Dan Ives on Anthropic, AI M&A, and OpenAI Losses

🔑 Enhanced Key Takeaways

•The Trump administration imposed export controls on Anthropic's latest AI models, Mythos 5 and Fable 5, due to a jailbreak vulnerability and concerns over military applications, leading Anthropic to disable global access to these models.
•OpenAI's significant operational losses, estimated at $8 billion in 2025 (excluding a $30 billion non-cash accounting adjustment), are largely driven by escalating inference costs, which reached $8.4 billion in 2025 and are projected to hit $14.1 billion in 2026.
•The AI M&A landscape in 2026 is characterized by increased discipline, with strategic acquirers prioritizing AI-native companies that demonstrate proprietary data moats, proven agentic capabilities, and high Net Revenue Retention (NRR) over those merely adding AI features.
•Anthropic was designated a 'supply chain risk' by the Department of Defense in February 2026, and federal agencies began phasing out its models after the company refused to remove contractual prohibitions on using Claude for mass domestic surveillance and fully-autonomous weapons.
•OpenAI confidentially filed its S-1 prospectus with the SEC in May 2026, targeting a public listing as early as September 2026 with an expected valuation exceeding $1 trillion, despite internal projections of continued losses until around 2030.

📊 Competitor Analysis▸ Show

Company	Model Focus / Key Strengths	Context Window (Tokens)	Pricing (Input/Output per million tokens)
OpenAI	General-purpose LLMs (GPT series), multimodal (GPT-4o), advanced reasoning (GPT-5 router)	Up to 128K (GPT-4), 131,072 (GPT-OSS)	Varies by model, e.g., GPT-5.4, o3/o4-mini
Anthropic	AI safety (Constitutional AI), strong coding (Opus 4), vision capabilities, long-running agents	Up to 1M (Sonnet 4/4.5 preview)	Opus 4: $15/$75; Sonnet 4: $3/$15
Google Gemini	Multimodal, rapid evolution, strong for multimodal applications and mobile integration	Varies by model (e.g., Gemini 3 Pro)	Varies by model
xAI Grok	Aggressive pricing, large context, strong reasoning	2M (Grok 4.1)	$0.20/$0.50 (Grok 4.1)
DeepSeek	High intelligence at lower cost, competitive benchmarks, open-weight models	Varies by model (e.g., V3.2, R1)	Varies by model, generally lower cost
Cohere	Enterprise RAG specialist, retrieval-augmented generation, citations, tool use	128K (Command R+)	Varies by model

Note: Gartner Peer Insights rated Anthropic higher than OpenAI in service and support, and evaluation and contracting.

🛠️ Technical Deep Dive

Anthropic Claude Models:
- Trained using 'constitutional AI' to enhance ethical and legal compliance.
- Typically released in three sizes: Haiku, Sonnet, and Opus, with Opus being the most capable.
- Claude Opus 4 is recognized as a leading coding model, achieving 72.5% on SWE-bench and 43.2% on Terminal-bench.
- Latest models like Sonnet 4 and 4.5 offer an expanded context window of up to 1 million tokens (preview), equivalent to approximately 150,000 words or over 500 pages.
- Possess best-in-class vision capabilities, accurately transcribing text from imperfect images and understanding diverse visual formats like charts and diagrams.
- Utilize cloud computing resources from Amazon Web Services and Google Cloud Platform, supported by PyTorch, JAX, and Triton development frameworks.
OpenAI GPT Models:
- Based on the transformer deep learning architecture, specifically using only the decoder part for next-token prediction.
- GPT-4 (released 2023) adopted a Mixture of Experts (MoE) architecture, reportedly with approximately 1.8 trillion total parameters across 120 layers and 16 expert networks, and introduced multimodal input (vision) with a 128K token context window.
- GPT-4o (released May 2024) is a multilingual and multimodal model capable of processing audio, visual, and text inputs in real-time.
- GPT-5 (released August 2025) features a real-time router that dynamically switches between a faster mode for simple queries and a 'thinking' mode for complex reasoning, achieving high scores on math (94.6% on AIME 2025) and coding (74.9% on SWE-bench Verified).
- GPT-OSS (September 2025) incorporates RoPE for context lengths up to 131,072 tokens and uses banded window attention for efficient long context processing, also employing an MoE architecture.

🔮 Future ImplicationsAI analysis grounded in cited sources

AI regulatory frameworks will become more interventionist.

The Trump administration's direct export controls on Anthropic's models signal a shift from light-touch regulation to active government intervention in AI development and deployment, particularly concerning national security and safety.

The AI industry will see continued consolidation and strategic acquisitions.

The trend of accelerated AI M&A, driven by the need for scale, capabilities, and talent, along with AI becoming a core valuation driver, indicates that larger players will continue to acquire specialized AI-native companies.

OpenAI's IPO will test investor appetite for high-growth, high-loss AI companies.

Despite rapid revenue growth and a projected trillion-dollar valuation, OpenAI's substantial operational losses and massive infrastructure spending commitments will scrutinize public market investors' tolerance for long-term unprofitability in the pursuit of AI dominance.

⏳ Timeline

2015-12

OpenAI founded as a nonprofit organization.

2021-01

Anthropic founded by former OpenAI employees.

2023-03

Anthropic releases its Claude AI chatbot.

2025-08

OpenAI releases GPT-5, featuring a real-time router for dynamic reasoning.

2026-02

Anthropic designated a 'supply chain risk' by the Department of Defense.

2026-06-08

OpenAI publicly confirms confidential S-1 filing for IPO.

2026-06-12

Trump administration issues export control directive on Anthropic's Mythos 5 and Fable 5 models.

Dan Ives on Anthropic, AI M&A, and OpenAI Losses

⚡ 30-Second TL;DR

🧠 Deep Insight

🔑 Enhanced Key Takeaways

🛠️ Technical Deep Dive

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline

📎 Sources (29)

👉Related Updates

Global Poll: China’s AI Leads in Capability, Lags in Trust

Man Group Warns of Bubble Risks in AI Bonds

Sam Altman Adopts ‘Emotional Fluidity’ Leadership Style

World Cup Highlights Global Rift Over Prediction Markets