OpenAI Using AI to Design Next AI Model, Says SoftBank CEO
6h ago|4 min readStandard
Fazen Markets Editorial Desk
Collective editorial team · methodology
artificial-intelligenceopenai
Fazen Markets Editorial Desk
Collective editorial team · methodology
Trades XAUUSD 24/5 on autopilot. Verified Myfxbook performance. Free forever.
Risk warning: CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. The majority of retail investor accounts lose money when trading CFDs. Vortex HFT is informational software — not investment advice. Past performance does not guarantee future results.
SoftBank Group Corp. CEO Masayoshi Son stated on June 5, 2026, that OpenAI is utilizing its own artificial intelligence systems to design its next flagship model. The announcement, made during a technology conference in Tokyo, implies a significant acceleration in the AI development cycle. This move could compress multi-year research timelines and redirect billions in technology sector capital expenditure toward AI infrastructure and specialized chips. The strategy represents a recursive approach to AI advancement, where current systems are tasked with creating more capable successors, a concept often termed recursive self-improvement.
The AI industry is approaching a perceived ceiling in scaling traditional transformer-based large language models. Model training costs have escalated exponentially, with projects like OpenAI's GPT-4 estimated to exceed $100 million. Recent breakthroughs in AI-powered scientific discovery, such as Google DeepMind's GNoME project discovering 2.2 million new crystals in 2024, demonstrated the potential for AI to accelerate research. This precedent likely encouraged OpenAI to apply similar AI-driven discovery techniques to the core problem of AI architecture design itself.
The current macro backdrop features elevated interest rates, with the Fed funds target range at 4.75-5.00% as of May 2026, pressuring speculative tech valuations. Against this, the promise of radical AI efficiency gains presents a compelling narrative for continued investment. The catalyst is the accumulation of evidence that AI can outperform humans in specific complex design and optimization tasks, making the concept of AI-designed AI transition from theoretical to operational.
The global AI chip market is projected to reach $263 billion by 2032, growing at a CAGR of 29.2% from 2023. Nvidia's data center revenue for Q1 2026 was $42.5 billion, a primary beneficiary of current AI training demands. A shift to AI-led design could alter this trajectory, potentially reducing raw compute requirements for future model development by optimizing architectures from the outset.
| Metric | Pre-AI-Design Paradigm | AI-Design Paradigm (Projected) |
|---|---|---|
| Architecture R&D Timeline | 18-24 months | 6-9 months |
| Estimated Training Compute | 10^26 FLOPs | 10^24-10^25 FLOPs |
| Human Researcher Focus | Foundational Design | Goal Specification & Validation |
Specialized AI accelerator startups like Cerebras Systems and SambaNova Systems, which focus on novel chip architectures, may see increased interest. Their market valuations, often private, are tied to the premise that non-Nvidia architectures can achieve superior performance for specific AI workloads.
The immediate second-order effect is a bullish signal for semiconductor capital equipment firms like ASML and KLAC, as AI-driven design could necessitate more exotic and precise chip manufacturing processes. AI software vendors MSFT and GOOGL stand to gain from more efficient, lower-cost model development, potentially expanding their cloud AI service margins. Conversely, a successful implementation risks creating a winner-take-most dynamic, further consolidating power among the few entities with the data and base models required to bootstrap this recursive loop.
A key risk is the black box problem; an AI-designed architecture may be less interpretable to human engineers, complicating debugging and safety verification. This could attract regulatory scrutiny from bodies like the US AI Safety Institute. Trading desks report increased flows into thematic ETFs like the Global X Robotics & Artificial Intelligence ETF (BOTZ) on the news, while long/short funds are establishing pairs trades, long semiconductor equipment and short traditional semiconductor foundries with less exposure to cutting-edge AI designs.
The next significant catalyst is OpenAI’s DevDay conference, typically held in November, where a technical paper or announcement detailing progress is likely. Investors should monitor Nvidia’s next earnings call on August 21, 2026, for commentary on demand shifts from brute-force training to optimized inference and design workloads. The 50-day moving average for the Nasdaq-100 index (NDX), currently at 19,200, serves as a key technical support level for the broader tech sector sentiment.
A breakthrough before year-end 2026 could trigger a re-rating of AI-related stocks, while a delay or failure to demonstrate tangible results would likely pressure high-multiple names. The 10-year Treasury yield, trading at 4.35%, remains a critical gauge for risk appetite; a sustained break above 4.50% could dampen capital flows into long-duration tech assets regardless of AI news.
The demand for top-tier AI research talent is unlikely to diminish but will shift in focus. Instead of manual neural architecture search, researchers will concentrate on formulating better objectives for the designing AI, curating high-quality training data, and implementing rigorous safety alignment protocols. The skill set will evolve from hands-on model building to supervisory and strategic roles, potentially increasing compensation for experts who can effectively guide AI development systems.
This development parallels the shift from manual circuit design to electronic design automation (EDA) in the 1980s. Companies like Cadence Design Systems automated chip layout, which did not eliminate chip designers but massively increased their productivity and the complexity of feasible designs. The key difference is the recursive potential; EDA tools are static programs, whereas an AI designing AI could improve its own design capabilities, creating a feedback loop with a much higher acceleration ceiling.
The primary risks involve unpredictability and control. An AI may discover architectures that are highly effective but whose internal workings are incomprehensible, making it difficult to guarantee the model's behavior or eliminate biases. There is also a risk of specification gaming, where the AI fulfills the literal technical objective but in an unintended, potentially harmful way. This underscores the critical importance of strong safety research running in parallel with capabilities research.
OpenAI's recursive AI development strategy marks a pivotal inflection point for the entire technology sector's capital and talent allocation.
Disclaimer: This article is for informational purposes only and does not constitute investment advice. CFD trading carries high risk of capital loss.
Vortex HFT is our free MT4/MT5 Expert Advisor. Verified Myfxbook performance. No subscription. No fees. Trades 24/5.
Position yourself for the macro moves discussed above
Start TradingSponsored
Open a demo account in 30 seconds. No deposit required.
CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. You should consider whether you understand how CFDs work and whether you can afford to take the high risk of losing your money.