Strategic Management Explorer

What the Hyperscaler Race Teaches Managers About AI Strategy

By Abhinav Ray

People are measuring the generative AI race in many ways: everyday use, AI use cases, capital expenditures, acquisitions, and workforce impact. Behind all of these measures is the same basic question: who is likely to win the AI race, and why? The answer has less to do with which firm builds the best model and more to do with which firm can orchestrate the ecosystem of suppliers and complements that surround it.

Hyperscalers, including Amazon, Microsoft, Google, and Meta, are driving much of the innovation in AI, either directly or through investments and partnerships. But they cannot succeed alone. To win, they need a broader AI ecosystem to innovate with them.

Ron Adner and Rahul Kapoor’s paper, “Value Creation in Innovation Ecosystems,” published in the Strategic Management Journal in 2010, offers a useful way to study this kind of innovation. Their framework helps managers understand where value is created, where bottlenecks appear, and why the AI ecosystem is evolving the way it is today.

The Practitioner “So What”

Find where the binding constraint sits in your ecosystem right now — because when it’s on the supply side you win by out-spending and in-sourcing, but when it’s on the adoption side, more capital buys you nothing until the workflows, tools, and users around your product catch up.

The Research: Understanding Adner and Kapoor’s Framework

‘Focal’ Firms

In Adner and Kapoor’s framework, focal firms are the main firms whose technology sits at the center of the ecosystem. Here that technology is the frontier AI models and the cloud platforms that serve them: Amazon’s Bedrock and AWS infrastructure, Microsoft’s Azure AI stack, Google’s Gemini models and Cloud platform, and Meta’s Llama family. Throughout this article, the Hyperscalers are the focal firms.

Components

Focal firms do not build everything themselves. They rely on suppliers for the inputs they need to build the focal technology. Those inputs include AI accelerators and GPUs, high-bandwidth memory, high-speed networking equipment, data center capacity, and electrical power. In the AI ecosystem, one important category of components is advanced memory chips and related hardware supplied by companies such as SK Hynix and Samsung Electronics.

Complements

On the demand side sit complements — the products, services, workflows, or capabilities that customers need in order to use the focal technology. In the AI ecosystem, complements include AI agents, AI-embedded workflows, and devices that make AI easier to use. Examples include AI-enabled PCs with on-device inference chips and smartphones with built-in assistants. These complements help end users and companies adopt AI in their daily work and business processes.

The framework does more than label the actors in an ecosystem; it explains how different kinds of bottlenecks shape who captures value as a technology develops. Adner and Kapoor’s hypotheses suggest that when the biggest technological challenges sit in components, technology leaders can widen their advantage by solving or securing access to those inputs. When the biggest challenges sit in complements, however, leaders may struggle to turn technical progress into end-user value because adoption depends on other products, workflows, or capabilities maturing alongside the focal technology.

Finally, as the technology life cycle progresses, vertical integration — a focal firm bringing an activity it previously bought from suppliers inside its own boundaries — can become increasingly valuable because it helps focal firms reduce uncertainty and control the parts of the ecosystem that matter most.

Applying the Hypotheses to the AI Ecosystem

Component challenges give Hyperscalers an opportunity: Rising input costs create a chance to build competitive advantage by investing heavily in capital expenditures to overcome scarce component challenges. Microsoft has signaled roughly $175 billion of capital expenditure in 2026, the majority directed toward AI infrastructure, while Meta has raised its 2026 capital expenditure to approximately $130–$145 billion to accelerate data center construction. Spending at that scale lets the largest firms secure scarce supply on terms smaller rivals cannot match. This fits Adner and Kapoor’s finding that supply-side challenges can allow technology leaders to move faster than rivals and strengthen their advantage.

Complement challenges are a risk to Hyperscalers: They also face a different problem — many complements are not yet mature enough to create clear value for end users who are willing to pay. For example, Microsoft, one of the Hyperscalers and focal firms, has had limited success generating revenue from Copilot because adoption inside organizations has been slow. The company reported that Microsoft 365 Copilot passed 20 million paid seats in April 2026, but that is still only about four percent of its roughly 450 million commercial Microsoft 365 seats. Unless organizations redesign workflows and encourage employees to use AI, Copilot usage will remain limited. That creates a risk for Microsoft’s broader AI ambitions.

Vertical integration provides an advantage as technology matures: According to Adner and Kapoor, vertical integration becomes more valuable as a technology matures. As component suppliers improve, they can help remove supply bottlenecks, but they do not eliminate contractual uncertainty. To reduce that uncertainty, focal firms may choose to vertically integrate. Hyperscalers have already started to do this selectively, designing custom chips in-house rather than buying from Nvidia. Google has built its own tensor processing unit (TPU) capability instead of relying on Nvidia’s GPUs, and Amazon is developing its Trainium and Inferentia chips.

Taken together, these examples show that the AI ecosystem is in a phase where component challenges are helping Hyperscalers take the lead, but complements need to mature quickly if that lead is going to last.

At the same time, Hyperscalers are trying to vertically integrate — designing their own accelerators, building and operating their own data centers, and contracting directly for power generation — so they can consolidate their position and stay ahead of competitors. Adner and Kapoor’s research gives business practitioners a framework for understanding this kind of innovation cycle and for asking better strategic questions about AI.

Practical Implications for AI Strategy: What Strategic Leaders Should Do

Managers of focal firms investing in AI should ask three questions during this stage of the innovation cycle:

Where are the bottlenecks in the ecosystem right now? AI strategy should start by identifying bottlenecks in the ecosystem, not only by evaluating a firm’s complements or internal capabilities.
Are we moving faster than the ecosystem? Being early does not always lead to value capture if the necessary components and complements are not ready.
Where can we vertically integrate to gain an advantage? As the technology matures, firms should weigh building capability in-house against acquiring it, taking control of the critical parts of the value chain while still relying on partners for the rest.

The AI race is often portrayed as a competition among Hyperscalers, but Adner and Kapoor offer a different perspective. Success does not depend only on who develops the best models or spends the most money. It depends on how effectively firms orchestrate an ecosystem of components and complements.

For managers, this shifts the strategic question from “How can we build better AI?” to “How can we position ourselves within an ecosystem that captures value?”

In other words, the winners of AI will not necessarily be the firms with the most advanced technology. They will be the firms that build, coordinate, and participate in the ecosystems that make AI valuable.

Abhinav Ray is a strategy practitioner with 12+ years of experience across corporate strategy and M&A. He has a keen interest in innovation ecosystems, competitive advantage, and the strategic implications of emerging technologies. He enjoys applying academic research to contemporary business challenges and industry development.