Conclusion

The more giants clash, the lower the AI costs for small and medium enterprises.

The resolution of the lawsuit between OpenAI and Musk, Anthropic’s acquisition of OpenAI-related startups, and the massive losses in the AI divisions of Microsoft and Google—these may seem like lofty matters. However, when we look at the structure, they represent extremely practical news for local small and medium enterprises.

As long as the giants continue their war of attrition, API prices will keep falling. This is the essence.

In 2022, the API usage fee for GPT-3.5 was about $0.002 per 1,000 tokens. As of 2024, GPT-4 Turbo offers similar processing for about $0.01, while its performance has increased severalfold. Anthropic’s Claude 3.5 Sonnet has also entered the same price range. Google’s Gemini 1.5 Flash is even aiming lower.

Performance is increasing while prices are decreasing. Is there any other market like this?

Whether you end your analysis with “Wow, that’s amazing” or delve deeper into “So, what can we use this for?”—that difference will determine next year’s profit margins.

—

What Happened—Reading Three News Items Through the Lens of ‘Cost Structure’

1. The OpenAI vs. Musk Lawsuit: Accelerating Monetization = Full-Scale Price Competition

A jury in Oakland, California, dismissed Musk’s claims. No breach of contract, statute of limitations has expired. OpenAI’s path as a for-profit entity has been legally confirmed.

What does this mean? OpenAI is steering full throttle towards ‘making money.’ With accountability to investors now clear, there will be pressure to lower prices and increase features to grow user numbers and revenue.

In fact, since the beginning of 2024, OpenAI has rapidly implemented strategies to “expand the base,” including the free release of GPT-4o, gradual reductions in API prices, and the introduction of the ChatGPT Team plan.

2. Anthropic’s Acquisition of Stainless: Lowering Developers’ ‘Switching Costs’

Anthropic’s acquisition of Stainless, a startup with a tool that automatically generates SDKs (Software Development Kits) for APIs, is significant. Moreover, this tool was also used to generate SDKs for OpenAI’s API.

In other words, Anthropic has acquired the infrastructure that makes it easier for OpenAI developers to “switch to Claude.”

This may seem subtle but is disruptive. When switching costs for developers decrease, API providers will have to compete solely on “price and performance.” As a result, price competition will accelerate even further.

3. Giants’ AI Losses: The War of Attrition Continues

Microsoft has invested over $13 billion in OpenAI. Google is pouring several billion dollars annually into DeepMind and Gemini development. Amazon is investing up to $4 billion in Anthropic. Meta has even open-sourced Llama 3 for free.

Everyone is digging themselves deeper into the red.

Why? Because if they gain dominance in the AI platform, they can reshape the revenue structures of cloud, search, advertising, and e-commerce. Therefore, no one can back down.

As long as this war of attrition continues, API prices will decrease, free models will proliferate, and development tools will continue to improve. For small and medium enterprises, it’s as if the giants are shouldering the ‘R&D costs’ on their own.

—

So, What Should Small and Medium Enterprises Focus On?—Three Specific Areas

Many articles end with “This is an opportunity.” Here, we will delve into “What to do starting tomorrow.”

Opportunity 1: Automating Customer Inquiries—A ’24/7 Response Desk’ for 50,000 Yen a Month

For local small and medium enterprises, the most immediate benefit is automating customer interactions.

Specifically, building an AI chatbot based on the company’s FAQ, manuals, and past inquiry history. By using the APIs of Claude 3.5 Sonnet or GPT-4o mini, the monthly API usage fee can be kept between 3,000 to 15,000 yen (for processing 1,000 to 5,000 inquiries per month).

Even when including the costs for building the chat UI, the initial cost would be between 100,000 to 300,000 yen, with ongoing monthly costs of 30,000 to 50,000 yen.

Traditionally, outsourcing a similar system would have cost around 3 million yen initially and 200,000 yen monthly. Costs have dropped to less than one-tenth.

You can achieve ’24/7 support’ with zero labor costs. Even a company with five employees can have a customer support system on par with large corporations. This is the structure of reversal.

Opportunity 2: Searching Internal Knowledge—A Weapon to Break Down Silos

“Only Mr. XX knows about that matter”—this is the biggest ailment of small and medium enterprises.

By inputting internal meeting minutes, manuals, daily reports, and emails into a vector database, and enabling natural language search through RAG (Retrieval-Augmented Generation), the technical barriers have significantly lowered. Vector databases like Pinecone, Qdrant, and Chroma offer free plans, and when combined with APIs, an internal search system can operate for 10,000 to 30,000 yen a month.

If you ask, “What estimate did we provide for the XX project last year?” the AI generates answers from past data. Information that was only in the minds of veteran employees becomes accessible to everyone.

Eliminating silos is a survival strategy for small and medium enterprises. We have entered an era where this can be done for just a few tens of thousands of yen a month.

Opportunity 3: Automating Sales Materials and Proposals—Turning ‘Time to Create’ into ‘Time to Think’

Many sales representatives spend three hours creating a single proposal. By feeding customer information and past proposal templates to AI, draft generation can take just 10 minutes.

The remaining 2 hours and 50 minutes can be spent thinking about “What proposal would truly resonate with this customer?”

API costs are just a few yen per proposal. Even if you create 100 proposals a month, it will only cost a few thousand yen.

—

To Wait or To Act Now

Many business owners think, “I’ll wait until it gets a bit cheaper” or “I’ll wait until the technology matures a bit more.”

I understand the sentiment. However, consider this:

API prices will indeed likely decrease next year. But is there any point in waiting for something that can already be started for just a few tens of thousands of yen to become a few thousand yen cheaper?

The real cost of utilizing AI is not the API usage fee. It is the time spent experimenting with ‘how to incorporate it into your business.’ This learning cost can be recouped more quickly by companies that start early.

The giants’ war of attrition will continue for at least the next 2 to 3 years. During that time, APIs will keep getting cheaper, and tools will continue to proliferate. However, the talent that can understand your business and determine where to integrate AI will not just appear out of nowhere.

Which company will survive in 2027: the one that started experimenting for 50,000 yen a month now, or the one that waits two years to say, ‘Let’s get started’?

The answer is clear.

—

Future Key Indicators—Just Keep an Eye on These Three

1. API Unit Price for GPT-4 Class (in $/1M tokens): Currently around $2.50. If it drops below $1, applications will explode.
2. Performance of Open Source Models: Meta’s Llama 3, Mistral, Google’s Gemma. The day when free models catch up to GPT-4 is near. The moment they do, the API billing model itself could collapse.
3. Maturity of No-Code/Low-Code AI Tools: Tools like Dify, Flowise, and n8n are reaching practical levels where AI workflows can be built without engineers. This is crucial for small and medium enterprises.

—

Summary: The Giants’ War is a Tailwind for Small and Medium Enterprises

The lawsuit between OpenAI and Musk, Anthropic’s acquisition of Stainless, and the giants’ ongoing losses—

All of these are acting as mechanisms that structurally reduce the costs of AI utilization.

Small and medium enterprises can leverage the technologies that large companies have burned through billions to create for just a few tens of thousands of yen in API fees. This is an unprecedented situation in historical terms.

The fruits to be picked are right in front of you. The only question is whether you will bend down to pick them up.

First, choose something in your business that you think, “What would happen if I left this to AI?” Experiment for 50,000 yen a month. If it doesn’t work out, you can stop next month.

Now is the best time for small and medium enterprises to stock up while the giants are clashing.