OpenAI is experiencing a wave of high-profile executive departures at a critical moment, with the company preparing for an initial public offering (IPO) while seeking to validate an $852 billion valuation. The exits come as the broader AI market undergoes a seismic shift: the cost of advanced AI models is approaching zero, fundamentally changing what companies should actually be paying for in the AI era.

Why Are OpenAI Executives Leaving Before the IPO?

The departures have raised alarm bells among industry watchers. Chief revenue officer Denise Dresser resigned after less than a year at the company, just two days after longtime COO Brad Lightcap announced his own exit after eight years. Before that, Fidji Simo, the former Instacart CEO, stepped away from her high-ranking executive role in July to focus on managing a chronic illness. Four other senior leaders, including former VP of Science Kevin Weil and CMO Kate Rouch, left the company in April.

The timing is particularly striking because executives typically wait for liquidity events like IPOs to realize the value of their equity stakes.

"The executives leaving OpenAI ahead of their IPO is a huge red flag," noted Kevin McCormick, founder of AI startup SignAudit.AI, pointing out that walking away before a liquidity event suggests underlying tensions.

Kevin McCormick, Founder at SignAudit.AI

The departures have prompted CEO Sam Altman and president Greg Brockman to shift into what analysts are calling "founder mode," with Brockman meeting with more customers and teams across the company than ever before to contain disruptions.

What's Driving the Shift in AI Pricing and Value?

Meanwhile, the AI market itself is undergoing a fundamental transformation that may be contributing to internal pressures at OpenAI. The cost of AI models is heading toward zero, with open-source alternatives becoming increasingly competitive. Alibaba's open-source model "Qwen" has surpassed 3 billion downloads, overtaking Meta's Llama to become the most downloaded open-source large language model (LLM) in the world. The latest Qwen3 offers eight sizes ranging from 0.6 billion to 235 billion parameters, and the 27 billion parameter model can run on a notebook PC with 16 gigabytes of memory, with zero cost and an Apache 2.0 license allowing free commercial use.

Just two years ago, when ChatGPT first launched, the API usage fee for GPT-4 was $0.03 per 1,000 tokens for input and $0.06 for output. Customization and fine-tuning often required initial costs of 3 to 5 million yen (roughly $20,000 to $35,000). Today, models with performance comparable to GPT-4 can be used for $0.1 to $0.5 per 1 million tokens through services like OpenRouter, which is less than a tenth of the cost from two years ago. If run locally, the cost is literally zero.

Where Is the Real Value in AI Now?

The infrastructure for delivering and utilizing models, rather than the models themselves, has become the battleground. Payment giant Stripe is negotiating to acquire AI gateway company OpenRouter for about $7 billion (approximately 1 trillion yen). OpenRouter provides a routing service that allows users to switch between multiple AI models via a single API. The fact that Stripe is willing to pay $7 billion for a service that connects models, rather than for the models themselves, reveals where the market sees future value.

For small and medium-sized enterprises (SMEs), this shift creates both opportunity and confusion. With model costs approaching zero, the question is no longer "Can we afford AI?" but rather "What should we actually be paying for?" According to analysis of the market, SMEs should focus their spending on three key areas:

• Business Process Identification: Inventorying operations and identifying tasks that can be replaced by AI, such as automatically generating estimate drafts that reduce time from 30 minutes to 3 minutes per item, or using retrieval-augmented generation (RAG) to handle 70 percent of 100 monthly customer inquiries automatically.
• Systematization and Prompt Engineering: Transforming employee know-how into reusable prompts and system instructions so that new employees can produce outputs of the same quality without relying on individual expertise, which requires investment in human time rather than expensive software licenses.
• Data Organization and Management: Building vector databases and retrieval-augmented generation pipelines using free tools like ChromaDB, FAISS, LangChain, or LlamaIndex to leverage company-specific data as a competitive moat, since a company's 20 years of customer interaction history cannot be replicated by competitors.

An AI utilization consultant to help design and maintain these systems costs 50,000 to 100,000 yen per month (roughly $350 to $700), or 600,000 to 1.2 million yen annually. This is often cheaper than full company adoption of ChatGPT at 20,000 yen per month per user, which for ten employees totals 2.4 million yen per year.

How to Transition Your Organization to Free or Low-Cost AI Models

For organizations looking to shift away from expensive proprietary models, the path forward involves several concrete steps:

• Audit Current Tasks: Document which tasks currently consume the most time and could benefit from AI assistance, such as report summarization, inquiry response, or estimate generation.
• Select and Test Open Models: Experiment with free models like Qwen3, Llama 4, or Gemini 2.5 Flash to identify which performs best for your specific use cases before committing budget.
• Build Internal Infrastructure: Set up local model deployment using tools like Ollama, or use low-cost API routing services to avoid vendor lock-in and reduce ongoing expenses.
• Document and Systematize Knowledge: Interview experienced employees to capture decision-making criteria and incorporate that knowledge into system prompts that new team members can use.
• Organize Company Data: Convert past documents to text or PDF format and build a vector database to enable retrieval-augmented generation, turning your company's historical data into a competitive advantage.

OpenAI's executive departures underscore a broader industry reckoning. As the cost of models themselves approaches zero, the companies that will thrive are those that can deliver and integrate those models effectively into business operations. For OpenAI, the challenge is clear: prove that the infrastructure and ecosystem around its models justify the company's valuation, even as the models themselves become commoditized. For SMEs, the opportunity is equally clear: the era of expensive AI is over, and the real investment should be in understanding how to use free or cheap models to transform their own operations.