Every Major Tech Partnership Ends the Same Way — Here's Why

In 1980, IBM needed an operating system for its new personal computer and didn’t have time to build one. Microsoft, then a small software company in Seattle, had a solution — or more precisely, had the ability to acquire one. For $50,000, Microsoft bought QDOS (Quick and Dirty Operating System) from a Seattle programmer named Tim Paterson, rebranded it as MS-DOS, and licensed it to IBM. The deal seemed to favor IBM: Microsoft was a tiny company, IBM was one of the most powerful corporations on Earth, and IBM’s PC would dominate the nascent personal computer market.

What IBM failed to notice was a single clause in the contract. Microsoft retained the right to license MS-DOS to other manufacturers. IBM thought it was buying an operating system. Microsoft understood it was building a platform. Within a decade, IBM’s PC business had been commoditized by clone manufacturers all running Microsoft’s operating system, and Microsoft had become one of the most valuable companies in the world. IBM eventually exited the PC business entirely. The partnership that IBM thought would secure its position in the emerging computer market had instead funded and legitimized the company that would ultimately subordinate it.

This story is not an anomaly. It is a template.

Every major tech partnership follows a recognizable lifecycle. Understanding the phases helps you see where current partnerships are in the cycle — and where they are headed.

The first phase is the value-add phase, and it genuinely is valuable. Two companies have complementary capabilities. One has distribution or scale; the other has technology or content. Together they can do something neither can do alone. IBM had manufacturing, marketing, and enterprise relationships. Microsoft had software. Apple had hardware design and a passionate user base. Google had search technology. The partnership delivers real value to both parties, and both parties grow. There is goodwill and genuine collaboration.

The second phase is dependency creation. As the partnership deepens, each company becomes structurally dependent on the other in ways that weren’t initially obvious. Users develop habits tied to both companies’ products. Third parties build on the combined platform. Switching costs accumulate. This phase looks like success, and it is — but it is also the moment when the terms of the relationship begin to shift. The company with more leverage starts to extract more value.

The third phase is the power struggle, triggered when one partner realizes the relationship has become extractive. The triggering moment is almost always the same: the junior partner becomes large enough to be a genuine competitive threat to the senior partner, or the senior partner tries to capture too much value from the relationship, or a new technological transition creates an opportunity for one partner to disintermediate the other.

The Apple-Google partnership illustrates the third phase clearly. Google Maps was for years the default navigation app on iPhones, and Google Search was the default search engine — generating billions of dollars in revenue for Google through a revenue-sharing arrangement with Apple. Apple and Google executives sat on each other’s boards. It was mutually beneficial, apparently stable, and seemed likely to persist indefinitely.

Then Google launched Android. What had been a partnership became, simultaneously, a deep financial relationship and an existential competitive threat. Apple, recognizing its dependency on Google services, launched Apple Maps in 2012 — famously broken, embarrassingly inferior to Google Maps at launch, but a strategic statement that Apple was willing to pay a short-term quality penalty to reduce its dependency on a competitor. Apple has since built its own search technology, its own maps, its own payment systems, and its own AI features. The partnership hasn’t ended — Apple still pays Google tens of billions of dollars annually to be the default iOS search engine — but it has been substantially reshaped by Apple’s deliberate effort to reduce the leverage Google holds.

The fourth phase — divorce, or at least separation — tends to happen slowly and then suddenly. The break rarely looks like a single dramatic event. It looks like a gradual withdrawal: a white-label relationship that gets taken in-house, a revenue-sharing agreement that gets renegotiated downward, a dependency that gets engineered away over several product cycles. IBM didn’t announce it was leaving the PC business on a specific date; it gradually exited over a decade as the economics made it impossible to compete. The divorce was the accumulation of many individual strategic decisions, each of which made sense in isolation.

The structural reason this pattern repeats is simple: in tech partnerships, the platform always has more power than the application. The platform controls the distribution channel, the customer relationship, the data generated by the interaction, and the terms on which the application can access users. The application provides value that the platform captures. As the application grows more successful, the platform has two options: extract more of the value, or try to internalize the function and eliminate the need for the partner. Both options are rational from the platform’s perspective. Neither is comfortable for the application partner.

This brings us to the Microsoft-OpenAI relationship, which is the most consequential tech partnership of the current moment and which follows the template closely enough to be instructive.

Microsoft invested heavily in OpenAI at a critical early stage, providing both capital and cloud computing resources through Azure. OpenAI, in return, gave Microsoft a substantial revenue share from its commercial products and, more importantly, exclusive rights to integrate OpenAI models into Microsoft’s products through the Azure OpenAI Service. For Microsoft, the investment transformed a stagnating productivity suite into an AI-first platform. For OpenAI, the investment provided the compute resources necessary to train frontier models without which the company couldn’t exist.

Both parties benefited enormously, and both parties are now navigating the dependency phase. Microsoft is clearly aware that a company as central to its AI strategy as OpenAI represents a significant dependency — one that competes with Microsoft’s own identity as a technology company rather than an AI service provider. Microsoft has hired AI researchers, built its own smaller AI models for specific use cases, and invested in alternative AI research. OpenAI, meanwhile, has been building out its own consumer products — ChatGPT, custom GPTs, voice interfaces — that increasingly bypass Microsoft’s distribution and compete directly with Microsoft’s own software products.

The power struggle phase is already visible. OpenAI has started offering enterprise contracts directly, competing with Microsoft’s Azure OpenAI Service. Microsoft has invested in competing AI companies and begun emphasizing its own AI research capabilities. The relationship has become simultaneously deeply financially entangled and increasingly competitive — exactly the structure that characterizes phase three.

The outcome is not predetermined, but the structural dynamics favor Microsoft over the long run, for the same reason IBM briefly favored over Microsoft: Microsoft is the platform, and OpenAI is the application. Microsoft controls the distribution channel — Windows, Office, Azure, Teams — through which most enterprise users access AI. OpenAI controls the technology, but technology advantages in AI have proven to be temporary and narrowing faster than almost anyone expected.

What’s genuinely uncertain is whether OpenAI can transition from being an application on Microsoft’s platform to being a platform itself — one that other applications are built on, and that users access directly rather than through Microsoft’s distribution. ChatGPT’s consumer scale suggests this is plausible. But becoming a platform requires a different kind of organizational capability than building research models, and the capital requirements are immense.

The IBM-Microsoft partnership ended with IBM structurally weakened and Microsoft dominant. The Apple-Google partnership is still nominally ongoing but has been substantially reshaped by Apple’s deliberate independence strategy. The Microsoft-OpenAI partnership will end — or be substantially renegotiated — within a decade, and the terms of that renegotiation will be determined by which company controls more of the distribution infrastructure for AI.

The pattern is clear. The mistake that losing partners consistently make is to confuse their current leverage — their technology, their content, their capabilities — with durable power. Technology advantages erode. Content gets replicated. Capabilities get internalized. Distribution, at the infrastructure level, tends to be far more durable than any of these things. If you are building a partnership in tech, the most important question to ask is: which of us will, in five years, be able to cut off the other’s access to users? The answer to that question tells you everything about who is really in control.

There is a variant of this dynamic that plays out not just between companies but between companies and developers — the ecosystem problem. Microsoft’s power over application developers in the 1980s and 1990s depended partly on its ability to set the terms on which developers could access Windows users. App stores have recreated this dynamic at scale: Apple and Google control the distribution channel for mobile software, and every application developer is in a partnership with Apple or Google whose terms Apple and Google dictate unilaterally. The antitrust fights over app store policies are, in structural terms, a version of the same power struggle that played out between IBM and Microsoft — except this time, the platform holders have learned from history and have worked hard to prevent any single application developer from accumulating enough leverage to become a genuine threat.

The question of when partnerships become antitrust problems is related but distinct. The IBM-Microsoft arrangement probably should have attracted more regulatory scrutiny than it did; Microsoft’s ability to exclude competition by controlling the operating system eventually did attract sustained antitrust attention in the late 1990s. The Google-Apple search deal has attracted similar scrutiny, with courts in multiple jurisdictions examining whether Apple’s receipt of billions of dollars to make Google the default search engine constitutes anticompetitive behavior. The Microsoft-OpenAI relationship will likely face its own regulatory examination as the partnership’s effects on AI competition become clearer.

What regulation cannot change is the underlying structural dynamic. Platforms will always have more leverage than applications because of their control over distribution. Regulation can limit the most egregious forms of extraction and can preserve space for competition, but it cannot make application developers into platform owners. The only viable strategy for the application partner is to build toward platform status — to accumulate your own distribution, your own user relationships, and your own ability to create switching costs for the users you serve. Companies that succeed in this transition graduate from being partners to being peers. Companies that don’t remain structurally subordinate, regardless of how valuable their technology is. The IBM-Microsoft story is, at its core, a story about what happens when you control the chokepoint everyone else depends on. Every major tech partnership is an argument about who that chokepoint belongs to.