OpenAI Breaks Funding Record, U.S. Clashes with Anthropic, and AI Agents Enter ‘Heavy Industry’

OpenAI Breaks Funding Record, U.S. Clashes with Anthropic, and AI Agents Enter ‘Heavy Industry’

In February 2026, the artificial intelligence market entered the ‘heavy industry’ phase^[1]. OpenAI announced a funding round worth 110 billion USD^[2], the largest in the history of private technology companies. According to analyses by AP and New York Times, the company valuation reaches about 730 billion USD before the round^[3] and up to 840 billion USD after its closure. The funding includes approximately 50 billion USD from Amazon^[4] (15 billion USD upfront and 35 billion USD conditional), Nvidia with 30 billion USD, and Soft. Bank contributing another 30 billion USD. OpenAI plans to spend about 600 billion USD on computing power by 2030^[5], which is already impacting the global energy, chip, and data center markets^[6], and raising costs of models and agents also for companies in Poland^[7].

Record Funding and Capital Concentration

The record funding of OpenAI fits into a broader picture of capital concentration^[8]. According to a Crunchbase News report, February 2026 brought 189 billion USD in global venture capital investments^[9], of which about 171 billion USD, or 90%, went to AI startups^[10]. Three deals — OpenAI (110 billion USD), Anthropic (30 billion USD), and Waymo (16 billion USD) — captured 83% of the total capital, amounting to 156 billion USD^[11]. Such concentration of funds in a narrow group of entities building infrastructure and agent systems deepens the divide between ‘AI giants’ and the rest of the market^[13] and pushes up prices for advanced GPUs and HBM memory^[12], which also hits resource availability for Polish projects.

U.S. Open Confrontation with Anthropic

Simultaneously, the U. S. presidential administration and U. S. Department of Defense^[14] entered an open dispute with Anthropic. The White House announced plans to blacklist the company^[15] as a supply chain risk, and The Pentagon declared intentions to terminate a contract potentially worth 200 million USD^[16] after Anthropic refused to lift restrictions on using the Claude model^[17] for mass surveillance and fully autonomous weapons. Claude has so far been the only major language model^[19] used in highly classified U. S. military systems, including operations against Nicolás Maduro. The blacklist decision allows a six-month transition period to migrate to alternative suppliers^[18], such as OpenAI, Google, or xAI, setting a precedent for state relations with AI providers in military applications.

Agents as Digital Workers

On the commercial market, AI agents began moving beyond experimental phases^[20]. OpenAI launched the Frontier Alliances program^[21] with consulting firms like Mc. Kinsey, Boston Consulting Group, Accenture, and Capgemini. These global companies are set to sell and implement the Frontier agent platform^[22] as a semantic layer over enterprise systems, covering sales, customer service, and IT^[23]. Perplexity AI introduced the ‘Computer’ service^[24], available from February 25, 2026, for Max plan subscribers, which breaks down complex business goals into tasks performed by parallel agents^[25] using about 19 models collectively^[26], including Claude Opus 4.6 as the master model and systems like Gemini, ChatGPT 5.2, Grok, Nano Banana, and Veo 3.1. Meanwhile, Service. Now is developing the Autonomous Workforce line^[27]—specialist agents such as Level 1 IT Service Desk AI, which according to internal data resolves 90% of employee IT tickets^[28] and operates 99% faster than a human, with planned general availability in Q2 2026.

Hardware and Industries for AI Agents

Concrete adoption metrics for agents were also provided by Salesforce^[29], which in its fiscal 2026 Q4 results emphasized growth in the Agentforce business. Annual recurring revenue (ARR) for Agentforce reached 800 million USD^[30], marking a 169% year-over-year increase^[31]. The combined ARR for Agentforce and Data 360 was 2.9 billion USD^[32], while the number of Agentforce contracts hit 29,000^[33], a 50% quarter-over-quarter rise^[34]. The system has processed about 19 trillion tokens since inception^[35] and delivered 2.4 billion agentic work units^[36]. Concurrently, Apple unveiled Mac. Book Pro laptops^[37] with M5 Pro and M5 Max chips designed for local language models. According to the company, the new chips accelerate LLM prompt processing up to 6.9 times^[38] compared to M1 Pro and 3.9 times versus M4 Pro, and provide up to 8 times faster image generation than M1 Max^[49]. The computers offer up to 64 GB of unified memory in M5 Pro version^[39] and up to 128 GB in M5 Max with bandwidth up to 614 GB/s, enabling running advanced agents locally without cloud data transmission.

New Deepfake Regulations and Market Impact

Amdocs and Google Cloud announced collaboration on ‘AI-first’ solutions^[40] for operators. The package includes preconfigured telco agents^[41] based on Amdocs Cognitive Core and Gemini Enterprise for CX, ready to handle customer service, billing, or sales within existing BSS/OSS systems. For Polish operators, this means quick access to ready-made agent scenarios^[42] via global supplier rollouts. At the regulatory level in the U. S., a key deadline approaches^[47] stemming from the federal ‘TAKE IT DOWN Act’. The law, signed May 19, 2025, requires platforms to implement effective notice-and-takedown procedures^[43] for non-consensual intimate content, including AI-generated, by May 19, 2026, with mandates to remove such content within 48 hours and keep records of reports and appeals. Meanwhile, 28 U. S. states have passed their own regulations on political deepfakes^[44], and 169 state laws regulating various types of deepfakes have been enacted^[45], including projects like Utah’s “Voyeurism Prevention Act” which requires consent for generating intimate images using AI^[48]. This necessitates adapting content moderation and governance of agents^[46] to diverse, often strict legal requirements.

Related posts:

• Breakthrough Week in AI: China’s Offensive, Agent Platforms, and Security Crises
• The Era of AI Agents Accelerates: Chinese Models, GLM-5, EU AI Grid, and Initial Security Incidents
• LLMs and AI Agents Accelerate: Record Contracts, New Models, and Regulatory Pressure
• NVIDIA, New Regulations, and a Wave of Funding: The Week AI Agents Enter the Mainstream