Every summer the Tour de France turns endurance into theater. For three weeks the world watches cyclists traverse mountains, windswept flats, and chaotic city finishes. If you want to see strategy in motion, watch the peloton. Riders from different teams—direct competitors with different sponsors, leaders, and ambitions—form a dense, fast-moving pack. Inside this temporary, informal alliance, everyone benefits from reduced wind resistance. The front rider does the most work as the riders behind conserve energy. Positions rotate. Over time the group moves faster than any rider could alone.
And then, suddenly, the truce dissolves. A climb steepens. Speeds drop. Aerodynamic advantages fade. The contenders must choose the optimal moment to break away. In the Tour, cyclists are constantly oscillating between two modes: collaboration when it makes the whole group stronger, and competition when differentiation decides the outcome.
In today’s markets, speed, interoperability, and trust determine whether new technologies gain traction. Most companies can’t afford to build a technological ecosystem alone. This is partly why we see so many circular deals in the AI sector. AI systems depend on shared infrastructure that no single company fully controls. Training models require chips, cloud capacity, data pipelines, software tools, and distribution channels, which are owned or supplied by different firms. Circular deals, in which companies invest in one another or sign mutually beneficial partnerships, help align incentives. Firms share key components, lock in supplies, and signal their commitment to a common ecosystem. Such commitments have led fierce competitors in this space to band together to create shared-standards groups (for example, the Agentic AI Foundation) that help ensure that in the future, ecosystems remain interoperable. In effect, this sharing reduces risk. Each participant knows the others have skin in the game, which makes it easier for the whole technology ecosystem to scale up quickly.
But if companies collaborate too broadly, they risk losing hard-won competitive advantages. That means the real strategic question is not whether to collaborate but where to do it. To answer that, it’s necessary to separate a company’s activities into two layers. The first, the core, includes the infrastructure, interfaces, standards, and safety practices that allow its ecosystem to function and grow. The second layer, the edge, is where companies create distinct value for customers by using proprietary workflows, integrated experiences, brand loyalty, data, and business models. The challenge is that the boundary between the two keeps moving. What differentiates you today may become commoditized tomorrow, and collaboration increasingly happens through informal ecosystems rather than traditional alliances. For that reason, leaders need a practical framework to decide, case by case, when collaboration strengthens their strategy and when it quietly erodes it.
What follows is the framework I use to help leaders make that call. It is based on my academic research, interviews with dozens of industry practitioners, and my work as a consultant. It’s built around five factors and grounded in identifiable patterns across both technology-heavy and low-tech industries. If the Tour has a lesson for business, it isn’t just “collaborate more.” It’s “collaborate intelligently.” Draft when it makes you faster. Break away when it’s time to win. Businesses must learn to do the same.
The Hidden Cost of Independence
Most leaders don’t struggle because they dislike collaboration. They struggle because there’s a high cost to failure, and failures happen frequently. If you undercollaborate, you can lose years to duplicated work, fragmented standards, and ecosystem distrust. If you overcollaborate, you can accidentally train your competitors, leak strategic know-how, or commoditize what you once owned. In companies’ practices, I see three predictable mistakes.
First, executives often talk about cooperation as if it’s binary: “We partner” or “we don’t.” But collaboration comes in many forms, ranging from lightweight coordination (shared standards or shared threat intelligence) to deeper codevelopment (formal R&D partnerships). The idea that collaboration must mean a full alliance with shared teams, road maps, and governance creates unnecessary fears about control, dependency, and value capture. The better model is to treat collaboration as a spectrum of options, each with different risk levels and different strategic payoffs.
Classic co-opetition guidance emphasizes how much of the challenge is about scope and control. How far does cooperation extend? Who governs the collaboration? And how can it be unwound if it stops making sense? Once leaders see collaboration as a set of tools, not a single big bet, they can choose the lowest-risk mechanism that still achieves their goal.
The second mistake is believing that collaborating on a shared core is mostly an act of goodwill. That common belief becomes self-fulfilling: Leaders starve collaborative efforts because they can’t see the ROI, and then those efforts fail because they’re underresourced. But many of the most impactful examples of collaboration are best understood as competitive moves designed to reshape the terrain of competition.
Apple and Google’s joint work on exposure notification during Covid-19 is an excellent example. Two companies that compete intensely at the platform level built an interoperability layer that allowed public-health authorities to broadly deploy Bluetooth-based exposure notifications. Apple framed the collaboration as an effort to help health agencies reduce the spread of the disease (a social good for everyone) while focusing on the enhanced privacy and security of the iPhone (a point of differentiation between iOS and Android often underappreciated by consumers). Even if the circumstances were unusual, the strategic logic is familiar: When customers’ adoption of a product depends on cross-platform interoperability and trust, the core is too important to remain fragmented. But collaboration on the core can offer an opportunity to shine a light on key aspects of differentiation.
The third mistake is blindly following the traditional cooperative strategy playbook. Classic cooperative strategy tends to emphasize formal alliances, joint ventures, and contractual partnerships. Those still matter. But in the digital age many of the most important cooperative dynamics occur outside a formal framework. As technologies become more modular, layered, and ecosystem-dependent, firms increasingly collaborate through informal or community-driven approaches. That creates a new challenge: Leaders can’t rely only on the familiar joint-venture checklist. They need a way to evaluate newer forms of collaboration.
The Five-Factor Framework
The economic logic behind the framework is that collaboration increases value for everyone. Collaborating on the core can either raise customers’ willingness to pay (through increased trust, interoperability, quality, and adoption rates) or lower costs (through shared R&D and reduced duplication). Competing on the edges is a value-capture play. It’s where you build the differentiated experiences, workflows, distribution, and brand promise that allow you to command premium pricing, lock in customers organically, and sustain or improve your competitive advantage. But where do you draw the line between collaboration and competition?
The good news is this line isn’t mystical. Across industries, it tends to be shaped by five predictable factors. Once leaders explicitly incorporate those factors into their deliberations, they stop arguing about vague philosophies (“open versus closed” or “partner versus build”) and start making clean, revisable decisions.
Now let’s look at how each factor shapes the boundary between the core and the edges and how real companies have used the framework.
Factor 1: Winner-take-all dynamics. Some markets reward interoperability and allow many winners. Others tend to tip toward one dominant platform, standard, or ecosystem. In winner-take-all (or winner-take-most) markets, collaboration can be unusually risky or unusually powerful, depending on how it affects the tipping point. If the market is likely to tip toward a single-owner-controlled platform, you must be careful not to fund your future captor. If the market is likely to tip toward an open or shared standard, collaborating early can help prevent fragmentation and ensure you’re helping control the standard while still competing on the edges that matter to customers. If the market is unlikely to become winner-take-all, then more collaboration can be helpful.
Consider the technical body EMVCo, which is overseen by major payment networks including American Express, Discover, JCB, Mastercard, UnionPay, and Visa. It develops global specifications so that card-based payments work securely and seamlessly. Payments aren’t a winner-take-all business. Most people have multiple credit cards and use more than one payment processor for purchases. The major payment networks collaborate on the core rails because they all benefit from interoperability. But they compete intensely on fraud tooling, issuing relationships, merchant services, pricing, and rewards. That’s a classic “open standard core, differentiated edges” move. Standardize what must work everywhere; compete on devices, ecosystems, and user experience.
In winner-take-all markets, the goal is neither to collaborate nor not to collaborate. The goal is to collaborate in ways that prevent the market from tipping toward competitors and to ensure the competitive battle happens where you can win.
Factor 2: Technology life-cycle stage. Are you growing the category or fighting for share? Technologies tend to follow an S curve: uncertain early days, rapid growth, maturity, and eventual decline. The balance between collaboration and competition shifts along that curve. In early stages, a company’s biggest problem is not usually beating rivals. It’s getting adoption: proving viability, establishing standards, reducing uncertainty, and building complements. Collaboration can be a way to accelerate the growth of the category itself. As a technology enters the growth and maturity stages, the strategic path moves toward differentiation: feature advantage, customer experience, cost position, and product stability. Collaboration still happens, but often narrowly and defensively (around shared risks like security, for example). Finally, in the decline stage, as a technology is being phased out or replaced, competitors often collaborate to extend its life, especially if they are not the ones building the replacement technology.
Let’s take a look at the electric vehicle (EV) industry. Early in the EV transition, the category faced a classic adoption barrier: Consumers, suppliers, and infrastructure providers needed confidence that EVs would be a viable market. Tesla decided to open access to certain battery and charging patents to encourage broader industry adoption and ecosystem investment. The company collaborated with other EV manufacturers on parts of the enabling environment while still competing on product, brand, and execution. For instance, Tesla supplied battery packs for Toyota’s RAV4 EV. Although many onlookers questioned this move at the time, it helped grow the EV market and had the added benefit of giving Tesla control of the standard for a growing industry.
Such collaboration early in the technology life cycle happens across a variety of industries. For example, in the pharmaceutical domain, “precompetitive consortiums” allow competitors to collaborate on fundamental research that will help them create new drugs and therapeutics.
The consortium led by Regeneron to sequence the genomes of 500,000 participants in the UK Biobank is one example. Regeneron and its competitors AbbVie, Alnylam, AstraZeneca, Biogen, and Pfizer each contributed $10 million to support the effort. Their collaboration sped up the sequencing timeline by three years. The companies that participated had access to this extremely important data and could use it to create new drugs.
For an example of a lack of collaboration in the growth stage, look at the streaming services industry in the late 2010s. As competition intensified, major players such as Netflix, Amazon Prime Video, and Hulu sought to differentiate themselves by offering unique user experiences and exclusive content and using proprietary algorithms. Because they were focused on capturing market share and establishing brand dominance, they competed with one another instead of collaborating.
As noted earlier, once a market reaches the decline stage, its players often revert to collaboration. For example, consider compact discs (CDs). As digital streaming rose to prominence and CD sales slowed, CD manufacturers and music labels collaborated on marketing efforts and special releases to try to extend the life of the technology in the face of disruption.
Life-cycle thinking prevents executives from making a common error: assuming that because a collaborative move made sense early, it remains the right move forever. The core often stabilizes over time, and the edge battle intensifies.
Factor 3: Technology stack position. Technologies live in stacks. Lower layers (or infrastructure) benefit most from interoperability, while higher layers (applications) benefit most from differentiation.
We see this dynamic regularly within big tech. For example, when Google developed the cloud technology Kubernetes (which automates the management of applications, among other things), the company understood that because it was at such a low level in the tech stack, it wasn’t going to drive competitive advantage. Therefore, Google shared it freely with the world and then donated it to the nonprofit Cloud Native Computing Foundation. Within a few years, Google’s primary competitors in the cloud computing space (Microsoft’s Azure and Amazon’s Web Services) adopted Kubernetes and started relying on it heavily. They also contributed to its development. Although Google had given away an important piece of technology to its competitors, doing so created value by helping to speed adoption of cloud computing. In addition, Google got a multiyear head start over competitors on deploying and integrating this now-crucial technology.
Conversely, consider Uber and Lyft, which compete at higher levels of the technology stack. Although they rely on similar underlying technologies for their mobile apps (GPS navigation and payment processing, for example), the two companies engage in fierce competition to differentiate their service offerings, user experience, and brand position using technology in the middle (pricing and routing algorithms) and top (user interface) of the stack.
What does this mean for executives? If you’re working on infrastructure layers, such as protocols, interfaces, safety practices, and enabling components, you should default to asking, “What must be interoperable to unlock adoption?” Those core areas are candidates for collaboration.
If you’re working closer to the end customer—for instance, on workflow integration, proprietary data, branding, or distribution—your default question should be, “What must remain unique for us to win?” The answers are your candidates for competition.
Stack position alone won’t make the decision for you, but understanding it dramatically improves the odds that you’ll draw the line separating the core from the edge in the right place.
Factor 4: Level of competition. When rivalries are intense, shared risk increases. It’s intuitive that low competition can make collaboration easier: Firms feel less threatened, and market-building benefits are clearer. Although the incentive to collaborate decreases as competition approaches moderate levels, what’s less intuitive, but extremely important, is that high competition can also increase collaboration in specific domains, especially around shared risk. Cybersecurity is one example.
Security is adversarial; attackers exploit the weakest link. If one firm learns about a new vulnerability and sits on it, an attacker can exploit it elsewhere, which ultimately hurts customers and erodes trust in the entire category. As a result, competitors often share intelligence on threats even while competing hard for customers. In 2014 Fortinet, Symantec, McAfee, and Palo Alto Networks created an informal agreement to share cybersecurity intelligence. The effort was so successful that the firms later formalized it by creating the Cyber Threat Alliance, which is still active today and has dozens of member companies, all of whom are competitors. From a strategy perspective, this is “collaborate on the core risk layer, compete on the edges of product and service.” You collaborate in areas where the marginal value of shared learning is enormous and where failure harms everyone’s legitimacy. You compete on what you can charge customers for (outcomes, integration, performance, and trust).
High competition also raises the bar for collaboration governance. If you’re collaborating in one area while fighting intensely elsewhere, you need sharper boundaries, clearer scope, and explicit paths for unwinding joint efforts.
It’s important to recognize that these principles apply to product market competition. With labor market competition (that is, when companies are competing for the same employees), the relationship is different. When labor competition is moderate, companies can offer collaborative opportunities (on R&D, for example, or on open source software) as a perk to attract better talent. However, when labor competition is low, companies don’t need to offer this benefit, and when it is high, they cannot afford to.
Factor 5: Social acceptance and regulation. When legitimacy is at stake, collaboration can help stabilize a market. Some technologies face skepticism, backlash, or regulatory uncertainty. In those cases, collaboration often becomes a way to build legitimacy, establish responsible standards, and avoid the chaos of conflicting rules.
Apple and Google’s cooperation on exposure notification is again instructive. The strategic subtext was simple: If public trust collapses, adoption collapses. Collaboration on a privacy-preserving core enabled broader trust and increased usability while leaving the companies free to compete vigorously in other areas.
Now zoom out beyond “technology” in the narrowest sense. Technology is broader than software and the digital economy. It includes products, processes, and even category-level practices. The five-factor framework still applies. Let’s look at the soft drink industry. As public concern about sugar and obesity rose, major beverage companies came under scrutiny. One response was collective action—industry commitments and formal and informal partnerships. A 2014 collaborative effort between fierce competitors Coca-Cola, Dr Pepper Snapple Group, and PepsiCo aimed to reduce beverage calories consumed per person nationally. In one TV campaign, delivery people from each company happily walked down the street together to let consumers know they were all working collaboratively to address the public’s concern.
Similar cross-industry efforts are commonly seen in times of regulatory uncertainty, when competitors will work together to help shape the laws that govern their industry. From a strategy perspective, this follows the same pattern: Collaborate on the core legitimacy layer (shared commitments, transparency norms, public messaging, and measurable goals), and compete fiercely on products, branding, pricing, and distribution, which are the edges where customers choose who wins.
When social acceptance and regulation are central, treating legitimacy as a PR problem is the wrong move. It’s an ecosystem design problem. Collaboration can be the fastest way to set credible norms, especially when regulators are watching and the alternative is a patchwork of incompatible or punitive rules.
Anthropic and OpenAI’s Collaboration
To see all five factors in one contemporary case, look at Anthropic’s decision to share the Model Context Protocol (MCP), a standard that allows AI models like LLMs and agents to talk to external datasets and tools. Anthropic donated the MCP to the newly formed Agentic AI Foundation (AAIF), which it has joined with its direct competitor OpenAI (as well as other competitors in the AI agent space, including Google and Microsoft).
Anthropic competes in a market in which ecosystems form around developer tools, integration, and compatibility. Using multiple platforms simultaneously is possible but costly for enterprises. Establishing standards can reduce market fragmentation, and whichever company shapes the standard influences the direction of the market. Anthropic’s announcement says it wants to keep MCP “open, neutral, and community-driven” as it becomes the established, enabling infrastructure. If we assess that move using the framework, all five factors suggest this was the right decision.
Winner-take-all: If the agent ecosystem tips, it will likely tip toward shared interfaces and developer defaults. By sharing MCP with competitors, Anthropic put itself in a strong position to help set the direction of the standard.
The tech life-cycle stage: Agentic AI systems are in an early phase, which means market growth depends on reducing friction and uncertainty.
Stack position: MCP sits in the enabling layer, an interface and protocol that allows tools and models to work together; competitive applications can be built on top of it.
Level of competition: Rivalry is intense. No firm wants to bet the future on a competitor’s proprietary interface. Neutral governance by the AAIF lowers that risk.
Social acceptance and regulation: As AI scrutiny grows, credible governance and openness can improve trust and accelerate responsible adoption.
Anthropic isn’t collaborating to be nice. It recognizes that collaborating on the core can be the fastest way to grow a market and prevent it from moving in a direction Anthropic doesn’t want. It has plenty of room to compete on the edges with things like models, products, safety systems, enterprise workflows, pricing, distribution, and brand.
How to Get Started
A good place to start is to map your technology stack. Identify the layers that you operate in—from the infrastructure and interfaces that enable your technology to the user-facing experience. Consider what customers value at each layer. That will help clarify which parts function as shared foundations and which ones shape the customer experience.
Next, list potential core candidates. These are components for which interoperability, trust, or shared risk reduction could increase adoption or lower costs. At this stage it helps to cast a wide net. You can narrow the list later.
Then evaluate each candidate using the four other factors of the framework. Ask whether the market is likely to tip toward a dominant standard and where the technology is in its life cycle. Consider how intense competition is and whether rivals face shared risks like security threats. Next, think about the status of social acceptance and the regulatory environment.
Once you’ve done that, choose the lightest collaboration mechanism that will achieve the goal. Often you don’t need a formal alliance. A shared technical standard, a neutral-governed project, or shared threat intelligence may be enough. The key is defining the scope of collaboration and the control mechanisms clearly from the beginning and ensuring that you are complying with antitrust regulations.
You must identify where to draw the boundary between collaboration and competition. Be explicit about what information remains part of your competitive edge and stays off-limits to your collaborators, how decisions will be made, and how a collaborative arrangement could change if conditions do.
It’s important to design your edge strategy in parallel. A common mistake companies make is building a strong shared core but neglecting the differentiated capabilities that win customers. Decide early where you will be meaningfully different from competitors and invest there.
Finally, revisit the boundary regularly. As markets mature, yesterday’s edge often becomes tomorrow’s core. Reexamining the line every six to 12 months helps ensure that your strategy evolves with the market.
. . .
The Tour de France is not a lesson simply in teamwork or in individual heroics alone. It’s a lesson in switching modes at the right moment. And it’s one that companies should take to heart. Build everything yourself and you’ll arrive late; collaborate without boundaries and you’ll help rivals catch up or surpass you.
The strategic skill is drawing the line: Collaborate on the core to grow the market, reduce risk, and accelerate adoption, then compete on the edges to deliver unique value and capture it. In other words, ride in the peloton when it makes you faster, then choose your moment and break away.
Copyright 2026 Harvard Business School Publishing Corporation. Distributed by The New York Times Syndicate.
