How long do you think this sell-off will continue? Is this the “bursting of the AI bubble”? It’s probably too early to say.
I think PKSHA $3993 is a prime example of a company that makes money with AI. Their customers actually save money with PKSHA’s AI applications and, as a result, are spending more and more (high NRR). The flywheel is working perfectly at PKSHA. The company’s philosophy is also very human-centered. The goal is the co-evolution of humans and AI, rather than simply seeking to replace human labor. I think PKSHA still has a lot ahead of it. The numbers back this up as well.
I think PKSHA $3993 is a prime example of a company that makes money with AI. Their customers actually save money with PKSHA’s AI applications and, as a result, are spending more and more (high NRR). The flywheel is working perfectly at PKSHA. The company’s philosophy is also very human-centered. The goal is the co-evolution of humans and AI, rather than simply seeking to replace human labor. I think PKSHA still has a lot ahead of it. The numbers back this up as well.
•
11
•
@PikaPika0105 Hello, my friend, That’s a good question. Right now, several factors are coming together to form a toxic cocktail: rising inflation, interest rate hikes, and the complex web of financing. The last point, in particular, could be the one that causes a bubble to burst. I think CoreWeave has contributed significantly to the correction. The volume of debt financing is growing steadily. And an interest rate hike could make it impossible to repay loans. That’s why I think interest rates should be raised cautiously. Because no one can really gauge how many banks are involved here. And if even one bank is thrown into turmoil as a result, the whole thing could come crashing down. The most important thing should be to ensure that companies like Anthropic, OpenAI, etc., don’t run into trouble. Even at SoftBank, as many as 20 banks are reportedly involved in AI financing. And now Masayoshi Son wants to raise up to $100 billion from investors in the Gulf region.
••
@Tenbagger-Capital Yes, I also think the whole structure is very unstable. I’ve significantly reduced the AI exposure in my portfolio. I still hold SoftBank, but it’s now my smallest position at 3%. I didn’t want to sell everything because I still believe in many of the projects, but it’s a dance on a knife’s edge. My other positions are all performing independently of the AI boom, though many are sensitive to interest rates and general market turmoil. In the long run, though, I think I’m still on track. Hopefully, we’ll get a soft landing. CoreWeave, Nebius, IREN, etc., can all go under for all I care; others like NVIDIA will get their fair correction, and the market will return to reason. The hyperscalers will be able to weather such a blow thanks to their strong core businesses. Others won’t be able to. Perhaps then we’ll finally be ready to create real value with AI instead of empty promises.
••
@PikaPika0105 For me, AI is and will always be a revolution. Setbacks are bound to happen. And risks and regulation will be major issues. As long as Trump is at the helm, AI—or let’s call it “superintelligence”—will continue to be pushed forward anyway. Ultimately, it’s also about not losing the economic competition with China. As I’ve written many times before, AI will have the greatest impact in the defense sector. When it comes to large AI models, we in Europe may have already fallen behind. That’s why I consider it extremely important to leverage our expertise in mechanical engineering and the industrial sector. To become a leader in the field of industrial AI. Here, I also see Japan far ahead. Because I personally see enormous potential in this area.
••
@PikaPika0105 By the way, $3993 really caught my eye at first glance. Great chart pattern with a breakout above the 200-day moving average. Nice growth potential for the next few years. I’ll have to take a look at the margins and debt levels. This might be something for me. Thanks for the tip.
••
@Tenbagger-Capital It’s also a great company. I’ve been on board for a while now, and things are slowly starting to pick up. I had my latest analysis of AI summarized for you in bullet points:
* Macroeconomic leverage & labor market shifts:
* A decline in Japan’s working-age population of about 11 million people by 2040 will lead to an acute labor shortage in administrative white-collar roles (call centers, human resources, clerical work, administration).
* PKSHA doesn’t address the limited IT budgets of 1 to 2% of revenue, but instead directly targets the 30 to 50% of corporate costs spent on shrinking staffing capacities.
* Structural transition from traditional per-seat pricing to value- and transaction-based pricing models for autonomous software agents, ensuring that declining customer headcount does not diminish revenue potential.
* Technological foundation & independence from Big Tech:
* Avoiding capital-intensive frontier base models: PKSHA does not train general $100-billion models from scratch but instead uses open-source architectures and trains compact, domain-specific Small Language Models (SLMs) at its own expense.
* Dominance in the voicebot market with an 83.8% market share in Japan through proprietary speech-to-text engines trained on billions of minutes of real call center data, resolve phonetic Kanji ambiguities with latencies under 300 milliseconds, and are natively integrated into legacy telephone systems (PBX/CTI).
* Full integration into Japanese governance structures through the digital mapping of multi-level consensus and approval processes (Ringi) directly within Microsoft Teams and Slack.
* Full data sovereignty through “PKSHA Private AI Agents,” enabling regulated customers in the financial, insurance, and government sectors to run generative agents locally on their own servers without forwarding data to U.S. cloud infrastructures.
* Three-pillar operational flywheel:
* AI Research & Solution (operating margin around 25%): Major clients such as NTT Docomo and Nomura Real Estate pre-fund customized algorithm development; intellectual property (IP) rights remain entirely with PKSHA.
* AI SaaS (gross margin over 70%): Standardization of proven algorithms into recurring software modules with an ARR of over 11.1 billion JPY (+32% YoY), over 4,900 corporate customers, a net retention rate of 107.7%, and a value-up rate of 141% for upgrades to agent solutions.
* AI Powered Worker (APW): The acquisition of Circulation provides access to more than 4,000 small and medium-sized business customers without significant customer acquisition costs (CAC); consultants act as on-site implementation partners and transform traditional consulting and staffing services into high-margin AI-powered BPO.
* Accounting & Valuation Baseline:
* IFRS accounting avoids scheduled goodwill amortization from acquisitions that affects net income, allowing operational growth to flow directly into earnings per share (EPS).
* The market capitalization of approximately 100 billion JPY (~600 million €) does not reflect the underlying software profitability, as the market mistakenly classifies the company as a labor-intensive IT service provider.
* Ongoing disciplined capital return through share buybacks, as well as the first-ever introduction of a regular cash dividend of 10 JPY per share for fiscal year 2026.
* Detailed Forecast – Fiscal Year 2026e (Actual Run Rate):
* Consolidated revenue: approximately 36.5 billion JPY (~228 million €)
* Adjusted operating profit (EBIT): approximately 5.7 billion JPY (margin of 15.6%)
* Net income after taxes: approximately 3.2 billion JPY (~25 million € at 128 JPY/EUR)
* SaaS customer base: approximately 5,100 companies with an average annual revenue per user (ARPU) of 2.30 million JPY
* Base dividend: 10 JPY per share
* Detailed Forecast – Fiscal Year 2030e:
* Consolidated revenue: 73.8 to 82.0 billion JPY (~575 to 640 million € at 128 JPY/EUR)
* Segment revenues: AI SaaS 33.3 billion JPY, AI Research & Solution 21.0 billion JPY, AI Powered Worker 19.5 billion JPY
* SaaS customer base: approximately 7,400 companies with an ARPU of 4.50 million JPY per year
* Adjusted operating profit (EBIT): approximately 16.5 billion JPY (margin of 22.4%)
* Net income after taxes: 11.4 to 13.0 billion JPY (~89 to 102 million € at 128 JPY/EUR)
* Expected share price: approximately 10,466 JPY
* Dividend per share: approximately 75 JPY (payout ratio of 20%)
* Detailed Forecast – Fiscal Year 2035e (Base Case Bottom-Up):
* Consolidated revenue: approximately 157.3 billion JPY (~1.23 billion € at 128 JPY/EUR)
* Segment revenue: AI SaaS 89.3 billion JPY (56.8% of total revenue), AI Research & Solution: 34.0 billion JPY, AI Powered Worker: 34.0 billion JPY
* SaaS customer base: approximately 10,500 companies (corresponding to approximately 18% market penetration of the relevant Japanese enterprise and SMB market)
* ARPU growth: Increase to 8.50 million JPY per year driven by large-scale company-wide rollouts (similar to the implementation at Nomura Real Estate for 10,000 workstations) and cross-module usage
* Adjusted operating profit (EBIT): approximately 47.5 billion JPY (consolidated margin of 30.2%, driven by an expansion of the SaaS segment margin to 45% as a result of declining token costs and optimized SLMs)
* Net income after taxes: approximately 32.8 billion JPY (~256 million € at 128 JPY/EUR or ~205 million € at 160 JPY/EUR)
* Expected market capitalization: approximately 850 to 940 billion JPY (~6.6 to 7.3 billion € at a P/E ratio of 26x and a reduced number of shares to 30 million)
* Expected share price: approximately 28,427 JPY
* Dividend per share: approximately 328 JPY (payout ratio of 30%)
* Detailed Forecast – Fiscal Year 2040e (Maturity Phase):
* Consolidated revenue: approximately 289.0 billion JPY (~2.26 billion € at 128 JPY/EUR)
* Segment revenues: AI SaaS 182.0 billion JPY (63.0% of total revenue), AI Research & Solution 52.0 billion JPY, AI Powered Worker 55.0 billion JPY
* SaaS customer base: approximately 14,000 companies with an ARPU of 13.00 million JPY per year
* Adjusted operating profit (EBIT): approximately 100.0 billion JPY (consolidated margin of 34.6%)
* Net income after taxes: approximately 69.0 billion JPY (~539 million € at 128 JPY/EUR or ~431 million € at 160 JPY/EUR)
* Expected market capitalization: approximately 1.65 to 1.70 trillion JPY (~13.0 to 13.5 billion € at a P/E ratio of 24x)
* Expected share price: approximately 55,200 JPY
* Dividend per share: approximately 920 JPY (payout ratio of 40%)
* Macroeconomic leverage & labor market shifts:
* A decline in Japan’s working-age population of about 11 million people by 2040 will lead to an acute labor shortage in administrative white-collar roles (call centers, human resources, clerical work, administration).
* PKSHA doesn’t address the limited IT budgets of 1 to 2% of revenue, but instead directly targets the 30 to 50% of corporate costs spent on shrinking staffing capacities.
* Structural transition from traditional per-seat pricing to value- and transaction-based pricing models for autonomous software agents, ensuring that declining customer headcount does not diminish revenue potential.
* Technological foundation & independence from Big Tech:
* Avoiding capital-intensive frontier base models: PKSHA does not train general $100-billion models from scratch but instead uses open-source architectures and trains compact, domain-specific Small Language Models (SLMs) at its own expense.
* Dominance in the voicebot market with an 83.8% market share in Japan through proprietary speech-to-text engines trained on billions of minutes of real call center data, resolve phonetic Kanji ambiguities with latencies under 300 milliseconds, and are natively integrated into legacy telephone systems (PBX/CTI).
* Full integration into Japanese governance structures through the digital mapping of multi-level consensus and approval processes (Ringi) directly within Microsoft Teams and Slack.
* Full data sovereignty through “PKSHA Private AI Agents,” enabling regulated customers in the financial, insurance, and government sectors to run generative agents locally on their own servers without forwarding data to U.S. cloud infrastructures.
* Three-pillar operational flywheel:
* AI Research & Solution (operating margin around 25%): Major clients such as NTT Docomo and Nomura Real Estate pre-fund customized algorithm development; intellectual property (IP) rights remain entirely with PKSHA.
* AI SaaS (gross margin over 70%): Standardization of proven algorithms into recurring software modules with an ARR of over 11.1 billion JPY (+32% YoY), over 4,900 corporate customers, a net retention rate of 107.7%, and a value-up rate of 141% for upgrades to agent solutions.
* AI Powered Worker (APW): The acquisition of Circulation provides access to more than 4,000 small and medium-sized business customers without significant customer acquisition costs (CAC); consultants act as on-site implementation partners and transform traditional consulting and staffing services into high-margin AI-powered BPO.
* Accounting & Valuation Baseline:
* IFRS accounting avoids scheduled goodwill amortization from acquisitions that affects net income, allowing operational growth to flow directly into earnings per share (EPS).
* The market capitalization of approximately 100 billion JPY (~600 million €) does not reflect the underlying software profitability, as the market mistakenly classifies the company as a labor-intensive IT service provider.
* Ongoing disciplined capital return through share buybacks, as well as the first-ever introduction of a regular cash dividend of 10 JPY per share for fiscal year 2026.
* Detailed Forecast – Fiscal Year 2026e (Actual Run Rate):
* Consolidated revenue: approximately 36.5 billion JPY (~228 million €)
* Adjusted operating profit (EBIT): approximately 5.7 billion JPY (margin of 15.6%)
* Net income after taxes: approximately 3.2 billion JPY (~25 million € at 128 JPY/EUR)
* SaaS customer base: approximately 5,100 companies with an average annual revenue per user (ARPU) of 2.30 million JPY
* Base dividend: 10 JPY per share
* Detailed Forecast – Fiscal Year 2030e:
* Consolidated revenue: 73.8 to 82.0 billion JPY (~575 to 640 million € at 128 JPY/EUR)
* Segment revenues: AI SaaS 33.3 billion JPY, AI Research & Solution 21.0 billion JPY, AI Powered Worker 19.5 billion JPY
* SaaS customer base: approximately 7,400 companies with an ARPU of 4.50 million JPY per year
* Adjusted operating profit (EBIT): approximately 16.5 billion JPY (margin of 22.4%)
* Net income after taxes: 11.4 to 13.0 billion JPY (~89 to 102 million € at 128 JPY/EUR)
* Expected share price: approximately 10,466 JPY
* Dividend per share: approximately 75 JPY (payout ratio of 20%)
* Detailed Forecast – Fiscal Year 2035e (Base Case Bottom-Up):
* Consolidated revenue: approximately 157.3 billion JPY (~1.23 billion € at 128 JPY/EUR)
* Segment revenue: AI SaaS 89.3 billion JPY (56.8% of total revenue), AI Research & Solution: 34.0 billion JPY, AI Powered Worker: 34.0 billion JPY
* SaaS customer base: approximately 10,500 companies (corresponding to approximately 18% market penetration of the relevant Japanese enterprise and SMB market)
* ARPU growth: Increase to 8.50 million JPY per year driven by large-scale company-wide rollouts (similar to the implementation at Nomura Real Estate for 10,000 workstations) and cross-module usage
* Adjusted operating profit (EBIT): approximately 47.5 billion JPY (consolidated margin of 30.2%, driven by an expansion of the SaaS segment margin to 45% as a result of declining token costs and optimized SLMs)
* Net income after taxes: approximately 32.8 billion JPY (~256 million € at 128 JPY/EUR or ~205 million € at 160 JPY/EUR)
* Expected market capitalization: approximately 850 to 940 billion JPY (~6.6 to 7.3 billion € at a P/E ratio of 26x and a reduced number of shares to 30 million)
* Expected share price: approximately 28,427 JPY
* Dividend per share: approximately 328 JPY (payout ratio of 30%)
* Detailed Forecast – Fiscal Year 2040e (Maturity Phase):
* Consolidated revenue: approximately 289.0 billion JPY (~2.26 billion € at 128 JPY/EUR)
* Segment revenues: AI SaaS 182.0 billion JPY (63.0% of total revenue), AI Research & Solution 52.0 billion JPY, AI Powered Worker 55.0 billion JPY
* SaaS customer base: approximately 14,000 companies with an ARPU of 13.00 million JPY per year
* Adjusted operating profit (EBIT): approximately 100.0 billion JPY (consolidated margin of 34.6%)
* Net income after taxes: approximately 69.0 billion JPY (~539 million € at 128 JPY/EUR or ~431 million € at 160 JPY/EUR)
* Expected market capitalization: approximately 1.65 to 1.70 trillion JPY (~13.0 to 13.5 billion € at a P/E ratio of 24x)
* Expected share price: approximately 55,200 JPY
* Dividend per share: approximately 920 JPY (payout ratio of 40%)
•
11
•@PikaPika0105 That sounds very promising. You've actually introduced it before. But the cyclical chart put me off back then. So what happened in 2024?
••
@Tenbagger-Capital I didn’t get on board until fall 2025, after all. I think in 2024 they just got totally swept up in the AI hype, but way too early. Of course, the numbers have to back it up first—that’s a common problem.
By the way, I highly recommend Gemini 4 (“Argon”). The analysis results are surprisingly good (always with “high” cognitive performance, of course). I once had Argon run its own analyses on my positions (initially independent of my own views), and the results pointed in a very similar direction. Previously, most AIs were always too “dumb” to connect all the relevant points and be far-sighted enough. I’ll reach my goal and be completely financially independent by age 30. To do that, I’d only need to achieve half of my goals. By age 40, I’ll essentially be able to live off dividends.
By the way, I highly recommend Gemini 4 (“Argon”). The analysis results are surprisingly good (always with “high” cognitive performance, of course). I once had Argon run its own analyses on my positions (initially independent of my own views), and the results pointed in a very similar direction. Previously, most AIs were always too “dumb” to connect all the relevant points and be far-sighted enough. I’ll reach my goal and be completely financially independent by age 30. To do that, I’d only need to achieve half of my goals. By age 40, I’ll essentially be able to live off dividends.
••
@PikaPika0105 Thanks for the tip. I'll check it out. But at the end of the day, I only use AI as a tool and rely on my own research.
••
@Tenbagger-Capital That goes without saying, but it's a huge help when gathering information. It lets me analyze media reports and scan quarterly reports very quickly.
••
@PikaPika0105 That's going to become important again after the earnings report. Thanks, my friend. Maybe we should set up a dedicated agent just for the quarterly reports. I'm sure @Raketentoni already has something like that.
••