (the enterprise plan) will be available across multiple clouds."
"ChatGPT Enterprise" is a secure enterprise plan that does not train AI on a company's confidential data.
Until 2025, it was "Azure-only" — difficult to adopt for companies without an Azure contract.
Following the April 2026 revision, deployment via Google Cloud and AWS is planned, with Japanese region (Tokyo/Osaka) processing also available.
It's the kind of improvement in convenience akin to "an app that previously only worked on Android smartphones now also working on iPhones."
Japanese companies that already have Google Cloud contracts — NEC, KDDI, SoftBank, and others — will find it much easier to use ChatGPT.
Companies with existing AWS contracts, such as Rakuten, CyberAgent, and Mercari, will also face lower barriers to ChatGPT adoption.
In the second half of 2026, an Oracle Cloud version of ChatGPT is also expected to launch, further expanding options.
The era of "giving up on ChatGPT because we don't have Azure" is over.
It is a turning point at which AI utilization among Japanese companies will accelerate all at once.
Another impact is "the full-scale emergence of multi-cloud strategies among Japanese companies."
Until 2025, many Japanese companies concentrated on "one cloud" for cost and operational simplicity.
From 2026 onward, "multi-cloud × multi-AI model" becomes the new standard, with optimization by use case.
For example: coding with Claude (AWS), image generation with Gemini (Google Cloud), general-purpose chat with GPT-5.5 (Azure).
It's the kind of freedom to mix and match like "family restaurant for a packed lunch, company cafeteria for lunch, eating out in the evening."
Major SIers including Fujitsu, NTT Data, and Hitachi will also offer AI implementation services built on multi-cloud, multi-model approaches.
As of April 2026, it is estimated that over 50% of Japan's AI budgets will be invested in multi-cloud strategies.
"Skills to manage multiple clouds" will become essential for IT departments.
Price competition among cloud providers is projected to allow Japanese companies to reduce IT budgets by 10–20% compared to 2025.
The peace of mind that comes with escaping the risks of "concentration in one cloud."
2026 is a turning point in the DX strategies of Japanese companies.
The third impact is "price competition in API fees."
With four-way competition among Azure, AWS, Google, and Oracle, AI API prices have fallen 30–50% compared to 2025 in some areas.
For example, GPT-5.5 API pricing is subject to price competition across Azure, AWS, Google Cloud, and Oracle.
Customer companies can switch to "the cheapest cloud," strengthening their negotiating position.
It's a favorable environment like "the same PC being sold at four electronics retailers, and you can buy from whichever is cheapest."
Small and medium-sized enterprises face lower barriers to AI adoption — meaningful AI implementation is possible for a few thousand to tens of thousands of yen per month.
Startups can develop AI products with lower initial costs, also reducing the barrier to fundraising.
Individual developers and freelancers can broaden their use of AI for side work through API pay-as-you-go pricing.
In 2026, "AI democratization" is in full swing in Japan as well.
From "AI is for large corporations" to "a tool anyone can use."
First-mover advantage still remains substantial.
Now is the best time to start leveraging AI.
Chiemi is a section manager in the IT systems department of a mid-sized manufacturing company in Tokyo, and she had been wrestling with how to roll out ChatGPT internally.
Her problem: "Our main cloud is Google Cloud, so adopting ChatGPT kept getting pushed back because we'd need an additional Azure contract."
After seeing the April 2026 news, she confirmed that deploying ChatGPT Enterprise via Google Cloud would become possible.
No additional cloud contract needed — consolidated into the existing Google Cloud invoice.
Data processed in the Tokyo region, clearing internal security and compliance standards.
The ease of it was like "choosing a model of new appliance that works with your existing power strip."
As a result, deployment of ChatGPT Enterprise was shortened from three months to two weeks.
80 out of 100 employees in the company started using ChatGPT, with average work efficiency improving by 30%.
The burden on the IT department was also significantly reduced, with cloud contracts and billing consolidated.
She also formalized internal guidelines for "using ChatGPT, Claude, and Gemini for different purposes," establishing AI governance.
Six months later, she was recognized by management as a DX success story.
"Multi-cloud × multi-AI" is the new standard for mid-sized companies.
Ryota is the CTO of a SaaS startup in Osaka, and he had been working to optimize the costs of his AI infrastructure.
His situation: "ChatGPT API fees exceed ¥1 million per month, putting pressure on our profits."
After the April 2026 multi-cloud agreement news, he tested ChatGPT usage via AWS Bedrock.
The AWS version of ChatGPT was priced about 15% cheaper than the Azure version, and it integrated with the company's existing AWS infrastructure.
He spent a week on the migration and cut monthly AI-related costs by 30%.
It felt like "switching from buying drinks at a convenience store to buying in bulk at a wholesale supermarket."
Furthermore, he implemented a three-way AI approach based on use case: GPT-5.5 for summarization, Claude for code generation, and Gemini for image generation.
AI infrastructure costs were compressed from ¥1 million to ¥600,000 per month, improving profit margins by 10 percentage points.
The freed-up budget accelerated new feature development, with two new features released in three months.
User satisfaction improved, and the monthly churn rate dropped from 3% to 1%.
Several months later, his "multi-cloud AI strategy" was evaluated positively during fundraising, resulting in ¥2 billion raised.
He embodies the new startup standard of 2026.
Yuka is a freelance engineer in Fukuoka who was independently developing a SaaS tool as a side project.
Her problem: "Using the ChatGPT API via Azure costs ¥30,000 per month — a heavy burden for an independent project."
After the April 2026 news, she discovered that ChatGPT usage via AWS and Oracle would become available.
By utilizing AWS's free tier, she could use the ChatGPT API effectively for free in the first month.
And with pay-as-you-go pricing, she could keep costs to a few hundred to a few thousand yen per month based on usage.
It was the kind of cleverness of "making full use of free coupons handed out on the street to cut living expenses."
The cost optimization gave her the bandwidth to work on multiple personal projects in parallel.
Two months later, she launched the SaaS tool, which went viral on X (formerly Twitter), and she achieved ¥400,000 in monthly revenue in the first month.
She also differentiated from competitors by promoting "multi-cloud AI compatibility."
"Strategies for large corporations" benefited individuals too — she experienced firsthand the democratization of AI infrastructure.
It was a moment when the barrier to AI utilization lowered for freelancers and side-workers alike.
2026 marks the dawn of a golden era for individual creators.
It's the best time to seize first-mover advantage.
A. To "avoid the risk of depending on one company" and "secure long-term profits."
Being 100% dependent on OpenAI means that problems at OpenAI (such as the Sam Altman firing incident) become Microsoft's problems too.
The internal turmoil in 2024–2025 caused Microsoft's stock price to move sharply in tandem, bringing home the magnitude of the risk.
It's the business maxim of "don't put all your eggs in one basket" put into practice.
With this agreement, Microsoft began offering Anthropic's Claude on Azure as well, promoting AI model multi-vendor adoption.
In exchange, IP usage rights were extended through 2032, securing stable long-term access to OpenAI technology.
Additionally, "the elimination of Microsoft's payments to OpenAI" reduces Microsoft's financial burden.
It's a smart move of "giving up exclusivity while strengthening its position as a long-term partner."
Microsoft's stock price remained stable even after the April announcement — the market approves.
A strategy that is quintessentially Microsoft: "diversify and extend."
A decision that will be recorded in business history.
A. The second half of 2026 through 2027 is considered most likely.
With the PBC conversion in October 2025, the legal groundwork for an IPO is complete.
The April 2026 revision has now organized the appearance of "an independent company" through steps such as ending Microsoft's exclusivity.
OpenAI's valuation stands at $500 billion (approximately ¥75 trillion) as of April 2026, with $1 trillion (approximately ¥150 trillion) potentially in sight at IPO time.
It's roughly the stage of "a high school student completing their final preparations for university entrance exams."
An IPO would be the largest tech company listing in history, with a major impact on the stock market as a whole.
However, CEO Sam Altman has publicly stated "there's no rush on IPO timing," reflecting a stance of assessing market conditions and business maturity.
Rival Anthropic is also eyeing an IPO in 2026–2027 — the race between the two companies is worth watching.
After an IPO, OpenAI's financial information will be public, making revenue structure and growth rates transparent.
Individual investors will also be able to purchase OpenAI stock.
The second half of 2026 is the beginning of the "AI company IPO rush."
A historic turning point.
A. The basic approach is to "align with existing cloud contracts." For new contracts, compare based on cost, region, and features.
Companies with existing Azure contracts can continue via Azure — the fastest route with no additional contracts needed.
Companies with Google Cloud contracts will be able to use ChatGPT Enterprise via Vertex AI (expected to be available in the second half of 2026).
Companies with AWS contracts can use ChatGPT via Amazon Bedrock — already partially available.
Companies with Oracle Cloud contracts can use it via Oracle Cloud Infrastructure (second half of 2026).
It's the kind of abundant options where "you can just go to whichever type of convenience store is nearby."
All options support processing in Japanese regions (Tokyo/Osaka), with no data cross-border risk.
Technical documentation is also available in Japanese, with plenty of information on Zenn, Qiita, and official cloud blogs.
It's recommended to start by building a prototype using the free tier.
API usage fees are pay-as-you-go, allowing a small start at a few thousand to tens of thousands of yen per month.
Fees differ between cloud providers, so the tip is to compare periodically.
In 2026, simply "giving it a try" already yields significant first-mover advantage.
A. It was changed to "an independent panel of experts makes the determination."
"AGI" refers to AI capable of handling a wide range of tasks at or above the level of humans — something that would transform society if achieved.
Under the previous contract, the rule was: "If the OpenAI board declares AGI has been reached, Microsoft's usage rights terminate."
Since OpenAI alone could make the determination, the structure was disadvantageous to Microsoft.
The October 2025 revision changed this to a system where "an independent panel of experts" makes the determination.
Panel members are selected by mutual agreement of both companies, ensuring objective evaluation by a third party.
It's the same structure as "having a neutral panel of referees judge a match rather than the coach of one of the teams."
Furthermore, "Microsoft retains IP usage rights through 2032 even after AGI is reached" has been explicitly written in.
Even in a post-AGI world, Microsoft can continue to access the technology — a reassuring provision.
OpenAI's revenue share (20%) also continues through 2032, but with a total cap.
An ingenious design in which both companies mutually hedge long-term risks.
The relationship feels like "a couple who won't divorce but have arranged to live separately for a time."
It is also a judgment that contributes to long-term stability for the AI industry overall.
A. Yes — in fact, individuals and small businesses stand to benefit the most.
With four-way competition among cloud providers (Azure, AWS, Google, Oracle), AI API prices have fallen 30–50% compared to 2025 in some areas.
Specifically, ChatGPT API (GPT-5.5) fees allow meaningful AI utilization for a few thousand to tens of thousands of yen per month.
By using the free tier, you can try AI APIs effectively for free in the first month.
For example: the AWS Bedrock free tier, the Google Cloud Vertex AI free tier — each offers several hundred API calls per month at no charge.
It's a democratization where "high-end restaurant cuisine can now be casually enjoyed as delivery."
Concrete use cases: website creation for sole proprietors, inquiry response AI for small businesses, meeting minute summarization for freelancers.
The prevailing style is "subscribe to an API contract and use it with prompts" rather than "hire an AI engineer."
Technical barriers have also fallen — subscribe to a monthly service for a few tens of thousands of yen and implementation is possible without specialized knowledge.
Price competition between rivals (OpenAI, Anthropic, Google) is expected to drive usage fees even lower.
For individuals and small businesses that want to "use AI to improve work efficiency," 2026 is the perfect moment.
The right answer is to start by trying the free plan to see whether it fits your own business operations.
There is still plenty of first-mover advantage to be had.
The April 27, 2026 OpenAI × Microsoft partnership restructuring rewrote the rules of the AI industry in one stroke.
The end of Azure exclusivity, the revenue share cap, the non-exclusification of IP usage rights, the shift to an expert panel for AGI determination — every one of these is a historic change.
Yet Microsoft remains as a 27% shareholder, with IP usage rights extended to 2032, securing long-term practical benefits.
A paradigm shift from "exclusivity to openness," "concentration to distribution," and "short-term to long-term."
It is an event that symbolizes a new kind of relationship — like "a couple who continue their marriage while each having time that is truly their own."
For Japanese companies as well, ChatGPT becoming available via Google Cloud, AWS, and Oracle means that options have exploded.
An era in which everyone — startups, large corporations, and individual developers — can benefit from a multi-cloud AI strategy.
Your project today might open new possibilities simply by trying ChatGPT or Claude aligned with your company's cloud contract — so take that small first step to avoid being left behind by the wave of AI infrastructure democratization.
This article is a cross-post from AI Friends.
