ChatGPT

OpenAI Feels Pressure from Google's Surge: Altman's Internal Memo and Shallotpeat Project

As Google's Gemini 3 tops benchmarks, OpenAI CEO Sam Altman warns of 'economic headwinds' in internal memo. OpenAI plans comeback with new Shallotpeat project, recognizing the continued importance of pre-training.

OpenAI Google ChatGPT Gemini Gemini 3 AI Competition Sam Altman AI LLM Large Language Models Deep Learning GPT-5 GPT

The battle for leadership in the AI industry has entered a new phase. As Google tops nearly all benchmarks with Gemini 3, OpenAI CEO Sam Altman warned in an internal memo that Google’s surge could bring “temporary economic headwinds” to the company. Amid this crisis awareness, OpenAI has launched a new project called “Shallotpeat” to mount a comeback in the AI race.

Sam Altman’s Internal Memo: A Sign of Crisis Awareness

According to an internal OpenAI memo obtained by The Information, Sam Altman frankly acknowledges the threat from competitors.

Key Contents of the Memo

Altman’s Message:

  • Economic Headwinds: Recent AI progress by Google and Anthropic could create “temporary economic headwinds”
  • Shrinking Technical Lead: OpenAI’s technological lead over competitors (like Google and Anthropic) is shrinking
  • Optimistic Outlook: However, emphasizes that OpenAI is “catching up fast”
  • Future Confidence: Expects to emerge as the leader in the AI race

Timing of the Memo

Importantly, this memo was written before the Gemini 3 release.

Timeline:

  1. Memo Creation: Altman issues internal memo before Gemini 3 release
  2. Gemini 3 Release: November 18, 2025, Google announces Gemini 3
  3. Benchmark Results: Gemini 3 tops nearly all benchmarks

Altman’s concerns became reality with the Gemini 3 release.

Unusual Nature of the Internal Memo

It’s unusual for a CEO to acknowledge “economic headwinds” from competitor success internally.

What This Indicates:

  • Serious Crisis Awareness: OpenAI truly takes Google’s threat seriously
  • Transparency: Attitude of frankly communicating risks to employees
  • Motivation: Sharing crisis awareness to inspire employees
  • Investor Consideration: Anticipating memo leaks and showing there are countermeasures

Google’s Surge with Gemini 3

Altman’s concerns proved accurate with the Gemini 3 release.

Overwhelming Advantage in Benchmarks

Google achieved top scores on nearly all major benchmarks with Gemini 3.

Key Benchmark Results:

  • LMArena: Industry-leading score
  • Coding Tasks: Superior performance in many areas
  • Math Tasks: High accuracy
  • Reasoning Tasks: Complex problem-solving ability
  • Multimodal: Integrated processing of text, images, and video

Impact on OpenAI and Anthropic

According to The New York Times, the Gemini 3 announcement caused anxiety at OpenAI and Anthropic.

Industry Insider Concerns:

  • Autonomous Coding: Gemini 3 could potentially surpass OpenAI and Anthropic in this area
  • Image Generation: Enhanced multimodal capabilities could threaten competitors
  • Business Impact: If Google’s model proves superior, it could negatively impact OpenAI and Anthropic’s businesses

Google’s Structural Advantages

Google has structural advantages that OpenAI and Anthropic lack.

Google’s Strengths:

  1. Massive User Base: Billions of Google Search, Gmail, Google Docs users
  2. Immediate Deployment: Integrating new model into Search from day one
  3. Data Advantage: Access to vast user data and search data
  4. Integrated Ecosystem: Complete integration of search, cloud, productivity tools
  5. Financial Power: Alphabet/Google’s enormous financial resources

OpenAI’s Challenges with Pre-training

The Information’s report revealed technical challenges OpenAI faces.

Pre-training Scaling Problems

OpenAI was reportedly struggling to make progress in pre-training.

Problem Details:

  • Scaling Limitations: Optimizations stop working when scaling up models
  • GPT-5 Struggles: This problem was prominent during GPT-5 development
  • Unexpected Behavior: Optimizations that worked at small scale failed at large scale
  • Development Delays: This problem delayed GPT-5 development beyond schedule

Shift to Reasoning Models

Difficulties with pre-training led OpenAI to focus on reasoning models.

Strategy Change:

  • o1 Series: Development of models specialized for reasoning
  • Chain-of-Thought: Explicitly modeling step-by-step thought processes
  • Quality Focus: Focusing on improving reasoning quality rather than simple scaling

Pre-training is “Not Dead”

Altman’s memo and reports indicate OpenAI is recognizing the importance of pre-training.

Importance of Pre-training:

  • Foundation Building: Learning knowledge and patterns that form the basis of all capabilities
  • Generality: Acquiring general-purpose capabilities for broad tasks
  • Scaling Laws: Predictable performance improvements through proper pre-training
  • Source of Competitiveness: Excellent pre-training is essential for cutting-edge models

Solving this problem is OpenAI’s top priority.

Shallotpeat: OpenAI’s Comeback Project

OpenAI has launched a new project called “Shallotpeat.”

Overview of the Shallotpeat Project

Project Details:

  • Codename: Shallotpeat
  • Purpose: Counter Google and regain leadership in the AI race
  • Focus: Improving pre-training and developing new models
  • Timeline: Details unclear but progressing rapidly

Project Goals

What Shallotpeat Aims to Achieve:

  1. Solve Pre-training Problems: Overcome optimization issues during scaling
  2. Counter Google: Develop models with performance surpassing Gemini 3
  3. Restore Technical Leadership: Regain technological advantage in the AI industry
  4. Stabilize Business: Overcome “economic headwinds” from competitor success

Meaning of the Project Name

The name “Shallotpeat” is interesting.

Possible Interpretations:

  • Shallot: A small but strongly flavored ingredient
  • Peat: Foundational soil or fuel
  • Combined Meaning: Perhaps intending that small improvements accumulate to create a strong foundation

However, since OpenAI’s internal project codenames are typically not publicly explained, this remains speculation.

Structural Changes in the AI Industry

The competition between OpenAI and Google indicates structural changes in the AI industry.

Shift in Technical Leadership

OpenAI has long been the technical leader in the AI industry, but that position is wavering.

OpenAI’s Past Advantages:

  • GPT-3 (2020): Demonstrated the potential of large language models
  • ChatGPT (November 2022): Mainstreamed AI chatbots
  • GPT-4 (March 2023): Multimodal and advanced reasoning capabilities
  • GPT-5 (2025): Further performance improvements

Current Situation:

  • Google Gemini 3: Top in nearly all benchmarks
  • Anthropic Claude: Surpasses OpenAI in some tasks
  • xAI Grok: Temporarily topped benchmarks

Diversification of Development Approaches

AI companies are adopting different development approaches.

Main Approaches:

  1. Large-scale Pre-training: Google’s strength, successful with Gemini 3
  2. Reasoning Models: OpenAI’s recent focus, o1 series
  3. Safety-focused: Anthropic’s approach, Constitutional AI
  4. Real-time Data: xAI Grok’s strength, access to X data

These approaches each have different strengths and weaknesses, showing the diversity of the AI industry.

Intensifying Benchmark Competition

Benchmark competition among AI companies is progressing at unprecedented speed.

November 2025 Benchmark Changes:

  • Early November: OpenAI GPT-5.1 among the top
  • November 17: xAI Grok 4.1 tops LMArena
  • November 18: Google Gemini 3 reclaims first place

This speed of competition demonstrates rapid AI technology advancement, but simultaneously creates significant pressure for companies.

OpenAI’s Response Strategies

OpenAI is taking multifaceted responses to Google’s surge.

Technical Responses

Shallotpeat Project:

  • Solve pre-training problems and develop new models
  • Aim for performance that can compete with Google

Continued Development of Reasoning Models:

  • Improvements to o1 series and new versions
  • Advanced reasoning techniques like Chain-of-Thought

Business Responses

Ecosystem Expansion:

  • Hardware partnership with Foxconn (announced November 21)
  • $3 billion data center investment
  • Addition of ChatGPT group chat features

Partnership Strengthening:

  • $38 billion contract with AWS
  • Strategic partnerships with companies like Emirates
  • Collaborative relationships with educational institutions

Communication Strategy

Improved Transparency:

  • Frank situation explanation through Sam Altman’s internal memo
  • Sharing crisis awareness and inspiring employees

Maintaining Presence Through Multiple Announcements:

  • Multiple major announcements on November 21
  • Countering Gemini 3 news

Google’s Strategic Moves

Google has gained the upper hand with the Gemini 3 release.

Advantages in Development Methodology

Gemini 3’s success demonstrates that Google’s development methodology is working.

Google’s Success Factors:

  1. Continued Large-scale Pre-training: Achieved excellent results in pre-training
  2. DeepMind Expertise: World-class AI research team
  3. Abundant Resources: Alphabet’s enormous financial power and computational resources
  4. Data Advantage: Vast data from Google Search, YouTube, etc.
  5. Integrated Ecosystem: Integration with search, cloud, productivity tools

Immediate Large-scale Deployment

Google integrated Gemini 3 into Google Search from day one.

Deployment Impact:

  • Reach to Hundreds of Millions: Available to hundreds of millions of users from day one
  • Real-world Feedback: Data collection from large-scale real usage
  • Brand Strengthening: Enhanced brand value through integration of Google Search and AI

Continuous Improvement

Google is continuously improving Gemini 3.

Expected Improvements:

  • Corrections based on user feedback
  • Addition of new features
  • Public release of Gemini 3 Deep Think

Impact on the Entire AI Industry

Competition between OpenAI and Google affects the entire AI industry.

Acceleration of Development Speed

Competition has dramatically accelerated AI technology development speed.

Impacts:

  • Accelerated Innovation: New ideas are implemented rapidly
  • Faster Benchmark Updates: Top positions change every few weeks
  • Increased Pressure: Companies must constantly produce latest results

Impact on Smaller Companies

Intense competition among major companies affects smaller AI firms.

Challenges:

  • Resource Gap: Growing gap in computational resources, talent, funding with major players
  • Difficulty Differentiating: Cannot compete with major players in general-purpose models
  • Acquisition Potential: May become acquisition targets for major companies

Opportunities:

  • Specialization in Niche Markets: Models specialized for specific industries or applications
  • Pursuit of Expertise: Advantages in specialized fields major players don’t address
  • Cost Efficiency: Providing more efficient models or services

Benefits for Users

This competition is ultimately beneficial for users.

User Benefits:

  1. Performance Improvements: Competition leads to development of better AI models
  2. New Feature Additions: Companies add new features for differentiation
  3. Price Reductions: Competition makes services more affordable
  4. Increased Choice: Can choose from multiple powerful AI platforms
  5. Accelerated Innovation: New use cases and applications emerge

Future Development Predictions

Competition between OpenAI and Google is expected to intensify further.

Short-term Predictions (Next 3-6 Months)

Expected Developments:

  1. Shallotpeat Results: OpenAI announces new model or significant improvements
  2. Google’s Response: Improved version of Gemini 3 or new features added
  3. Anthropic’s Moves: Claude model updates to compete with OpenAI and Google
  4. Continued Benchmark Competition: Companies successively release new models

Medium to Long-term Outlook (6 Months - 2 Years)

Industry Changes:

  1. Ecosystem Competition: Competition at integrated ecosystem level, not single models
  2. Balancing Specialization and Generalization: Combination of general-purpose and specialized models
  3. Hardware-Software Integration: Each company develops proprietary hardware
  4. Strengthened Regulation: Development of regulations accompanying AI technology advancement
  5. Emergence of New Players: Startups and existing tech companies enter AI competition

OpenAI’s Challenges and Opportunities

Challenges OpenAI Faces:

  • Solving pre-training scaling problems
  • Countering Google’s integrated ecosystem
  • Restoring technical leadership
  • Stabilizing business

OpenAI’s Opportunities:

  • Comeback through Shallotpeat project
  • Differentiation through hardware entry
  • Leveraging strong partnerships
  • Maintaining advantage in reasoning models

Renewed Recognition of Pre-training Importance

This competition has renewed recognition of pre-training’s importance.

”Pre-training is Not Dead”

While some industry insiders believed “pre-training is no longer important,” Gemini 3’s success proved this wrong.

Why Pre-training is Important:

  1. Foundational Knowledge: Acquiring broad knowledge and patterns
  2. Generality: Basic capabilities for various tasks
  3. Scalability: Predictable performance improvements through proper pre-training
  4. Downstream Task Performance: Pre-training quality affects all downstream task performance

Balancing Reasoning and Pre-training

Cutting-edge AI models require both excellent pre-training and reasoning capabilities.

Ideal Approach:

  • Strong Foundation: Large-scale, high-quality pre-training
  • Advanced Reasoning: Reasoning techniques like Chain-of-Thought
  • Efficient Architecture: Efficient use of computational resources
  • Continuous Improvement: Improvements based on user feedback

Summary

The AI competition between OpenAI and Google indicates a new phase for the industry.

Key Points:

  1. Leadership Shift: Google tops nearly all benchmarks with Gemini 3
  2. OpenAI’s Crisis Awareness: Sam Altman’s internal memo warns of “economic headwinds”
  3. Technical Challenges: OpenAI faces pre-training scaling problems
  4. Shallotpeat Project: OpenAI launches new project as comeback strategy
  5. Importance of Pre-training: Proven that “pre-training is not dead”

Sam Altman’s frank acknowledgment of competitor threats internally demonstrates that OpenAI has a serious sense of crisis. Simultaneously, launching the Shallotpeat project shows that OpenAI is not merely taking a defensive stance but actively planning a comeback.

Google’s success proved that large-scale pre-training remains essential for building cutting-edge AI models. It became clear that OpenAI’s focus on reasoning models stemmed from scaling problems with pre-training. The Shallotpeat project aims to solve this problem and restore OpenAI’s position as the leader in the AI race.

While this competition is intense and sometimes harsh for companies, it’s ultimately good for users. As multiple powerful companies compete with each other, AI technology advances rapidly, and better products are delivered faster and at more affordable prices.

Over the coming months and years, it will be very interesting to see what results OpenAI’s Shallotpeat project produces, how Google improves Gemini 3, and how other players like Anthropic and xAI respond. The AI industry can be said to be at a historic turning point.