Accuracy exceeds 70%! Theory Ventures' 12 ultimate predictions for 2026: AI, IPOs, and stablecoins

Tomasz Tunguz, Founder of Theory Ventures

Translation: Yuliya, PANews

At the end of each year, Tomasz Tunguz, founder of Theory Ventures, systematically reviews his predictions from the past year and offers new insights for the year ahead.

In his review of 2025, Tunguz scored his ten predictions (each out of 1) with an average of 7.85 points. This year, profound structural changes occurred across fields such as artificial intelligence, capital markets, data infrastructure, and crypto finance. These changes not only validated most of his predictions but also laid clear foreshadowing for development trends in 2026. This article will revisit the market upheavals of 2025 and look ahead to the twelve major trends coming in 2026. PANews has compiled and organized these two articles.

2025 Top 10 Predictions Review

Prediction 1: The IPO market will experience an explosion. Score: 0.6

Overall, the IPO market in 2025 saw a significant rebound. A total of 46 software companies went public, raising $12.3 billion, markedly higher than 21 companies and $3.8 billion in 2024, but still far below the peak of tech IPOs in 2021.

CoreWeave and Circle successfully went public, with strong market capitalization performance and post-listing trading; meanwhile, companies like Figma and Chime traded below their previous private valuation, indicating a more rational market valuation judgment. Some high-profile companies such as SpaceX, Stripe, and Databricks did not go public in 2025 but accumulated substantial potential momentum for future years.

Prediction 2: Google will continue expanding in artificial intelligence. Score: 1

Google has reclaimed its top position in AI, leading in nearly all major AI categories. Its Gemini3 model achieved a fundamental leap in pretraining efficiency and multimodal integration. Gemini3 Flash redefined industry benchmarks for performance and latency, becoming the default engine for high-frequency intelligent workflows. In open-source, the Gemma series models consistently top their weight classes, providing inference capabilities comparable to 70B models with only 27B parameters. Even in creative media, Google’s video models rank among the top three globally, with a focus on temporal consistency and role stability, making them highly suitable for enterprise applications.

Prediction 3: Voice will become an important interface for human-AI interaction. Score: 1

OpenAI’s reports show that by October 2025, voice chats via ChatGPT accounted for 19% of user interactions. The number of voice assistants worldwide reached 8.4 billion, with 153 million users in the US. It is expected that 80% of enterprises will integrate AI-driven voice features into their operations by 2026. Using Whisper, WisprFlow for voice input, and conversing with agents like Gemini Live has become routine.

Prediction 4: Total US venture capital investment will stay between $210 billion and $230 billion, but fundraising by VC funds will grow by 20%. Score: 0.5

In venture capital, total US VC investment for the year was about $220 billion, aligning with predictions, mainly driven by mega-scale AI financings. However, fundraising did not grow; instead, it declined by about 20% year-over-year, with total fundraising around $65 billion. Although deal volume rebounded early in the year, prolonged liquidity constraints and a sluggish exit environment kept LPs cautious.

Prediction 5: Integration is the theme of the modern data stack. Score: 1

2025 was a record year for data infrastructure mergers and acquisitions. The “modern data stack” shifted from a collection of “best-in-class” tools to a competition to build vertically integrated platforms.

This wave of integration extended downstream, demonstrating that the current competitive core lies in compute power, computational resources, and integrated software. Notably, CoreWeave’s series of acquisitions marked the rise of “full-stack hyperscale cloud providers,” with complete tech stacks from GPUs to MLOps layers.

Prediction 6: The first company with annual recurring revenue (ARR) exceeding $100 million and fewer than 30 employees will emerge. Score: 1

AI-native teams have redefined business efficiency. Cursor achieved $100 million ARR with only 12 employees in January 2025, while Midjourney reached $500 million ARR with about 100 team members. This efficiency far surpasses traditional SaaS companies, demonstrating the enormous capital efficiency advantage of agent software. For comparison, Slack had 650 employees when reaching $100 million ARR, Ramp had 275, and Wiz had 400.

Prediction 7: After years of decline, as governments embrace cryptocurrency and Web3, the number of Web3 engineers in the US grew by 25%. Score: 1

In 2025, Web3 job openings in the US increased by 26%, reaching 21,600. Significant regulatory shifts sparked a wave of institutional adoption and fostered a new wave of decentralized tech stack-based consumer applications.

Prediction 8: AI competition will drive GPU demand, with data center spending by hyperscale providers exceeding $125 billion annually. Broadcom will become the hottest semiconductor stock of the year. Score: 0.75

Compute infrastructure investments far exceeded expectations. In 2025, hyperscale cloud providers’ capital expenditures reached between $315 billion and $350 billion, with Amazon around $100 billion, Microsoft about $80 billion, and Google approximately $75 billion. In semiconductors, Broadcom’s stock surged due to AI network demands, outperforming even Nvidia in the second half of the year. Despite a year-to-date increase ranking third industry-wide behind Micron and Google, it remained impressive.

Prediction 9: Stablecoin supply will grow by 50% to $300 billion, with trading volume over three times that of Visa. Score: 1

Stablecoins are becoming a key component of global payments. By December 2025, total stablecoin supply reached $310 billion, with on-chain annual trading volume exceeding $46 trillion, nearly three times Visa’s transaction volume. As enterprises seek faster, cheaper cross-border settlement methods, adoption of stablecoins in B2B payments is accelerating.

Prediction 10: Observability, SIEM, and business intelligence will begin sharing the same data lake. Score: 0

This prediction did not materialize. Although usage-based pricing models have indeed driven demand for unified data lakes, and data lake architectures are increasingly dominant across workloads, the idea of observability, Security Information and Event Management (SIEM), and Business Intelligence (BI) sharing a single data lake did not become a reality in 2025.

Twelve Major Predictions for 2026

Based on observations from 2025, predictions for 2026 are more systematic and in-depth, with the core logic being: AI is evolving from an auxiliary tool to autonomous systems, shifting from cutting-edge experiments to core infrastructure. 2026 will be the year when enterprises large-scale deploy AI into actual production.

First, enterprise spending on AI agents will surpass spending on human labor for the first time. Consumer-side has already seen this phenomenon, e.g., Waymo’s autonomous driving costs are higher than Uber’s, yet demand continues to grow. After considering recruitment, training, and management costs, enterprises will accept premium pricing for agents handling repetitive tasks.

Second, 2026 will be a record year for liquidity. Companies like SpaceX, OpenAI, Anthropic, Stripe, and Databricks are expected to go public collectively, with SpaceX and OpenAI’s IPOs potentially ranking among the top ten in history. Meanwhile, facing disruptive threats from AI, traditional companies will launch over $25 billion in defensive M&A, choosing to “buy” rather than “build” AI capabilities.

Third, vector databases will re-emerge as critical infrastructure in the AI tech stack. Multimodal and world models demand new data structures, and as the central hub connecting foundational models with enterprise data, revenue from vector databases will explode.

Fourth, the time AI models can autonomously perform tasks will exceed a full workday. According to METR data, AI task durations double every 7 months. Following this trend, by the end of 2026, AI agents will be able to continuously complete workflows over 8 hours, fundamentally changing project configuration methods.

Fifth, AI budgets will undergo systematic review for the first time. Boards and procurement committees will question AI spending, with small models and open-source solutions gaining popularity due to cost advantages. Research teams will specialize tasks, achieving performance comparable to cutting-edge models at lower costs, enabling developers to reduce costs by orders of magnitude.

Sixth, Google will further widen its lead over competitors through extensive and deep AI deployment. Breakthroughs across multiple areas—frontier models, edge inference, video generation, open-source weights, and search integration—will force companies like OpenAI, Anthropic, and xAI to focus on niche segments, ending the era of full-spectrum competition.

Seventh, agent observability will become the most fiercely contested layer in the reasoning stack. As AI agents penetrate enterprise operations, traditional engineering, security, and data observability will merge into a unified discipline. Enterprises will need end-to-end monitoring of AI code execution, security threats, and data lineage. The three observability domains forecasted in 2025 will find their convergence here.

Eighth, by December, 30% of international payments will be settled via stablecoins. The efficiency advantage of cross-border settlement is so significant that, as major markets clarify regulations, stablecoins will move from crypto periphery to the core of global trade finance, replacing some B2B transactions under SWIFT.

Ninth, AI agent data access patterns will overwhelm existing database architectures. Their query frequency and concurrency demands will be at least an order of magnitude higher than humans, forcing a fundamental overhaul of transactional and analytical databases to meet ongoing automation needs.

Tenth, data center construction will reach 3.5% of US GDP in 2026. To support exponential AI compute demands, data center investments will hit historic levels, comparable to the railway expansion of the past. The only potential slowdown could come from credit market risks, especially rising default rates in private credit, which may become a bottleneck for these highly capital-intensive projects.

Eleventh, the internet will shift toward “agent-first” design. Future websites and developer documentation will primarily be designed and optimized for AI agents rather than humans. As more business procurement decisions are made based on AI-driven information gathering and comparison, websites will need to open “front doors” for robots, reserving “side doors” for human visitors.

Twelfth, Cloudflare will become a key hub for agent payments. The x402 protocol will reactivate the long-idle HTTP 402 “Payment Required” status code, enabling AI agents to pay for API access in real-time. With its position in network infrastructure, Cloudflare will naturally become the gateway for this new business model, sparking a new round of debates on openness versus centralization.