Key takeaways

• AI startups are reaching revenue milestones faster than ever.
• A new,AI-native social media giant could emerge.
• Startups must be ready for lots of acquisition interest 

In itsannual overview of artificial intelligence (AI),venture capital firm Bessemer Venture Partners on Wednesday said that startups it funds -- such as Anthropic,Perplexity,and Canva -- are reaching meaningful turning points in their revenue faster than at any other time in the history of the firm's financing of startups.

"AI-native businesses are scaling from$0 to$100 million in ARR [Annual Recurring Revenue] faster than any other companies in cloud history," write lead author Kent Bennettandteam inan accompanying essay . 

Also: Will AI replace all software? Why GPT-5 emboldens the doomsayers

"It's the highest number of hyper-growth startups we have ever seen."

AI supernovasandshooting stars

The firm, which has spent$1 billion on startups so far to fulfill a planannounced two years ago , categorizes startups into three classes according to their speed to revenue: supernovas, shooting stars, and cloud centaur. Supernovas are companies that have reached, on average,$40 million in annual revenue after one year in business.

Also: Don't fall for AI-powered disinformation attacks online - here's how to stay sharp

The Bessemer team does not cite any revenue data for individual companies. According to an AI model run by FactSet Systems, four-year-old Anthropic is estimated to have$53 million in annual revenue. According to the same model, Perplexity, founded in 2022, may be generating$10 million a year. Canva's last disclosed revenue count was for$1.7 billion in 2022, according to FactSet. The company was founded in 2012.

Bessemer has a long track record of backing some of the most successful tech firms, including Twilio, LinkedIn, Yelp, Pinterest, and Wix. (See the full list here .)

AI state of play

The report summarizes AI into five categories: infrastructure, development tools, enterprise AI, vertical-market AI, and consumer AI.  

The big platforms, provided by OpenAI, Google, and other giants, mean that "a new infrastructure layer has emerged --spanning models, compute, training frameworks, orchestration, and observability," they write. 

Also: Why GPT-5's rocky rollout is the reality check we needed on superintelligence hype

But a "second act" is unfolding, they argue, wherein the AI giants move beyond benchmark tests to "building systems that define, measure, and solve problems with experience, clarity, and purpose," including establishing the connections from AI models to systems for "knowledge retrieval, memory, planning, and inference optimization."

In the dev tools market, the Model Context Protocol (MCP) created by Anthropic is establishing itself as an industry standard, a kind of USB-C to hook things together in AI. 

"For developers, MCP radically simplifies integration. For founders, it opens the door to building truly agentic products -- where AI doesn't just assist users, but acts on their behalf across systems," they write. 

An important element going forward will be how tools incorporate memory and storage, moving beyond early efforts such as Retrieval Augmented Generation (RAG). Bennett and team relate that "While the foundational model companies are working on memory, so too are startups likemem0 , Zep , SuperMemory , and LangMem byLangchain ."

Also: RAG can make AI models riskier and less reliable, new research shows

In enterprise AI, traditional "system-of-record" software such as Salesforce is under attack because AI can make it easier to move away from that software by lowering switching costs, the authors argue.

"For decades, SoRs like Salesforce, SAP, Oracle, and ServiceNow held firm thanks to their deep product surfaces, implementation complexity, and centrality to business-critical data. "With AI's ability to structure unstructured data and generate code on demand, migrating to a new system is faster, cheaper, and more feasible than ever."

That observation is especially interesting given how much concern there is nowadays that AI may replace traditional commercial software. 

For vertical markets, AI will succeed where broad-based software failed to capture the complexity of individual industries and practices. 

Also: This is the fastest local AI I've tried, and it's not even close - how to get it

"SaaS failed to solve high-value vertical-specific tasks that were multi-modal or language heavy," they write. "Vertical AI is finally meeting these users where they are, with products that feel less like software and more like real leverage."

The authors believe the payoff from AI's vertical market expertise is strong for buyers. "ROI is clear from day one and there's no Excel spreadsheet needed to explain it to the user," they write. "These tools unlock 10x productivity, reallocate labor to higher-value work, reduce costs, or drive topline growth. The value is immediate, not a "nice to have."

As far as consumer AI, the average consumer up to this point has been exploring "the novelty and utility of AI," and engaging in some limited productivity activities with AI's help, "such as writing, editing and searching." 

As the interactions have become habitual for consumers, new modes of interaction are emerging beyond simply chatting with a bot. "Platforms likeVapi in the Voice AI space are helping power consumers' abilities to interact with machines in a way that spans language, context, and emotion," they note.

Also: How AI-enabled autonomous business will change the way you work forever

And Perplexity "has emerged as a breakout darling," leading "one of the most meaningful shifts is in how consumers search for information and interact with the web altogether."

Emerging successes are focused on specific use cases for the consumer, they write. "This includes AI journals and mentors likeRosebud and gamified self-care companions like Finch."

They note some areas that are ripe for opportunity between the large AI platforms and startups. Two include travel booking and shopping. Travel is "fragmented and time-consuming" today, they observe. "The opportunity for a personalized, end-to-end travel concierge is enormous, but still unclaimed."

And shopping can be "fundamentally reshaped when the starting point is no longer Google but agents that handle browsing, price comparison, and even checkout on the consumer's behalf."

Five predictions for 2026

Moving from the funding data and state of play, the Bessemer report details five predictions for the year ahead in AI:

1. The browser will emerge as a dominant interface for agentic AI. Browsers designed for AI agents include Perplexity'sComet and The Browser Company of New York'sDia browser , and they "will be much more than plug-ins," able to "embed AI at the operating layer." The team expects new browsers from OpenAI and Google and others to become more agent-smart.
   
2. 2026 will be the year of generative video. Generative AI's video quality will shine in 2006 as "Model quality -- across Google's Veo 3, Kling, OpenAI's Sora, Moonvalley's Marey, and emerging open-source stacks -- is accelerating," they note. "We're nearing a tipping point in controllability, accessibility, and realism, that will make generative video commercially viable at scale" in entertainment, education, marketing, social media and retail.
   Open questions are whether video will be dominated by the largest AI models, whether open-source will be competitive in video generation, and whether real-time and low-latency uses of video will become important.
   
3. Evals and data lineage will become a critical catalyst for AI product development.Current ways of evaluating Gen AI are only offering "coarse-grained signals at best," they write. "Most every company still struggles to assess whether a model performs reliably in their specific, real-world use cases." The result is going to be a big advance in custom evaluations, they believe. "AI evals will go private, grounded, and trusted -- and enterprise deployment will 10x because of it." Companies profiting from that wave may include startups asBraintrust , LangChain , Bigspin.ai and Judgement Labs .
   
4. A new AI-native social media giant could emerge. Technology waves always lead to successful startups, so they expect the same will be the case in social media. "PHP enabled Facebook. Mobile cameras made Instagram possible. Advances in mobile video propelled TikTok," the team observed. "It's hard to imagine that the new capabilities enabled by generative AI won't lead to a similar breakout." The young contestants includeCharacter.ai and Replika , but "we don't yet know what form the next social media giant will take," they muse. It could be a platform of helpful agents doing things for you, or it could be one made up of "emotionally intelligent AI influencers and AI clones." 
   
5. The incumbents strike back as AI M&A heats up. The existing software vendors "face a stark choice," they write, "evolve or become obsolete." The team's prediction is that public tech firms will go on a frenzy of acquisitions. "In 2025 and 2026, we expect to see a surge in M&A activity as incumbents move aggressively to buy their way into the AI era."

Also: How AI amplifies these other tech trends that matter most to business in 2025

Get ready for the merger and acquisition frenzy

The authors offer AI startups a few words of advice on how to prepare for the merger and acquisition feeding frenzy.

"SaaS giants are buying their way into AI," the team writes. "Build technical and data moats. Be M&A-ready, but operate like you'll own the category."

Also: Why Perplexity is going after Google Chrome - and yes, it's serious

Their specific recommendations include:

• "Be ready for strategic interest: If you're building a domain-specific or infrastructure-layer AI product, expect inbound from legacy players looking to fill gaps."

• "Play for leverage: The best-positioned startups will have strong technical moats, customer traction, and embedded workflows that make them hard to replicate."

• "Know your acquirer's roadmap: Understand where incumbents are falling behind in your space. If you can deliver what they can't build fast enough, you're valuable."

Artificial Intelligence

Coding with AI? My top 5 tips for vetting its output - and staying out of trouble

The best AI for coding in 2025 (and what not to use)

I found 5 AI content detectors that can correctly identify AI text 100% of the time

I'm an AI tools expert, and these are the only two I pay for (plus three I'm considering)

• Coding with AI? My top 5 tips for vetting its output - and staying out of trouble
• The best AI for coding in 2025 (and what not to use)
• I found 5 AI content detectors that can correctly identify AI text 100% of the time
• I'm an AI tools expert, and these are the only two I pay for (plus three I'm considering)