Tech Stocks Have Plummeted. So, What Should Your Company Do?

The market capitalization of major tech companies on NASDAQ has returned to levels seen before the AI boom. It’s estimated that several trillion dollars have vanished since the peak.

The media is buzzing about the “AI bubble burst.” But hold on a second.

The drop in stock prices and the unavailability of AI technology are two completely different matters.

In fact, it’s the opposite. With stock prices down, we have now entered a phase where only the “practical value” of AI remains. What’s left after speculative money has pulled out are genuinely usable tools and an environment where they can be acquired at a low cost.

For local SMEs, this could be the biggest opportunity in the past two years. Let me explain three specific reasons why.

—

1. The Cost of Using AI Has Dropped to Laughable Levels

First, let’s look at the numbers.

At the beginning of 2023, integrating a GPT-4 class model into business via API required a monthly budget of several hundred thousand yen. If you wanted to set up a GPU server in-house, an NVIDIA A100 cost over 2 million yen. Creating a decent inference environment would set you back between 5 to 10 million yen.

By June 2025, the situation has changed dramatically:

• MiniMax M2.7 (Open Source): A model comparable to GPT-4 has been released for free. No licensing fees. You can download it and run it on your own server.
• Intel Arc Pro B70: A local AI inference GPU with 32GB of VRAM. Priced at $949 (about 140,000 yen), which is less than a third of comparable NVIDIA products.
• Latest Ollama: The tool for running LLMs locally has evolved significantly. With just one command, you can download and run a model immediately. It’s usable even for those without engineering skills.

In other words:

What used to cost 3 to 5 million yen to “implement AI” two years ago can now be done for 150,000 to 300,000 yen.

Just buy one GPU, install an open-source model, and run it with Ollama. That’s all it takes to acquire a dedicated AI environment for your company. There’s no need to keep paying monthly fees to the cloud, nor to send data outside.

For local manufacturers or a ten-person professional office, this has become an affordable option. This is the answer to what happens when “costs go down.”

—

2. The Structure of Talent Has Changed—”AI Talent” is No Longer Necessary

During the bubble period, salaries for AI talent were outrageous. Even new graduates were earning 10 million yen, and experienced professionals over 20 million yen. It was impossible for SMEs to compete in the hiring race.

With the crash of tech stocks, this structure is beginning to collapse. Major tech companies are laying off employees, and talented engineers are entering the job market. However, let’s be honest: What SMEs should focus on now is not hiring “AI talent.”

Why? Because the tools have evolved too much.

Try out the latest version of Ollama. Just typing `ollama run` in the terminal gets a high-performance language model running. The GUI has also improved significantly. Programming knowledge is almost unnecessary.

Even with MiniMax M2.7, if you follow the instructions to download it from Hugging Face and run it locally, it takes less than 30 minutes.

In other words, the barrier to “using AI talent” has dramatically lowered.

What’s needed now is not a 20 million yen AI engineer, but an employee who understands the company’s operations and is comfortable using new tools. There’s likely one or two such people in every company.

In fact, at a local construction company we support, a site supervisor started using Ollama to automatically summarize daily reports. The setup took just two hours. No outsourcing costs. Twenty hours of administrative work were eliminated each month.

We’ve shifted from an era of “hiring AI talent” to one where internal staff “uses AI.” This change presents a structural opportunity for SMEs that can’t compete with large companies in hiring.

—

3. The “Preparation Time” is Indicated by Stock Price Charts

There are rules regarding the timing of technology adoption.

If you adopt during the peak of a bubble, it’s expensive. If you adopt after a crash, it’s cheap. However, the performance of the technology is better after the crash.

Consider this: Companies that decided to implement AI at the peak in 2024 ended up signing overpriced annual contracts for SaaS, paying high consulting fees, and receiving abstract reports under the guise of “AI strategy formulation.” Hundreds of thousands of yen vanished, and nothing changed on the ground. We’ve seen many such companies.

On the other hand, as of June 2025, the same things can be achieved at less than a tenth of the cost, and with more powerful models.

Specifically, what can be done?

• Automatic processing of estimates and invoices: Run MiniMax M2.7 locally to extract information from PDFs and input it into accounting software. No need to pay monthly fees to external services.
• First response to customer inquiries: Build a chatbot fed with your company’s FAQ data on your own server. Customer data doesn’t leave your premises.
• Automatic generation of minutes and reports: Transcribe audio data with Whisper and summarize it with LLM. Five hours of administrative work are eliminated each week.
• Inventory forecasting and demand analysis: Analyze past sales data with a local model. Expensive BI tools become unnecessary.

All of these can be achieved with just one GPU and an open-source model. The initial investment is 150,000 yen, and the running cost is just the electricity bill.

—

Now is the Time for SMEs While Large Corporations Hesitate

Another structural change that cannot be overlooked is this:

Large corporations are currently burdened with the “accountability” of AI investments. They are being questioned by shareholders about the ROI of AI projects costing billions of yen, making it harder to get new investment approvals. Decision-making is slowing down.

SMEs are different. If the president says, “Let’s do it,” they can act the next day. There’s no need for approval documents or board meetings. No company takes three months to approve the purchase of a 150,000 yen GPU.

While large corporations hesitate due to shareholder concerns, SMEs can finish implementation.

This “speed of decision-making” is the greatest weapon that SMEs possess. When the cost of technology drops, the first beneficiaries are the organizations that can move quickly.

—

So, What Should You Do?

Just three things:

1. First, create a local AI environment.
Buy one Intel Arc Pro B70 (about 140,000 yen), install Ollama, and run MiniMax M2.7. It will take half a day.

2. Choose the “most tedious routine task” in your company.
It can be anything—creating minutes, processing estimates, handling inquiries. Experiment with letting AI take over that task for a week.

3. Measure the results numerically.
“How many hours were reduced per month?” “How much did costs decrease?” If you have the numbers, you can make the next investment decision.

All it takes for the first step is 140,000 yen and half a day of time.

The crash of tech stocks may be bad news for investors. However, for SMEs that are steadily doing business in local areas, this is the best “preparation time” possible.

Technology has become cheaper. Tools have become easier to use. Large corporations are hesitating.

If you’re going to act, now is the time.