Major Companies Withdraw from AI Investment. So, What Should Small and Medium Enterprises Do?

To get straight to the point: Now is the most reasonable time for small and medium enterprises to adopt AI.

In August, spending on AI by major companies slowed down. According to a report by TechCrunch, AI spending per employee has clearly decreased. Furthermore, OpenAI has temporarily halted new subscriptions for its GPT-6 Pro plan ($200 per month) due to demand exceeding infrastructure supply.

Some media outlets are speculating whether this indicates an ‘AI bubble burst.’ However, that’s not the essence of the matter.

The reasons why major companies are pulling back and why small and medium enterprises should act now are actually two sides of the same coin.

—

Why Did Major Companies Cut Spending?

The reason is simple: “They paid a lot for high-end models but are not utilizing them effectively on the ground.”

Much of the AI investment by major companies has this structure:

• Company-wide contracts for top-tier models like GPT-4
• Annual API costs ranging from tens of millions to hundreds of millions of yen
• Yet, only a few departments are actually using it in daily operations
• When ROI is questioned, reports indicate “no visible effects”

In other words, they are finally beginning to realize that “buying high-performance AI” and “achieving results with AI” are entirely different matters.

What major companies have done is akin to buying the ‘highest grade’ of AI for everyone without considering the cost-effectiveness. Naturally, this doesn’t add up.

—

AI Costs Are Plummeting to ‘Collapse’ Levels

This is the most crucial point.

While major companies were investing in high-cost models, the cost structure of AI has dramatically changed. Let’s look at some specific numbers:

• API cost of GPT-4: Approximately $30 per 1M tokens at its launch in March 2023 → In 2024, GPT-4o mini costs $0.15. About 200 times cheaper.
• Open-source models: High-performance models like Llama 3.1, Mistral, and Gemma are now available for free for commercial use. If run on in-house servers, API costs can be zero.
• Cost of implementing AI chatbots: Two years ago, outsourcing could cost around 3 million yen, but now, there are cases where it can be built for under 50,000 yen using no-code tools.

What does this ‘cost collapse’ mean?

It signifies that the premise that ‘companies with financial power will win’ in AI adoption has collapsed.

Small and medium enterprises can now acquire functionalities equivalent to AI systems that major companies invested 100 million yen in for just a few tens of thousands of yen per month. This reversal is happening right now.

—

The Temporary Halt of OpenAI’s Pro Plan Indicates ‘Another Truth’

The temporary halt of OpenAI’s GPT-6 Pro plan ($200 per month) also needs to be interpreted calmly.

This is not a story of “AI has failed.” It indicates that demand surged and infrastructure couldn’t keep up. In other words, there is certainly demand for top-tier models. However, the costs and technology to supply them stably have not yet caught up.

The lesson small and medium enterprises should learn from this is clear.

“Using the top-tier model” is not always the right answer.

In fact, 80% of business tasks can be sufficiently handled without the top-tier model. Responding to inquiries, summarizing meeting minutes, drafting emails, organizing data—full specifications of GPT-6 are not necessary for these daily tasks. GPT-4o mini, Claude 3.5 Haiku, or open-source Llama 3.1 are more than adequate.

Here, the concept of “smart model routing” is noteworthy.

This is a system that automatically switches the AI model used based on the difficulty of the task. For simple questions, a cheaper model is used; for complex analyses, a high-performance model is employed. This alone can reduce API costs to one-fifth or one-tenth in some cases.

Major companies are too large to optimize these details easily. They implement a unified platform across the organization, forcing everyone to use the same model. That is ‘governance.’

On the other hand, small and medium enterprises are different. If the CEO says, “Let’s use this,” the entire company can start using it from tomorrow. This speed of decision-making is the greatest weapon of small and medium enterprises.

—

Three Things Small and Medium Enterprises Should Do Now

Enough with the abstract discussions. What should they specifically do?

1. Start with “under 5,000 yen per month”

ChatGPT Plus ($20 per month ≈ 3,000 yen), Claude Pro ($20 per month), Google Gemini Advanced ($2,900 per month). Just pick one. First, the business owner should use it daily. Without using it, they won’t see “what it can be used for.”

2. Replace just one “personalized task” with AI

Reports that only veteran employees can write, estimates that only specific individuals can create, know-how that exists only in the CEO’s mind. Choose one of these tasks that “cannot function without that person” and try to replicate it with AI.

Write the know-how into the prompt to create a state where anyone can produce the same quality output. This is what “systematization” means. Outsourcing would cost 500,000 yen, but doing it yourself is virtually free.

3. Try models that “major companies are not using”

Due to vendor contracts and security reviews, major companies often end up choosing only OpenAI or Microsoft. However, now there are plenty of options like Anthropic’s Claude, Google’s Gemini, Meta’s Llama, and Mistral’s Mistral.

Small and medium enterprises have the freedom to “try everything and choose the one that fits best.” This freedom is something major companies absolutely do not have.

—

Not a ‘Bubble Collapse’ but ‘Democratization of Costs’

Finally, how should we perceive this situation?

It’s easy to call the reduction in AI spending by major companies a ‘bubble collapse.’ However, looking at the structure, what is happening is “the costs of AI have dropped so much that high-cost investments have lost their meaning.”

This is not a bubble collapse. It is the democratization of costs.

Two years ago, a budget of several million to tens of millions of yen was necessary for serious AI adoption. Now, it can start from just a few thousand yen per month. The speed of this change is faster than during the internet boom.

And looking back in history, when cost democratization occurs, it is always the ‘have-nots’ who benefit the most.

When the cost of website creation dropped from 1 million yen to 10,000 yen, small and medium enterprises could finally stand on the same playing field as major companies. When the cost of managing social media became zero, individual stores could reach a national audience.

Now, the same thing is happening with AI.

While major companies are hesitating, saying “the ROI is unclear,” small and medium enterprises can change their operations with tools costing 3,000 yen per month. They don’t have to wait for a system development costing 3 million yen; they can start tomorrow.

The question is simple: “So, what will your company start doing tomorrow?”

Now is the time for small and medium enterprises to take action while major companies are hesitating.