Interest in AI continues to grow throughout the current year. The main areas of focus today are large language models, data centres, chips and AI-based business solutions. Quarterly dynamics have shown positive trends, including new funds, mega rounds and the attraction of corporate investors. By 2026, AI is expected to become a key component of the entire innovation ecosystem.

Current status? Record investments

Total volume and key figures

To understand the scale, it is sufficient to examine the structure of the deals. By the end of the third quarter of 2025, more than 4,400 investment deals with AI startups had been recorded, 120 of which were valued at over $100 million.

Geographically, the leaders remain the US (accounting for more than 60% of the total volume), China (15%) and Europe (around 12%). Meanwhile, the EU is actively supporting AI through state funds in an attempt to reduce its dependence on American players. The largest deals of the year include: OpenAI — $40 billion; Anthropic — $13 billion; and Mistral AI — $2.5 billion. All of these companies focus on developing large language models.

Investors have also become more attentive to efficiency. While growth was the main focus in 2022, today, funds are more focused on unit economics, model performance, and controlling inference costs. This means that 2025 will not only be a record year in terms of numbers but also a more mature one in terms of approaches to capital expenditure.

The share of total venture capital held by AI

By 2025, AI had effectively become the core of the venture market. Its share of total venture capital grew to 23%, up from 8-9% three years ago. This is an unprecedented leap.

Funds are changing their strategy, both thematically and structurally. Many are creating separate AI-focused funds or special AI opportunity funds for later rounds. These funds enable a rapid response to breakthrough developments requiring substantial capital.

Another sign of market maturity is the emergence of investments in multi-industry AI solutions, such as models capable of processing different types of data, including text, video and biomedical signals. This expands the market and attracts institutional investors who value risk diversification.

Even traditional venture capital players focused on hardware or biotech have begun to integrate AI components into their investment strategies, which is telling. In other words, it is now difficult to find an area of investment decision-making that AI does not influence.

Quarterly dynamics and mega rounds

Taking a quarterly view, 2025 appears to be a steady period of growth. The first quarter saw a breakthrough, driven by activity in the US and the emergence of several new funds that focus exclusively on generative AI. The second quarter built on this momentum, with the Anthropic and Mistral AI deals setting the market standard.

The third quarter saw a slight slowdown as investors digested the previous mega rounds. However, rather than a decline in activity, we witnessed a shift in focus towards vertical solutions, such as medical diagnostics, cybersecurity, and autonomous systems. This period saw the most early-stage deals — more than 1,200.

The fourth quarter of 2025 will mark a culmination, with new players entering the market from Europe and the Middle East. Notable newcomers include the Middle East Fund and the Saudi Future Tech Fund, which have collectively announced plans to invest over $10 billion between 2026 and 2028.

Mega rounds are no longer seen as exceptional — they are becoming the new norm. Investors have realised that building competitive AI models requires enormous resources, so large investments are becoming a strategic tool.

If this trend continues, 2026 could not only see record deals but also a structural redistribution of venture capital. Artificial intelligence would finally transform from a separate sector into the foundation of the entire investment ecosystem.

Investment geography: regional leaders and trends

Sectors and technologies within AI — where capital is flowing

In 2025, the largest amount of capital — more than $50 billion — was directed towards infrastructure. Infrastructure provides the main resource — computing power. Startups that optimise the operation of data centres, GPU clusters, and energy consumption are now receiving financing with the same level of enthusiasm as software giants did in the 2010s. Venture capital funds understand that control over computing power means control over the entire ecosystem.

Next up are generative AI and large language models. These accounted for over 40% of venture deals in the AI sector. Startups are not only creating new LLMs, but also agent systems that can operate autonomously, learn from experience and perform complex business tasks. They are shaping the new AI labour market.

Another area is the development of practical solutions for various industries. AI is being actively integrated into sectors such as healthcare, finance, manufacturing, logistics and education. Here, investors are looking for practical efficiency: how AI can save money, reduce risks or generate new revenue models.

Clearly, investors are establishing a comprehensive vertical structure, from basic calculations to specific solutions. This structure indicates that the AI market is maturing, moving beyond hype to focus on real innovation.

Infrastructure and computing resources

Investments in artificial intelligence infrastructure became a key area for venture capital in 2025. Without a stable foundation of powerful chips, data centres and tools for training models, no AI company can grow. This is why venture capital funds are increasingly financing not only startups with smart products, but also those who create the foundation for them — computing solutions optimised for AI.

Demand for graphics processors and specialised AI chips has reached unprecedented levels. Leading companies such as Nvidia, AMD, and Tenstorrent have attracted billions of dollars in funding, and a new wave of startups specialising in everything from energy-efficient solutions to alternative architectures for inference and model training has emerged around them. Companies such as CoreWeave and Lambda Labs have secured significant funding precisely due to the market's capacity shortage.

A second powerful stream of investment has gone into data centres and cloud infrastructure, particularly in the AI-ready segment. Funds are investing in projects that optimise energy use, cooling and computing distribution. Capital is also being actively directed towards creating platforms for model management, GPU orchestration, and automating the training process.

The trend is clear: investors are no longer just looking for flashy AI products. They are seeking control over the basic level of the ecosystem because infrastructure determines who will be able to withstand the scaling race. According to forecasts, this segment will receive over $60 billion in new investments by 2026, making competition for access to computing resources a key strategic factor in the world of AI.

Generative AI, LLMs and agents

Generative AI has become emblematic of the current era and the primary focus of venture capital in 2025. It accounted for over 40% of all artificial intelligence investments. Startups working with large language models (LLMs) and agent systems have become the new centre of gravity for innovation.

The main players — OpenAI, Anthropic, Mistral AI and Cohere — have attracted funding rounds totalling hundreds of millions to several billion dollars. However, the most interesting developments are happening further down the chain, among startups that create specialised agents for specific professions, such as finance, medicine, education, and marketing. These agents don't just respond to requests; they perform tasks autonomously and integrate themselves into companies' workflows.

Investors see revolutionary potential here. LLMs are no longer just toys; they are becoming a new type of workforce — digital employees who work alongside humans. This has a significant economic impact, particularly in SaaS, analytics, and customer service.

Investors are paying particular attention to multimodal models that combine text, images, audio, and video. Demand for content and personalised services is driving rapid development in this area. A new category of startups is also emerging: 'AI agents as infrastructure', which enables companies to develop their own agents without requiring in-depth technical expertise.

AI in healthcare, fintech, and the manufacturing industry

By 2025, AI had moved beyond laboratories and had become a key element in applied industries. Most venture capital went to medicine, fintech, and manufacturing industry sectors in which the introduction of artificial intelligence has a direct economic impact on real-life situations.

In medicine, for example, AI has transformed the way diseases are diagnosed and treated. Startups such as Insilico Medicine, PathAI and Tempus have attracted significant investment thanks to their systems for predicting diseases, analysing medical images and creating personalised treatment plans. Investors see enormous potential in reducing the cost of medical research and improving the accuracy of doctors' decisions.

Fintech has become the second magnet for capital. Here, AI is used for risk management, customer analysis, fraud detection and the automation of regulatory processes (RegTech). Companies such as Stripe, Plaid, and Unit21 demonstrate that integrating AI into financial services can boost profitability and cut operating costs.

In the manufacturing industry, AI is being implemented through robotisation, computer vision and predictive maintenance models for equipment. This helps companies to avoid downtime and optimise costs. Investors are particularly active in smart factories and supply chain automation.

The overall trend is clear: venture capital is shifting from experimentation to achieving tangible results. Funds are seeking out startups that can implement AI in a specific industry and create measurable value. By 2026, these areas will be driving the stable growth of the entire AI ecosystem.

Key venture capital funds and AI investors in 2025–26

Artificial intelligence will continue to attract venture capital in 2025-2026. However, unlike in 2021 when investments were made indiscriminately, the focus has now shifted towards the quality of technologies and teams.

Key investors include Andreessen Horowitz, Sequoia Capital, Index Ventures, Lightspeed Venture Partners and Khosla Ventures. These funds actively support startups that create critical infrastructure, such as models, data centres and cloud services for training LLMs.

Nvidia's Inception Programme deserves a special mention as it has evolved into a comprehensive venture ecosystem hub for hundreds of AI companies, offering not only a partnership platform but also valuable support and resources. In Europe, Atomico, Lakestar and EQT Ventures have become more active, while in Asia, the SoftBank Vision Fund, Tencent Investments and state funds from Singapore and the UAE are financing infrastructure for national AI models.

In the corporate sector, venture divisions such as GV (Google), M12 (Microsoft), Alexa Fund (Amazon) and OpenAI Startup Fund are shaping a new market dynamic by integrating startups into their ecosystems. This creates a vicious circle where the best resources go to startups close to the giants.

The main trend for 2026 is the professionalisation of investors in the AI sector. Funds are forming specialised teams to analyse models, energy costs, and scaling risks. Investment is becoming more selective, and competition for quality deals is intensifying.

Barriers, risks, and the AI bubble

Despite record investment volumes, the artificial intelligence venture market will face increasingly significant barriers in 2025–26.

The first of these is the high cost of computing. Training a large language model (LLM) can cost tens of millions of dollars, which is out of reach for most startups. Access to Nvidia's computing power and cloud technologies from AWS, Google Cloud and Azure is becoming a strategic resource controlled by a select few. This creates an asymmetry whereby innovation is concentrated in a limited number of companies.

The second barrier is the lack of high-quality data. Startups face legal and ethical restrictions on using data for model training. Lawsuits against Stability AI, OpenAI and Midjourney regarding copyright have demonstrated that regulatory risks can significantly increase the cost of capital.

Additionally, there is a risk of an 'AI bubble'. Investors are increasingly drawing parallels with the dot-com boom of the 2000s. Some companies with 'AI' in their name receive inflated valuations despite having no proven business model. According to PitchBook, over 40% of AI startups at the seed stage have no real revenue. While this does not necessarily lead to collapse, it does signal an upcoming market consolidation.

Another risk is dependence on cloud platforms and large model APIs. Most young companies build products on top of GPT, Claude or Gemini, thereby losing control over margins and technological independence.

High levels of competition and capital concentration

Today's AI market is a battleground for the industry's biggest players. By 2025–2026, venture capital will be concentrated in a few key locations: the United States, China, the United Kingdom, Israel and the United Arab Emirates. This is where the large deals, corporate partnerships and funds that dictate terms will be concentrated.

Competition among startups is becoming almost ruthless. Every niche, from generative design to medical analysis, has dozens of similar projects. This is driving up the cost of user acquisition and reducing margins. Investors are now looking not just for AI products, but also for teams that can monetise and integrate them into specific business processes.

Capital is also becoming concentrated within the funds themselves, with a few top players taking most of the rounds. This creates a 'winner takes most' effect, whereby the best startups receive excessive funding, while hundreds of others are excluded from the market.

Furthermore, the AI market is becoming increasingly dependent on strategic partnerships with major tech companies. Microsoft, Google, Amazon, and Meta are actively buying up shares in startups to form ecosystems around themselves. While this accelerates technological development, it also limits the independence of innovation.

The main task for venture investors is to find unique niches where competition has not yet turned into a price war. These could be vertical AI solutions for industries such as energy, education, and agriculture — this is where the next generation of unicorns will emerge in 2026.

Valuation risk

From 2021 to 2024, ambitious startups with big ideas could receive high valuations even in the early stages. Now, however, investors demand proof that the technology generates real money. A company's valuation depends on three things: monetisation metrics (MRR/ARR), long-term contracts and barriers to entry, such as unique data, patented algorithms and infrastructure. If any of these elements are weak, the valuation will be revised downwards.

There are several dimensions to the risk of overvaluation:

• Business risk. Startups without real revenue have limited time before investors demand cost cuts or a sale.
• Technological risk: Many solutions are based on the APIs of large external models, which creates an economic vulnerability when the cost of inference increases or the supplier raises the price.
• Market risk: A prolonged slowdown in hiring, coupled with a decline in investment in general technology, reduces demand and puts pressure on valuations.

So, what should startups do? They should focus on unit economics, prove LTV/CAC, build a contract base with corporate clients and create technical barriers to entry, such as inference optimisation and unique datasets.

Regulation, ethics, and security

Regulators influence the ecosystem, causing it to slow down or speed up. There are three levels of influence: legal (e.g. legislation regarding data, copyright, and liability); ethical (e.g. transparency, bias, and explainability); and security (e.g. protection against abuse and model cybersecurity). When a regulator requires audit logs or explainability, this incurs additional development and audit costs, as well as delaying the launch of the product.

Ethical risks are important not only from a moral standpoint, but also because they affect investment attractiveness. Systems capable of generating deepfakes or making autonomous decisions in high-risk areas such as medicine, finance, and justice become less attractive in the absence of clear policies for control and bias testing. Investors demand responsible use policies, penetration tests and independent audits.

Another dimension is national security and geopolitics. Countries may ban the export or use of certain models, block access to cloud services, or require data localisation. This creates the risk of market fragmentation, so startups must have a plan for operating locally if global access is restricted.

Technological limitations and infrastructure challenges

Technological limitations comprise economic, physical, and engineering issues that impact the ability to scale solutions. Training large models consumes a lot of energy and incurs significant operating costs. In the long term, profitability depends on the efficiency of inference and model optimisation.

Another challenge is the scarcity and centralisation of hardware. When access to key resources is controlled by a few large suppliers, this increases the barrier to entry for new market entrants and creates the risk of supplier shock.

Furthermore, not all companies have sufficient amounts of high-quality, ethical and legal data with which to train competitive models. Collecting, labelling and protecting data are costly and time-consuming processes.

Forecasts for 2026

2026 will be a year of discipline and market organisation. Three parallel trends are expected. Firstly, the market is expected to consolidate, with weak startups without revenue or technical barriers merging or being acquired. Secondly, infrastructure will attract investment in its own right: solutions that reduce inference costs or provide an alternative to centralised GPU clusters will be prioritised. Thirdly, demand will grow for specialised solutions with proven economic value in sectors such as medicine, energy, and industry.

The role of responsible AI will be strengthened by regulatory and political factors: companies with transparent data policies and audits will receive better financing terms. Technically, those who invest in efficiency will succeed. Economically, business models with recurring revenue and corporate contracts will prevail.

For investors, 2026 will be a time for analytics, requiring thorough technical due diligence, verification of unit economics, scenario modelling and stress testing of infrastructure costs.