OpenAI CEO confirms GPT-5 underperformed in 2025, shifts focus to GPT-6 with faster launch aiming 2026

OpenAI’s recent admission that its GPT-5 model fell short of expectations has sparked fresh debates about the pace of artificial intelligence innovation and its implications for the broader tech sector. As the company pivots towards developing GPT-6, investors are scrutinising whether this shift signals deeper challenges in scaling large language models or merely a temporary setback in a rapidly evolving field.

The Underwhelming Reception of GPT-5

The rollout of GPT-5, announced earlier in August 2025, was intended to mark a significant leap in AI capabilities, building on the successes of previous iterations. However, feedback from users and industry observers has highlighted inconsistencies in performance, particularly in areas like instruction following and real-world task execution. Benchmarks, while showing improvements in specific metrics such as tool calling (reaching up to 96.7% on certain tests), have not translated into the transformative user experience many anticipated. This has led to a perception that the model, despite its technical advancements, struggles with reliability and adaptability in practical applications.

From a financial perspective, this development underscores the high-stakes nature of AI investments. OpenAI, valued at over $150 billion as of late 2024 based on funding rounds, relies heavily on demonstrating continuous progress to justify its capital-intensive operations. The company’s compute demands have escalated dramatically with each model generation, often requiring 100 times more resources than predecessors. Yet, recent reports suggest diminishing returns on these investments, with performance gains following a logarithmic pattern rather than the exponential growth seen in earlier phases.

Implications for AI Scaling Laws

The challenges with GPT-5 align with broader industry trends indicating that simply scaling up compute may no longer yield proportional improvements. Historical data from AI research shows that models like GPT-4, released in 2023, achieved breakthroughs in multitask learning, but subsequent versions have faced hurdles in saturating benchmarks such as MMLU (around 95%) and MATH (40-50%). Analysts project that without innovations in areas like reinforcement learning and efficient compute utilisation, future models could encounter similar plateaus.

Investor sentiment, as gauged by reports from credible sources like CNBC and The Washington Post, reflects a mix of caution and optimism. For instance, market observers note that while GPU shortages continue to constrain development timelines, advancements in reasoning models could enable smaller, more efficient systems to rival larger ones. This could potentially reduce costs, which have ballooned due to the energy and infrastructure needs of training massive neural networks.

Shifting Focus to GPT-6: A Strategic Pivot?

In response to the GPT-5 feedback, OpenAI has signalled an accelerated push towards GPT-6, with indications that this next iteration could arrive sooner than expected. Company leadership has emphasised customisation features, such as allowing users to tailor the model’s style and responses, aiming to address criticisms of GPT-5’s perceived emotional flatness and inconsistency. This move is seen as an attempt to regain momentum in a competitive landscape where rivals like Anthropic and Google are advancing their own AI offerings.

From an investment standpoint, this pivot could have mixed outcomes. On one hand, a faster development cycle for GPT-6 might demonstrate OpenAI’s agility and commitment to iteration, potentially boosting user adoption—ChatGPT now boasts around 700 million users as of mid-2025. On the other, persistent supply chain issues, particularly around GPUs, could delay meaningful progress. Analyst-led forecasts, drawing from models like those discussed in AI research forums, suggest that GPT-6 might incorporate enhanced agentic capabilities, improving on GPT-5’s benchmarks by 10-20% in areas like multi-step reasoning, but only if compute efficiency improves substantially.

Economic and Market Ramifications

The broader economic implications extend to the semiconductor industry and cloud computing providers. Nvidia, a key supplier of GPUs, has seen its market capitalisation soar to over $2 trillion by 2024, largely driven by AI demand. However, if OpenAI and peers encounter scaling limits, this could temper growth expectations. Historical trends show that AI hype cycles have led to valuation bubbles; for example, the dot-com era of the early 2000s saw similar over-optimism followed by corrections.

Valuation metrics for AI firms remain elevated, with price-to-sales ratios often exceeding 20x for leading players. Investors should monitor key indicators such as R&D spending as a percentage of revenue—OpenAI’s has hovered around 40% in recent years—and partnerships with hardware giants. Sentiment from financial outlets like Yahoo Finance indicates growing wariness of an “AI hype bubble,” with warnings that GPU constraints could slow advancements towards more ambitious goals like artificial general intelligence.

Opportunities and Risks for Investors

For those eyeing AI-related investments, the GPT-5 to GPT-6 transition highlights opportunities in ancillary sectors. Companies specialising in AI optimisation tools, data centre infrastructure, and energy-efficient computing stand to benefit. For instance, advancements in smaller models achieving high performance could democratise AI access, expanding markets in regions like India, where affordable subscriptions (around Rs 399) are being introduced.

• Diversification Strategy: Consider exposure to AI enablers rather than pure-play developers to mitigate risks from model-specific setbacks.
• Long-Term Thesis: Despite short-term hiccups, the AI market is projected to grow to $1.8 trillion by 2030, per analyst models from firms like McKinsey.
• Risk Factors: Regulatory scrutiny, ethical concerns over AI “emotional lobotomies,” and competition could cap upside.

In a nod to dry humour, one might say that AI’s path to superintelligence resembles a marathon runner hitting the wall—plenty of sweat, but the finish line keeps receding. Yet, for discerning investors, these moments of recalibration often present the best entry points, provided the underlying technology continues to evolve.

Forecasting the AI Landscape

Looking ahead, analyst models forecast that GPT-6 could debut by late 2026, incorporating feedback loops for self-improvement, though experts caution that true recursive self-enhancement remains elusive. Progress is expected to be gradual, spanning years rather than explosive leaps. Market sentiment from sources like BleepingComputer suggests confidence in iterative improvements, but with caveats around resource ceilings.

Ultimately, OpenAI’s navigation of these challenges will test its resilience and influence investor confidence in the AI sector’s long-term viability.

References

• OpenAI. (2025). Introducing GPT-5. https://openai.com/index/introducing-gpt-5/
• OpenAI. (2025). Introducing GPT-5 for developers. https://openai.com/index/introducing-gpt-5-for-developers/
• OpenAI. (2025). GPT-5 overview. https://openai.com/gpt-5/
• Fast Company. (2025). OpenAI gave GPT-5 an emotional lobotomy. https://www.fastcompany.com/91388232/openai-gave-gpt-5-emotional-lobotomy-crippled-model
• Botpress. (2025). Everything you should know about GPT-5. https://botpress.com/blog/everything-you-should-know-about-gpt-5
• CNBC. (2025, August 7). OpenAI launches GPT-5 model for all ChatGPT users. https://www.cnbc.com/2025/08/07/openai-launches-gpt-5-model-for-all-chatgpt-users.html
• Y Combinator. (2025). User commentary on GPT-5. https://news.ycombinator.com/item?id=44826997
• The Washington Post. (2025, August 17). OpenAI GPT-5 and superintelligence. https://www.washingtonpost.com/technology/2025/08/17/openai-gpt5-chatgpt-superintelligence/
• The Hindu. (2025). OpenAI CEO Sam Altman on GPT-6. https://www.thehindu.com/sci-tech/technology/openai-ceo-sam-altman-bats-for-more-personal-gpt-6-after-mixed-reactions-to-gpt-5-report/article69959404.ece
• PYMNTS. (2025). 10 ways to get better results from GPT-5. https://www.pymnts.com/artificial-intelligence-2/2025/10-ways-to-get-better-results-from-gpt-5/
• Wired. (2025). OpenAI’s GPT-5 is here. https://www.wired.com/story/openais-gpt-5-is-here/
• BleepingComputer. (2025). OpenAI says GPT-6 is coming. https://www.bleepingcomputer.com/news/artificial-intelligence/openai-says-gpt-6-is-coming-and-itll-be-better-than-gpt-5-obviously/
• Finance Yahoo. (2025). OpenAI CEO Sam Altman concedes limitations in GPT-5. https://finance.yahoo.com/news/openai-ceo-sam-altman-concedes-171805241.html
• Indian Express. (2025). OpenAI introduces ChatGPT Go in India at Rs 399. https://indianexpress.com/article/technology/openai-introduces-chatgpt-go-in-india-at-rs-399-upi-10197639/