Comparative Study on Innovation Strategies of Ecosystem Companies with Generative AI Platforms

2-1 The Importance of Platform Ecosystem Innovation Strategy and Dynamic Capabilities

Within platform ecosystems, a company's innovation strategy is considered a key element for securing sustainable competitive advantage. The core driver of corporate performance lies in the business model (BM), defined as an interconnected system of activities through which a company creates, delivers, and captures value[4]. That is, BM is emphasized as a new analytical unit that holistically describes a company's strategy and operations, extending beyond mere products or technologies. From this perspective, BM Innovation (BMI) is recognized as an essential strategy for companies to secure competitive advantage and achieve long-term performance. Meanwhile, research demonstrates that companies successfully implementing BMI have shown outstanding market performance, and that BMI plays a crucial role in complementing or replacing technological, product, and service innovation[5]. Furthermore, AI technology is seen as a catalyst that brings radical changes to existing products and services, maximizes operational efficiency, and transforms traditional business models into digital platforms[1]. Thus, prior studies emphasize the decisive impact of BM innovation on a company's sustainable performance and productivity, indirectly proving that the productivity gap among generative AI companies addressed in this study is deeply related to BM innovation.

In this business environment, Teece introduced the concept of dynamic capabilities to explain a firm's long-term success. Dynamic Capabilities refer to a firm's ability to integrate, build, and reconfigure internal and external capabilities to respond to rapidly changing environments. Teece categorizes these dynamic capabilities into three core elements. First, Sensing is the ability to identify and evaluate market opportunities and threats, requiring deep awareness of market trends, technological advancements, and shifts in customer preferences. Second, Seizing is the ability to mobilize resources and take appropriate actions—such as investments, partnerships, or strategic moves—to respond to sensed opportunities. Third, Reconfiguring is the ability to continuously strengthen, combine, and, when necessary, reconfigure a firm's asset structure to maintain competitiveness[3]. This can be supported through decentralization, autonomy, and strategic alliances.

2-2 Previous Research on Corporate Performance Evaluation Indicators

Corporate performance evaluation is crucial for measuring a company's efficiency, and Revenue Per Employee (RPE) is one of the most widely used indicators. RPE is a useful metric for assessing a company's return on labor input and measuring employee efficiency, particularly valuable in service and technology industries where labor costs constitute a significant portion of corporate expenses[6]. Meanwhile, research indicates that this labor productivity is a key indicator for evaluating corporate performance, and that higher labor productivity increases the likelihood of a company reinvesting in innovation and securing a competitive advantage[7]. However, limitations of RPE include the fact that its interpretation can vary depending on industry characteristics or a company's growth stage. For example, in traditional manufacturing, RPE can be heavily influenced by capital expenditures, whereas companies with business models that are less labor-intensive and easier to scale, such as SaaS, tend to exhibit high RPE[8].

Meanwhile, traditional methodologies have limitations in assessing the growth and valuation of technology startups. The price-to-sales ratio (PSR) is considered particularly useful for valuing emerging companies like startups, which possess high growth potential but have yet to generate significant profits[9]. Tech startups, in particular, often exhibit high PSRs driven more by growth potential than current profitability. Companies in high-growth sectors, especially tech firms, tend to command higher PSRs due to strong future growth expectations, with investors willing to pay a premium on current earnings anticipating substantial future growth. This explains why PSR serves as a useful valuation tool for companies that are not yet profitable, unlike the price-to-earnings ratio (PER)[10]. Thus, corporate valuation metrics must be appropriately selected based on a company's growth stage and business model. For technology startups, including AI companies, a comprehensive evaluation considering both technological capabilities and market scalability is particularly necessary.

3-1 Selection of Research Subjects and Data Collection Methods

This study selected five representative companies leading the generative AI industry as research subjects. Model-centric startups include OpenAI, Perplexity AI, and Anthropic, while large-scale platform companies include Google and Microsoft. This classification clearly distinguishes between companies that directly sell generative AI technology as their core product and those that utilize it as a tool to enhance the competitiveness of their existing platforms.

Data for this study was gathered from publicly available sources such as company websites, media reports, and news articles to conduct in-depth case studies on each company's business model, organizational structure, and strategic actions. Specifically, financial information and workforce data were collected from credible financial information platforms including PitchBook and U.S. Securities and Exchange Commission (SEC) filings for the period from 2023 to 2025.

3-2 Dynamic Capabilities-Based Analytical Framework

This study analyzes the innovation strategies of companies based on Teece's Dynamic Capabilities framework discussed in Chapter 2. This framework explores corporate cases centered on three key elements constituting corporate agility. First, regarding Sensing, it analyzes how each company perceived opportunities and threats in the generative AI market and what efforts they made to capture market changes. Second, regarding Seizing, it examines the business models and strategic investments each company implemented to capture the sensed opportunities. Third, regarding Reconfiguring, it analyzes how companies restructured their organizational structures, workforce composition, and technological infrastructure to execute new strategies. Specifically, in the qualitative analysis process, the three components of dynamic capabilities (Sensing, Seizing, and Reconfiguring) were designated as the core codes for the Qualitative Coding Procedure. The collected textual data was analyzed through an iterative review to extract key patterns and themes concerning each company's strategic actions, thereby allowing for the investigation of the specific manifestations of each dynamic capability component. Through this case analysis, we derive the characteristics of each company's innovation strategy and interpret how these strategies translated into performance outcomes to draw final implications. Such case analysis applying the dynamic capability framework contributes to logically arguing corporate performance as the result of strategic choices, not mere phenomena.

Table 1.

Analytical framework

4-1 OpenAI Innovation Strategy (Founded in 2015)

OpenAI's innovation strategy originates from its founding mission: developing artificial general intelligence (AGI) that benefits all of humanity. From the outset, they recognized AGI's potential reality and prioritized AGI safety as their foremost task. OpenAI's sensing capabilities extend beyond merely identifying market opportunities, focusing instead on deeply recognizing AI technology's potential risks (misuse, hallucination, etc.). For example, OpenAI conducted research to detect and reduce dangerous AI behavior modes like scheming (calculated and deceptive actions), thereby establishing the safety frameworks necessary for AGI development. Furthermore, they recognized early on that AGI development would demand enormous capital and computational resources, acknowledging the limitations of the existing non-profit research institute framework. This was a key Sensing outcome that triggered the shift to a capped-profit structure in 2019. Second, OpenAI deployed an unconventional Seizing strategy to capture the detected opportunity. By transitioning from non-profit to capped-profit, it established a unique business model capable of attracting massive capital. This model structures funding around the long-term mission of AGI development rather than limiting investor returns. Furthermore, it resolved the computing resource challenge through a strategic partnership with Microsoft. Microsoft invested over $ 13 billion in OpenAI, becoming its exclusive cloud partner, enabling OpenAI to secure the massive infrastructure needed for model development. It then generated revenue through an Application Programming Interface (API) -first business model, selling the models directly. Generative Pre-trained Transformer (GPT) models are priced on a per token basis, while image generation models like DALL-E are priced per image. Additionally, it launched the $ 20/month ChatGPT Plus subscription model and an Enterprise Plan for businesses, maximizing revenue across diverse market segments. Third, OpenAI continuously reconfigured its organization and assets to achieve its goal of developing AGI. Transitioning from a non-profit to a for-profit subsidiary enabled it to offer stock compensation to technical talent, which proved crucial in attracting talent from competitors like Google and DeepMind. Structurally, it maintains a Lean and Research and Development (R&D) -centric culture where Research and Engineering collaborate closely. This organization prioritizes the ability to think from scratch and rapid market learning. Furthermore, on the infrastructure front, OpenAI is moving beyond its exclusive partnership with Microsoft. It is collaborating with Oracle, SoftBank, and others to pursue the Stargate project, a $ 500 billion data center initiative, aiming to secure its own computing capabilities. This is part of a strategic reconfiguration to reduce dependence on specific partners and accelerate AGI development. Consequently, OpenAI's innovation strategy has generated global ripple effects. ChatGPT's success has fundamentally changed public perception of generative AI and served as a decisive catalyst for corporate AI adoption. Furthermore, the launch of text-to-video models like Sora has revolutionized the marketing and advertising industries, offering small businesses and startups opportunities to produce high-quality content at low cost. Technologically, the latest models like GPT-5 have significantly enhanced coding and agent task capabilities, demonstrating AI's evolution beyond simple chatbots into tools capable of performing complex tasks. This innovation signifies OpenAI's success in both paving the way toward AGI and democratizing AI access for hundreds of millions of users worldwide.

4-2 Perplexity AI Innovation Strategy (Founded in 2022)

Perplexity AI first demonstrated exceptional capability in simultaneously identifying the shortcomings of the existing search market and the opportunities presented by generative AI. It recognized the inefficiency of traditional search engines listing countless links, forcing users to sift through them themselves, and keenly sensed the potential demand for an answer engine where AI comprehensively summarizes information while clearly citing sources. It also rapidly recognized the market reality that AI models themselves were already becoming commoditized. This led to the strategic insight that the key to competitive advantage lay not in improving model performance, but in effectively leveraging models to deliver a unique user experience. Based on this market understanding, Perplexity AI adopted a Sensing strategy, focusing on proprietary search technology and AI agent capabilities rather than entering the race to develop its own model. Second, Perplexity AI deployed an innovative Seizing strategy to capture the identified market opportunity. It adopted the wrapper business model, which generates unique value by leveraging AI models from competitors (e.g., OpenAI, Anthropic, Google) and integrating proprietary AI search engine technology, thereby significantly reducing the burden of massive model development costs. This reduced the burden of high-cost model development, allowing resources to be focused on service improvements and enhancing the user experience. Furthermore, in August 2025, it launched a new subscription model called Comet Plus, introducing a revenue-sharing program with publishers. This model allocates 80% of Perplexity's subscription revenue to content providers and also compensates them for AI-generated agent traffic. This innovative strategy is seen as alleviating media companies' copyright infringement concerns and encouraging ecosystem participation. Furthermore, in August of the same year, Perplexity AI made a bold proposal to acquire Alphabet's Chrome web browser for $ 34.5 billion. This aggressive market preemption strategy aimed to secure a user base of over 3 billion people, dramatically strengthening its position in the search market. Third, Perplexity AI reconfigured its organization and infrastructure through an asset-light strategy. Maintaining a small workforce of around 250 employees as of 2025, it operates an agile decision-making system to respond swiftly to rapidly changing market conditions. Regarding infrastructure, it prioritizes dynamically utilizing cloud partners' computing resources as needed over operating its own data centers or directly training large-scale AI models. This Reconfiguring strategy significantly reduces fixed costs and allows resources to be concentrated on core competencies: search technology and AI agent capabilities. Consequently, Perplexity AI's innovation strategy has successfully established a unique position in the AI search market. Their service has positioned itself as an interactive answer engine that transcends the limitations of traditional search engines, and is expanding into the business to business (B2B) market through Perplexity Labs, a high-value-added feature supporting strategic corporate decision-making. Furthermore, the launch of the Comet browser presents a new user experience by combining a web browser with AI, demonstrating a strong commitment to preempting the AI agent platform market, which is poised to become a key battleground in the future AI market. Its revenue-sharing model with publishers is regarded as an effective alternative that mitigates copyright controversies and seeks sustainability within the AI ecosystem.

4-3 Anthropic Innovation Strategy (Founded in 2021)

Anthropic focused first on early sensing of potential risks that could arise alongside the rapid advancement of generative AI technology. Based on a strong conviction that unresolved issues of AI safety and alignment could pose a serious threat to humanity, they recognized this field as both a core business opportunity and a mission. This sensing positioned Anthropic as a safety-first company, becoming the starting point for a unique innovation strategy that differentiated it from competitors. Furthermore, they recognized that in the B2B market, enterprise customers prioritize security and reliability above all else when adopting AI, and developed a strategy to address this unmet demand. Second, Anthropic employed a unique Seizing strategy called business-first to capture the identified opportunity. Through an API-centric business model, they generated 70~75% of revenue from enterprise customers, focusing on the high-value B2B market rather than competing in the consumer market. They generated substantial token-based revenue through high-value enterprise contracts tailored to specific industries like finance, healthcare, and law. Analysis suggests Anthropic's average revenue per API request is up to 10 times higher than ChatGPT's. Additionally, through a channel partnership strategy providing models to cloud platforms like Amazon's Amazon Web Services (AWS) Bedrock and Google's Vertex AI, it minimized sales and marketing costs while gaining easy access to large corporate clients. Furthermore, by collaborating with highly secure government agencies like the U.S. National Nuclear Security Administration (NNSA) to develop AI safety tools preventing the misuse of nuclear-related data, it successfully implemented safety as a core feature of its actual products. This demonstrates how Anthropic's safety-first strategy transcends mere philosophy, translating into tangible business value. Third, Anthropic reconfigured its organization and resources to maximize both safety and efficiency. Structurally, researchers and engineers share a single title, member of technical staff, creating a lean organization that maximizes synergy between R&D and engineering. This concentrates all personnel on the singular goal of safety research, enabling rapid and effective decision-making. On the infrastructure front, similar to OpenAI, it adopted an asset-light model relying on cloud partners rather than building proprietary infrastructure. This strategy reduces the burden of massive capital investment, allowing focus solely on core competencies: AI safety research and model development. Consequently, Anthropic's innovation strategy has successfully established unrivaled leadership in the enterprise AI market. By mid-2025, Anthropic captured a 32% share of the enterprise large language model (LLM) market, overtaking OpenAI—which previously held a 50% share—to take the lead. Notably, the engineering team reported a 30% improvement in code delivery speed through Claude Code. It demonstrated exceptional performance in automating high-value workflows like document analysis and research report generation, dramatically enhancing enterprise customer productivity. This proves the effectiveness of Anthropic's safety and B2B specialization strategies in the market.

4-4 Google Innovation Strategy (Founded in 1998)

Google first sensed that generative AI technology posed a significant threat to its core search advertising business model while also presenting a new growth opportunity. Particularly after OpenAI's ChatGPT emerged, Google detected a shift in user search behavior from the traditional link-based approach to conversational responses. This was perceived not merely as a change in the competitive landscape of the search market, but as a fundamental threat to the business model that could lead to a decline in search advertising revenue. To counter this threat, Google realized it must go beyond its existing 'adding AI' stance and revise its strategy toward integrating and embedding AI technology company-wide. Google also recognized AI as a powerful tool capable of resolving inefficiencies in existing businesses and creating new value. Second, to seize the perceived threats and opportunities, Google executed a powerful strategy characteristic of large platform companies. It consolidated all its AI capabilities into Google DeepMind to develop the Gemini model, then aggressively integrated it into all core products like Google Search, Chrome, and Workspace. This strategy aims for Gemini to evolve beyond a simple chatbot into an interactive AI platform by combining with Google's vast ecosystem. Furthermore, Google built and advanced its own computing infrastructure, essential for AI model development. Google trains large-scale models using self-developed AI accelerators like Tensor Processing Unit (TPU) v4 Pods, reducing reliance on external infrastructure and enabling the internalization of AI technology. Furthermore, it is targeting the AI market by launching specialized solutions like Gemini in Google SecOps, which leverages AI technology to enhance the productivity of existing B2B customers. Third, Google undertook a company-wide reconfiguration to transform into an AI-centric organization. First, it restructured its workforce, striving to flatten organizational hierarchies by having fewer managers oversee larger teams. This strategy aims to accelerate decision-making and enhance organizational agility. Furthermore, it centralized AI capabilities—such as integrating the Gemini app team into Google DeepMind—to accelerate the research-development-launch feedback loop. Internally, Google is implementing intensive change management to embed AI technology into workflows, including mandating AI tool usage for employees. This Reconfiguring demonstrates Google's effort to shift beyond merely adding AI to products, aiming instead to transform the organization's operational approach itself into an AI-first model. Consequently, Google's innovation strategy has yielded rapid results in building an AI ecosystem centered around Gemini. Gemini overcame initial slow adoption to rapidly expand its user base and is strengthening market competitiveness by offering diverse enterprise AI solutions through Google Cloud. Furthermore, examples like integrating AI into existing services—such as Gemini in Google SecOps—are creating tangible value by enhancing enterprise customers' security and operational efficiency. These efforts demonstrate that Google's AI technology is not confined to lab achievements but is translating into real business value.

4-5 Microsoft Innovation Strategy (Founded in 1975)

Microsoft was the first to sense the potential of AI technology and recognized it as an opportunity to strengthen its core businesses: cloud and productivity software. In particular, it recognized the technical potential of OpenAI's GPT model early on and sensed that they would require massive computing resources to pioneer the path toward AGI. Based on this sensing, Microsoft realized that, beyond developing AI technology internally, partnering with the most innovative AI startups was the most effective strategy to capture the market. This became the decisive motivation for pursuing a partnership with OpenAI. Second, Microsoft executed a bold seizing strategy, aiming to become a core pillar of the AI ecosystem, based on the insights gained through sensing. It made a strategic $ 13 billion investment in OpenAI and established a partnership to provide its Azure cloud platform as OpenAI's exclusive computing infrastructure. This partnership created a mutually beneficial structure: Microsoft gained priority access to OpenAI's technology, while OpenAI secured vast computing resources. Furthermore, through this partnership, Microsoft integrated OpenAI's GPT models into its Microsoft 365 Copilot and launched GitHub Copilot, dramatically enhancing the value of its existing productivity software and development tools using AI technology. This Copilot business model generates revenue through a per-seat monthly subscription, successfully increasing ARPU (Average Revenue Per User) for enterprise customers. Third, Microsoft reconfigured its organization and infrastructure to execute its AI-centric innovation strategy. First, it restructured its engineering organization to embed AI technology into existing businesses and announced a workforce restructuring affecting approximately 15,000 employees to strengthen AI capabilities. This strategy aims to make AI an essential competency for all roles and focus resources on new AI-related roles rather than traditional ones. Additionally, through programs like the AI Cloud Partner Program, Microsoft supports partners in developing and launching solutions using Azure AI, thereby expanding the AI ecosystem. On the infrastructure front, it is investing a massive $ 80 billion annually to expand data centers and develop proprietary AI semiconductors like Azure Maia AI Accelerators, aiming to secure the infrastructure leadership essential for the AI era. Consequently, Microsoft's innovation strategy has driven explosive growth in its Azure cloud business and secured leadership in the AI platform market. In April-June 2025, Azure cloud service revenue grew 34% year-over-year, with AI-related services driving this growth. Copilot has secured over 100 million users, dramatically enhancing enterprise customer productivity, and is driving results in existing businesses, such as boosting Microsoft's search advertising revenue by 21%. These achievements prove that AI is not merely a technology but a core business model that drives corporate revenue and creates new value.

4-6 Comprehensive Comparison of Innovation Strategies and Performance by Company Type

Model-centric startups' innovation strategies are broadly focused on creating hyper-scalable IP and capturing markets first. First, in Sensing, they identify unmet needs in existing markets and detect opportunities within the long-term technological vision of AGI development and AI safety. Second, in Seizing, they sell high-margin IP through an API-first business model and reduce infrastructure burden via an asset-light strategy. Third, in Reconfiguring, they maximize decision-making agility through small elite organizations that blur the boundaries between research and engineering. Consequently, these innovation strategies lead to overwhelming revenue growth rates and high PSR, as shown in Table 2 which is based on PitchBook data, resulting in the market highly valuing their growth potential and worth.

Table 2.

Comparison of innovation strategies and outcomes among generative AI companies

The innovation strategies of large platform companies are broadly focused on reorganizing existing platforms around AI and strengthening market dominance. First, in Sensing, they recognized generative AI technology as both a threat to existing businesses and a new growth opportunity. Second, in Seizing, they leveraged massive capital and infrastructure to aggressively invest in internalizing AI technology and preempting the market through strategic partnerships.

Third, in Reconfiguring, they reorganized their structures around AI, pursued intensive change management including workforce redeployment and mandatory AI adoption. Consequently, these strategies drove existing business performance and created new revenue streams, leading to stable sales growth and profitability.