New Power Dynamics in the AI Industry 2025: Anthropic vs OpenAI, Chinese Models and Cyberthreats for Business

Introduction: The AI Industry Is No Longer What It Was a Year Ago

The power dynamics in the AI industry have shifted dramatically — and this directly impacts the decisions you're making for your business today. Just 12 months ago, the picture seemed stable: OpenAI — the undisputed leader, Google — a powerful contender, everyone else — far behind. But 2025 has turned these assumptions upside down. Anthropic has for the first time surpassed OpenAI in monthly recurring revenue. Google is experiencing a significant crisis — the departure of key researchers and delays in releases are damaging the company's reputation. Chinese developers are offering models that match top players in quality but cost several times cheaper. And at the same time, the world's first fully autonomous AI-ransomware attack has emerged — an alarming signal for the entire market. If you manage a small or medium-sized business and use or plan to use AI solutions, this article is your guide to the new reality.

---

Anthropic Surpassed OpenAI: What It Means for the AI Tools Market

How It Happened and Why It Matters

The first quarter of 2025 brought a sensation: Anthropic reported monthly recurring revenue that exceeded OpenAI's equivalent metric. This is not just statistics — it's a signal of a fundamental shift in corporate clients' trust. Companies choosing an AI platform for critical business processes are increasingly stopping at Claude from Anthropic — primarily due to the model's predictable behavior, better handling of long documents, and clearer corporate security policies.

For business owners, this means one thing: the AI market is no longer monopolized. Competition between suppliers is your leverage in negotiations over price and terms.

Read more about Anthropic's competitive positions in our article: Anthropic surpassed OpenAI in revenue: what it means for business in Ukraine

What Changed in Anthropic's Product Strategy

Anthropic is betting on corporate security and transparency. Their "Constitutional AI" approach — where the model is trained to adhere to a clear set of principles — has proven extremely attractive to industries with strict compliance requirements: finance, law, medicine. Additionally, Anthropic is actively investing in expanding the context window (up to 200,000 tokens in Claude 3.5), making the model ideal for analyzing large volumes of documents — contracts, reports, technical specifications.

Meanwhile, OpenAI is not giving up: the GPT-5.6 line with models Sol, Terra and Luna deliberately attacks different market segments. Details about this — in the article GPT-5.6 (Sol, Terra, Luna): what the new OpenAI means for your business in Ukraine

---

Google's Crisis and Researcher Exodus: Should You Bet on Gemini Now

What's Happening Inside Google DeepMind

Google faces problems that seemed impossible for such a giant just a year ago. According to industry sources, in 2025 the company lost a significant number of leading AI researchers — some left for startups, some went to Anthropic and xAI. Delays in releasing new versions of Gemini, as well as high-profile failures during public demonstrations, have dealt a serious blow to market capitalization: the company lost over $225 billion in value over a relatively short period.

Notably, among those who moved to Anthropic is John Jumper, a Nobel laureate and one of the creators of AlphaFold. This is not just a personnel change — it's an indicator of where the best scientific talent in the industry is concentrated. Read more: Nobel laureate John Jumper at Anthropic: what it means for business and AI automation

Practical Conclusion for Business

Does this mean you should abandon Google Workspace and integrations with Gemini? Not necessarily. Google maintains powerful infrastructure and deep integration with its cloud services. However, if you plan to build critical AI processes based on Gemini, you should:

• Have a backup plan — document dependencies and keep an alternative supplier "on standby"
• Don't lock your entire stack to one provider — a multi-model approach reduces operational risks
• Monitor updates — Google has the resources to quickly recover its position, as it has done before
• Evaluate not just the model, but the ecosystem — BigQuery, Google Cloud, Vertex AI remain competitive

More details on Google's situation — in our article: A tough week for Google: Gemini crisis, researcher exodus and $225 billion capitalization loss

---

Chinese AI Models and xAI: Bottom-Up Competition You Can't Ignore

GLM-5.2 and Other Chinese Developments: Real Threat or Hype

While leading American labs battle each other, Chinese AI models quietly and methodically capture market segments where price matters. GLM-5.2 from Zhipu AI is one of the most striking examples. The model demonstrates results approaching GPT-4-class performance on a number of standard benchmarks, but costs 3-5 times cheaper per million tokens.

For small and medium-sized businesses where AI budgets are limited, this can be a significant argument. But there are nuances:

• Data sovereignty: Chinese models may fall under Chinese law regarding data storage and processing
• Compliance: for some industries (finance, government procurement) using such models may violate regulatory requirements
• Support: the ecosystem of tools and documentation is mostly in English or Chinese, complicating implementation

Detailed analysis of Chinese developers' competitive positions in the corporate market: Chinese AI models capture US corporate market: what it means for your business

xAI and Grok: An Unexpected Contender

xAI — Elon Musk's company — is also actively fighting for market share. Grok 4.5 is positioned as a model with minimal censorship and maximum response speed. For some business applications requiring high throughput and lower latency (for example, real-time customer support), this could be a justified choice. Everything about Grok's capabilities for Ukrainian business — in the article Grok 4.5 from xAI: what the new model means for business in Ukraine

The Chip Race as the Foundation for Price Changes

Behind the decline in token prices is not just competition between labs, but a real inference chip arms race. NVIDIA, Broadcom and OpenAI itself are investing billions in their own specialized processors for AI computing. The result for business — gradual reduction in API access costs to models, making automation more accessible to companies of any size. Technical context of this race: The inference chip race accelerates: OpenAI, Broadcom and NVIDIA change the rules for business

---

First Autonomous AI-Ransomware and Regulatory Shifts: Security and Regulation Can't Keep Up

JADEPUFFER: A New Level of Cyberthreats

Among all the news of 2025, one stands apart — and it should force every executive to reconsider their cybersecurity approach. JADEPUFFER — the first documented fully autonomous AI-ransomware attack. Unlike traditional ransomware programs that require human operator involvement at key stages, JADEPUFFER independently:

• Scans the network perimeter and finds vulnerabilities
• Develops personalized phishing messages for specific employees
• Chooses the optimal moment for activation and data encryption
• Adapts tactics in real-time depending on the response of defense systems

This is a fundamentally new level of threat. Traditional antivirus and even modern EDR solutions were not designed for an opponent that "thinks" and adapts. For small and medium-sized businesses, where there is often no dedicated security department, this is particularly dangerous.

Detailed breakdown of the attack and specific protection steps: JADEPUFFER: the first autonomous AI-ransomware attack and how to protect your business

What You Need to Do Right Now

• Access audit: review who has access to which systems — AI attacks often start with compromised credentials of rank-and-file employees
• Backup according to the 3-2-1 rule: three copies, two different media, one — outside the network
• Employee training: phishing generated by AI is much more convincing — traditional security training needs updating
• Network segmentation: limit lateral movement within infrastructure
• AI protection against AI attacks: consider machine learning-based solutions for monitoring abnormal behavior

Regulatory Response: Frontier Models Executive Order

In response to growing risks, the US administration signed an executive order requiring developers of frontier AI models to provide the government early access to new systems before their public release. This is the first systemic step toward regulating the most powerful AI systems, but critics already point out an obvious problem: the security assessment process takes months, while the development cycle for new models is shrinking to weeks.

For business, this means potential delays in accessing the latest capabilities, but at the same time — greater predictability and reduced regulatory risks when using certified solutions. Full analysis of regulatory implications: Trump's Executive Order on Frontier AI Models: what new AI regulation means for business

---

How Business Can Navigate the New Power Dynamics in the AI Industry

Strategy for Choosing an AI Platform in Times of Turbulence

The new power dynamics in the AI industry requires business to have not just technical competence, but strategic thinking. Here's a practical framework for decision-making:

1. Evaluate not the model, but the ecosystem. The capabilities of a specific LLM change monthly. It's more important to assess the reliability of the supplier's infrastructure, API quality, documentation, support and roadmap.

2. Diversify dependencies. Don't build critical processes on a single AI provider. Architecture with the ability to switch between models (for example, through a unified API layer) is insurance against risks.

3. Count the total cost, not just the token price. A cheap model with poor documentation and unstable API can cost much more to implement and maintain.

4. Invest in security in parallel with automation. Expanding AI use without corresponding strengthening of cybersecurity is like building a house without a foundation.

5. Monitor the regulatory environment. For industries with strict compliance requirements (medicine, finance, law), the choice of AI supplier should take into account not only technical characteristics, but also jurisdiction and certifications.

Practical Entry Points for SMB

If you're just starting out or scaling AI usage — start with automating repetitive customer communications. AI agents for processing incoming requests, lead qualification and customer support deliver measurable ROI within 4-8 weeks of implementation. More about practical applications — in the article AI automation for medium business: how to grow without increasing headcount

---

FAQ: Common Questions About Changes in the AI Industry and Their Impact on Business

1. Should I switch from OpenAI to Anthropic now because of news about the change in revenue leader? You shouldn't make decisions solely based on companies' market positions — both platforms remain powerful and are developing. Evaluate specific technical characteristics, price and support for your use case, not companies' financial ratings.

2. How dangerous are Chinese AI models for corporate use? The main risk is not model quality, but issues of data sovereignty and compliance with regulatory requirements. Before using GLM or analogues, consult a lawyer about processing customer personal data and compliance with Ukrainian and Euro-Atlantic legislation.

3. How do I protect my business from autonomous AI attacks like JADEPUFFER? Comprehensive protection includes network segmentation, regular offline backups, training employees to recognize AI phishing, and implementing anomaly monitoring systems. No single solution provides 100% protection — a layered approach is needed.

4. Will the decrease in AI token prices lead to significant savings for small business? Yes, but not immediately. While the chip race unfolds at the data center level, cost reductions gradually translate into lower API prices. Already now, the cost of automating routine tasks through AI has dropped 30-60% compared to 2023, and this trend will continue.

5. What does the executive order on frontier AI model regulation mean for Ukrainian business? The executive order does not directly obligate Ukrainian companies, but it affects the timeline for new capabilities from American providers. Indirectly, this may slow access to the latest features, but will make using top models more predictable in terms of security.

---

Conclusion

The new power dynamics in the AI industry in 2025 is not temporary turbulence, but a new permanent reality where leaders change faster than ever before, and risks grow alongside opportunities. For small and medium-sized business, this is both a challenge and an opportunity: those who learn to navigate this environment will gain a competitive advantage over those waiting for "stability." Want to understand what AI strategy is optimal for your business in the current conditions? Contact our experts for a free consultation. We'll help you choose the right tools, build a secure architecture, and launch your first automations within a few weeks.