Collaborative Competition: Winning in AI by building with your rivals

NVIDIA’s biggest customers, AWS, Google, Meta, and OpenAI, are all developing their own AI chips to compete with it.

OpenAI, once exclusively tied to Microsoft, has started diversifying its compute partners, including a new collaboration with Oracle.

Despite being rivals in the AI hardware, Google and NVIDIA partnered to bring agentic AI to enterprise customers through Google Cloud.

These aren’t anomalies. They’re features of a strategic shift happening across the AI industry.

In the AI landscape, defined by high fixed costs, rapid innovation cycles, and deeply entangled value chains, the boundaries between competitor, customer, supplier, and partner are blurring. What we’re seeing is the rise of co-opetition, a dynamic where companies compete and collaborate at the same time, with the same players.

This article explores:

• Why co-opetition is a necessary strategy

• How frameworks like the Value Net help us make sense of this shift

• What it means to “grow the pie” when everyone’s also trying to carve their slice

• And why this matters for anyone building in the AI ecosystem, from chipmakers to app developers

Traditional business strategy maximizes performance within existing market structures, aiming to capture a larger share of a fixed pie. The dominant mindset is zero-sum: for one player to win, another must lose.

But in the AI ecosystem, this logic is breaking down.

The lines between competitor, supplier, partner, and customer are increasingly blurred. What’s emerging is a positive-sum dynamic, where collaboration expands the pie and creates more value for everyone involved.

This is the strategic logic behind co-opetition. Rooted in game theory and captured in models like the Value Net, co-opetition redefines strategy as a dynamic interplay of competition and collaboration.

It recognizes that business relationships have more than one dimension. And it demands a mindset shift:

• From capturing value to creating it

• From fixed roles in the value chain to dynamic, fluid alliances

• From guarding advantage to growing ecosystems

In a co-opetitive world, the real edge lies in making the whole game more valuable and positioning yourself to benefit from that growth.

In most industries, business relationships are relatively straightforward, defined by a value chain. But in AI, those definitions are fluid.

For instance, a cloud provider is a compute supplier to a Foundational Model developer, a competitor with its LLMs, and a customer embedding foundational models into its services. A hardware manufacturer might enable the ecosystem with open tooling, while simultaneously climbing the stack to capture more value.

This is a feature, not a bug.

Source: Randy Fath on Unsplash

To make sense of co-opetition, we should refer to the Value Net Framework developed by game theorists Adam Brandenburger and Barry Nalebuff, a model that reframes business as a multi-player game, not just a competitive race.

At the center of the Value Net is a player (a company), surrounded by four key relationships: Customers, Suppliers, Competitors, and Complementors (whose offerings increase the value of yours).

The Value Net highlights both competitive and cooperative dynamics at play simultaneously.

It introduces several key ideas from game theory:

• Complementarity: Another player’s success increases your payoff. When LLM becomes more useful, it drives greater demand for GPUs.

• Strategic Interdependence: Your best move depends on what others do. Collaborating creates more value than going it alone.

• Coalition Formation: Relationships are multi-faceted. You might cooperate with another company on common protocols while competing on the core algorithm.

• Two-phased approach: In the value creation phase, players collaborate to make the pie bigger, and in the value capture phase, players compete for their share of value.

Using this approach, you can partner to grow the ecosystem and still compete aggressively to win within it.

Let’s walk through the AI stack to see how this co-opetitive dynamic plays out in practice.

NVIDIA dominates AI hardware with its GPUs, but its biggest customers like AWS, Google, Microsoft, and Meta are also developing their own AI chips.

Meanwhile, NVIDIA is expanding up the stack into AI frameworks and inference services, increasingly encroaching on the territory of the hyperscalers.

Foundation model companies like OpenAI and Anthropic rely on cloud providers for training and deployment, but these hyperscalers are also their customers and competitors.

Microsoft is OpenAI’s investor, exclusive cloud partner, and customer, while Microsoft supports other Foundational Models. Amazon has invested billions in Anthropic, giving AWS early access to Claude, even as Bedrock supports multiple models.

Enterprise software giants like Salesforce, SAP, and Oracle are embedding LLMs into domain-specific workflows. They use models from companies like OpenAI or Anthropic, but fine-tune them, wrap them in proprietary interfaces, and route customer data through their own layers.

Model providers are upstream enablers, but could be commoditized if value capture shifts entirely to the enterprise layer. Meanwhile, enterprise platforms are dependent on upstream innovation and new model breakthroughs.

Startups are building aggressively on open models and APIs. Whether it’s legal AI, coding copilots, or design tools, they rely on foundation models, but layer in data, UX, and orchestration to create defensible vertical solutions.

But as they scale, they may undercut the platforms they depend on. In response, platform players are moving downstream, building copilots, launching app ecosystems, or acquiring the startups outright. The relationship swings between symbiosis and substitution.

Co-opetition is the only viable strategy in this environment that is continuously evolving and announcing breakthroughs every other week. The winners focus on shaping the structure of the market beyond maximizing just their market share.

So, what does all this mean if you’re operating anywhere in this AI stack?

Pure-play competitive strategy is not enough. Whether you’re building chips, training models, shipping software, or orchestrating workflows, your success will depend on how well you collaborate, including with your fiercest rivals.

You could co-opetite in the following ways:

Don’t consider yourself as just a supplier, a competitor, or a customer. You’re all of them at the same time. Treat each relationship as multi-dimensional and value accretive.

Trust is a prerequisite for co-opetition. Build trust by aligning incentives, defining legal and technical guardrails. But don’t let risk-aversion impede progress.

Design to collaborate. Make it easy for others to build on, plug into, or extend your offering. Platforms that win in AI will act as a connective tissue across ecosystems.

Focus on increasing your added value to others in the network. That’s what makes you an indispensable partner. You don’t have to own or control everything.

The most strategic actors are influencing standards, shaping ecosystems, brokering alliances, and making headway on multiple fronts.

The AI future will be co-created. And co-opetition is a capability that will separate those who define the game from those that just ride along.

For those AI builders, ask not what your ecosystem can do for you – ask what you can do for your AI ecosystem!