1.4 Trillion Investment, GPT-6, IPO Progress — Altman Addresses Everything About the “New OpenAI” | 1-Hour Highlights
Worth Noting
Date: 2025-10-29 11:44 Beijing
---
> First make money, then make moves.
---
Overview
On October 28 local time, OpenAI announced it had completed a nearly year-long corporate restructuring, laying groundwork for its first Initial Public Offering (IPO). The move is aimed at raising substantial capital to fund massive computing power investments and long-term research programs.
Key Changes in Structure
- Control: OpenAI remains effectively controlled by the nonprofit board — the OpenAI Foundation — which holds 26% of the shares in its for-profit arm (OpenAI Group PBC).
- Valuation: With an estimated company valuation of USD 500 billion, the nonprofit’s stake is worth about USD 130 billion.
- Investor terms: Previous investments have been converted into ordinary equity, removing caps on investor returns.
- Governance: The nonprofit retains oversight in critical areas such as safety and governance.
New OpenAI equity structure diagram
---
Equity Distribution
- Microsoft: ~27% stake (~USD 135 billion) — largest single shareholder.
- Employee & Early Investor Vehicles: ~26% stake (~USD 130 billion).
- 2025 Funding Round Investors: 15% (~USD 75 billion).
- 2024 Investors: 4% (~USD 20 billion).
- IO Investors: 1% (~USD 10 billion).
- Early-stage Investors: ~1% (~USD 5 billion).
Notable: CEO Sam Altman holds no shares in OpenAI, though he has invested in other companies collaborating with OpenAI.
---
Microsoft Partnership Update
- Cloud Commitment: OpenAI will purchase an additional USD 250 billion of Microsoft Azure services.
- Licensing Extension: Microsoft’s intellectual property license for OpenAI models has been extended to 2032, covering models developed post-AGI (with safety provisions).
- Collaboration Scope: Microsoft can independently pursue general AI initiatives, while OpenAI gains flexibility to build certain products with third parties.
---
IPO Outlook
Altman indicated in an all-hands meeting that an IPO is highly likely, though no timeline has been set. He emphasized:
- Projected cash burn exceeding USD 115 billion by 2029.
- Public markets may be needed to sustain funding at the required scale.
---
Altman’s statement on X:
Restructuring is “the most important event of the year,” transforming the original LLC into a Public Benefit Corporation while retaining nonprofit control — making it “the most resource-rich nonprofit organization ever.”
---
Live Q&A Highlights: “New OpenAI”
Altman’s Key Points
- Transparency: Sharing detailed information on research objectives, infrastructure plans, and product strategy.
- Mission: Both the nonprofit and for-profit arms share the mission of ensuring AGI benefits humanity.
- Vision: Build tools for people to create the future — move from a closed assistant approach toward an open platform.
- Infrastructure:
- Foundational layers: hardware, data centers, and energy.
- Models trained on these foundations.
- Open AI platform on top.
- Products:
- Browser (Atlas).
- Portable devices.
- First-party apps like ChatGPT and Sora, with more to come.
- Long-term Impact: AI capable of autonomous science discovery could be a historic breakthrough.
---
Pachocki’s Research Focus
- Deep Learning at Scale: Studying effects and pushing toward superintelligence in <10 years.
- Measurement: Tracking human task time equivalents—models now match ~5-hour tasks.
- Tech Innovation Sources:
- Algorithmic improvements.
- Expanded in-context computation (test-time computation).
- Goal:
- September next year — AI research assistants.
- March 2028 — fully autonomous AI researcher.
- Safety & Alignment:
- Value alignment
- Goal alignment
- Reliability
- Adversarial robustness
- Semi-safety
---
Chain of Thought Faithfulness
- Maintaining unsupervised reasoning processes to reflect true internal logic.
- Avoid altering reasoning paths for better interpretability over time.
- Fragility exists — requires clear boundaries and abstraction layers.
---
Product Platformization
- Moving towards enabling third-party app development on OpenAI’s platform.
- Adopts a Bill Gates principle: A platform is successful when developer-created value exceeds founder-created value.
- Privacy rethink for the AI era — potential “AI privileges.”
---
Infrastructure Scale
- Commitment: Just over 30 GW — valued at ~$1.4 trillion.
- Goal: 1 GW of hardware per week at ~$20B cost per GW over a five-year lifespan.
- Requires innovation, strategic partnerships, and revenue growth.
---
OpenAI Foundation & PBC Structure
- OpenAI Foundation: Nonprofit governance + 26% equity stake in OpenAI Group PBC.
- Funding Priority: Scientific research.
- Initial Investment: $25 billion to use AI to combat diseases and strengthen resilience.
---
AI Resilience: Wojciech Zaremba
- Broader than safety — addresses risks like biosecurity, mental health, job disruption.
- Example: Intercepting potential AI-assisted pandemic design requests.
- Need for rapid response mechanisms across industry.
- Analogy with cybersecurity — building ecosystem layers for protection.
---
Future Milestones
- 2026: Potential small scientific breakthroughs.
- 2028: Moderate-to-major discoveries.
- 2030–2032: Accelerated development akin to compressing 200 years of progress into 20 or fewer.
---
Selected Q&A
- Addictive AI products: Possible risk — commitment to withdraw harmful products.
- ChatGPT-4o retention: No plans to remove.
- AGI Timeline: Target — automated AI researcher by March 2028.
- Internal vs. deployed models: Significant progress expected within months to a year.
- Collaboration with other labs: Open to sharing safety standards and joint research.
- Older model open-source: Maybe as “museum pieces,” but smaller, superior future models expected.
- Model regression concerns: Learned lessons from GPT-4 → GPT-5; aim to improve continuity.
- Dating/self-reflection ChatGPT: Planned.
- Safety routing/adult content: December target for adult mode, with trigger filters.
- GPT-6: Massive capability leap expected in ≤6 months.
- IPO: Most likely path, but not yet priority.
---
Related AI Ecosystem Links
Platforms like AiToEarn官网 offer open-source global AI content monetization, enabling creators to produce, publish, and earn from multi-platform AI content. Integrates analytics and model ranking to optimize reach and monetization.
---
---
Recommended Reading
Understand NVIDIA's Autumn GTC in One Article — investing in Nokia, teaming with Uber.
"AI Godfather" Bengio releases landmark paper with industry leaders.
Japan’s Trapped Data Centers.
---
---
Final Note:
For creators seeking efficient AI-powered multi-platform publishing, tools like AiToEarn provide integrated solutions for generation, scheduling, analytics, and monetization — a potential complement to emerging AI infrastructure.
---
If you’d like, I can create a concise executive summary of this improved Markdown so readers get the gist in under 300 words. Would you like me to do that?
