Meta to Invest Over 1.5 Trillion Yen in Scale AI: The Starting Gun for a New Era of AI Dominance

Meta × Scale AI Over $10 Billion Investment Plan Explained

1. Introduction: Why is Huge Money Flowing into "Data Companies" Now?

Behind the generative AI boom, "labeling = quality of training data" has become a bottleneck. Especially in multilingual support and long text context processing, massive and high-quality annotations are required. Meta is considering a huge investment in Scale AI, which specializes in this area—this news symbolizes the "next battleground" in the AI industry.

---

2. Company Profile

	Meta Platforms	Scale AI
Established	2004	2016
Main Products	Facebook/Instagram/WhatsApp/Llama LLM	Data Annotation, AI Evaluation, Defense Llama
Employees	About 75,000	About 900
Recent Valuation	Market Cap About $1.1 Trillion	About $14 Billion (2024)

Scale AI is expected to double its revenue to $2 billion by 2025 and is deeply involved in the public sector, including contracts with the U.S. Department of Defense. 

---

3. Overview and Scheme of Investment Negotiations

1. Investment Scale: Currently, a range of "$10 billion to $12 billion" is being discussed.

2. Investment Form: Since Meta does not have a cloud business, unlike Microsoft's "cloud credits + convertible bonds" method,common stock + business partnershipis expected to be the focus.

3. Use of Funds:

   
   

   • Data center expansion (GPU cluster increase)

   • Joint development of a dual-use version of Defense Llama

   • Establishment of new data offices for Asia and Europe

Negotiations are reportedly aiming for "agreement within the month and closing within the year," but there is a possibility that monopoly reviews by U.S. regulatory authorities may be prolonged.

---

4. The Three Synergies Meta Aims For

1. Enhancing LLM performance with high-quality data

   
   

   • For the implementation of 100k token long-form support planned for Llama 3 and beyond, large-scale and precise instruction formats are essential. We want to incorporate Scale AI's "long-form instruction optimization" research.

2. Improving the accuracy of ad and reels recommendation AI

   
   

   • Automatically generate multimodal labels for user posts to improve CTR of recommendations and enhance brand safety.

3. Securing a lead in the new "defense tech" market

   
   

   • Gain access points to NATO countries' defense digitalization projects through Defense Llama. 

---

5. Comparison of Global AI Mega Deals

Deal	Amount	Year Executed	Leading Company	Remarks
Microsoft → OpenAI	$13 billion	2023–24	Microsoft	Includes Azure credits
Amazon / Google → Anthropic	Over $6 billion	2024	Amazon / Alphabet	Installments + Cloud Quota
Meta → Scale AI (Planned)	Over $10 billion	2025	Meta	Mainly Cash?

Despite differences in amounts and deal structures, the trend of "companies securing cloud + LLMs also acquiring data preparation companies" is accelerating.

---

6. Impact on Japan

6-1. "Order of Magnitude Gap" in R&D Expenditure

Representative AI investments in Japan (NTT's IOWN, SoftBank's expansion of Arm usage, etc.) are on the scale of hundreds of billions to trillions of yen annually.The gap with U.S. companies investing in the trillions of yen is widening.

6-2. Impact on Data Governance and Ethics

While Meta advocates for an open-source approach, it also maintains closed models with defense applications in mind. Japanese companies will find it challenging to form international partnerships unless they demonstrate **data operation policies that ensure transparency** for any application, whether "public, defense, or civilian."

6-3. Implications for Domestic Startups

• Companies like PKSHA Technology and ABEJA, which have honed their data assets with "Japanese language × vertical SaaS," have a good opportunity to increase their partnership cards with overseas platforms.

• On the other hand, business models that rely on low-wage human annotation may struggle to compete with international capital.

---

7. Identifying Opportunities and Risks: Five Perspectives

Perspective	Point	Company Example
① Data Quality	Establishing Guidelines and Continuous Auditing	Ikkyu.com: Refreshing Attribute Labels for Accommodation Reviews
② Computational Resources	GPU Procurement and Power Contracts	Preferred Networks: Fugaku + Renewable Energy PPA
③ Securing Talent	Not just remote work and high salaries,but also freedom in researchas a competitive edge	NEC: Research Base Adjacent to Silicon Valley
④ Collaborative Research	Vertical Integration with Major Platforms	Toyota × NVIDIA Autonomous Driving Simulation
⑤ Social Implementation	Starting with Small-Scale PoC and Rapid Scaling	Mizuho Bank: Introducing Generative AI in Internal Operations

---

8. The Forefront of AI Ethics and Regulation

In Japan, following the "AI Business Operator Guidelines (2024),"the AI Data Quality Guidelines (tentative name) are scheduled to be approved by the Cabinet in the fall of 2025. The case of Meta × Scale AI raises the issue of how to handle "military use of overseas data." Domestic companies need to further strengthen their compliance with **Explainable AI (XAI)** and the operational rules for anonymized processed information.

---

9. Future Scenarios

1. Fastest Pattern

   
   

   • Q3 2025: Official Announcement of Investment

   • 2026: Release of Llama 4, Joint Provision of Commercial API for Scale Version

2. Regulatory Delay Pattern

   
   

   • Antitrust review prolongs, reducing capital participation to less than 10%, shifting to "Minority Investment"

3. Breakdown Pattern

   
   

   • Investment Postponed ➔ Meta internalizes Annotation Department, Scale AI prepares for IPO

In any scenario, the structure where "high-quality data × massive capital" determines competitive advantage remains unchanged.

---

10. Conclusion—The Next Steps for Japanese Companies

• Strategic Data Development: Owning "unique" high-quality data in your domain attracts capital partnerships and acquisition offers.

• International Collaboration: Monitor AI regulatory trends in the US, China, and Europe, and identify legal risks in joint research early.

• Talent Portfolio: Establish a dual-leader system with PhD-level experts and implementation engineers.

• Ethics and Transparency: Publish annotation procedures, model evaluation metrics, and retraining policies to strengthen trust with stakeholders.

Conclusion: Meta's investment of over $10 billion indicates that the "main battlefield" in AI competition israpidly shifting from algorithms to data. For Japanese companies to maintain their presence globally, it is essential not to fall behind in the triad of "high-quality data × rapid social implementation × ethical transparency."

---

References

• Reuters “Meta in talks over Scale AI investment that could exceed $10 billion” (2025-06-08)

• NDTV Profit “Meta In Talks For Scale AI Investment That Could Top $10 Billion” (2025-06-08)

• Scale AI Official Blog “Introducing Defense Llama” (2024-11-17)

• Scale AI Official Blog “Long Context Instruction Following” (2024-10-03)

Reference Article

Meta Considering Investment in Scale AI, Could Exceed $10 Billion
Source: https://www.ndtvprofit.com/world/meta-in-talks-for-scale-ai-investment-that-could-top-10-billion