In past innovation waves, VCs created public market indices to track emerging sectors, such as the Bessemer Cloud Index and the F-Prime Fintech Index. With the rise of AI and Flybridge’s focus on the market, we developed the Flybridge AI Index to track the capital markets performance of public companies in this space. We believe this trend will be one of the largest value creators in history and look forward to seeing how these companies, and those to come, perform over time.

We conducted this analysis for several reasons. As early-stage investors for many decades across many economic cycles, we have seen how crucial it is for founders to have a clear understanding of what drives value in the public markets. By closely monitoring public market trends and comparables, they (and we) can make more informed decisions as they scale their companies. The companies included in this analysis could also potentially become acquirers for some of our startups, making it important to track their growth and strategic priorities.

There is a prevailing notion that incumbent companies will emerge as winners in the GenAI space due to their vast data resources and established distribution channels. We believe that conventional wisdom is wrong. Although we see massive opportunities for founders, we want to closely monitor this trend to validate or challenge this assumption. Lastly, we found this exercise to be both engaging, intellectually stimulating, and fun!

Why This Index?

Flybridge AI Index Methodology

Criteria For Inclusion

Flybridge team select the top 20 U.S.-listed public companies that, in our view, have the highest impact on AI. A company qualifies by meeting several of the following criteria:

Core model development: Builds and trains its own foundation models.
Ecosystem influence: Demonstrably shapes the AI stack through open source projects, model hubs, SDKs, and widely used APIs that other builders adopt.
Critical infrastructure: Provides scarce or enabling layers of the stack, such as GPUs/NPUs.
Data infrastructure for AI: Operates platforms where customer data is stored, processed, and activated specifically for AI workloads.
Revenue impact: Derives a material share of revenue or growth from AI, or has a credible 12–24 month path to AI monetization at a scale that meaningfully impacts financials. AI also represents a significant market expansion potential for their business.
Strategic investment: Commits significant R&D and CapEx to AI, and expands capabilities through acquisitions or partnerships.
Talent and research density: Concentrates top AI researchers and engineers, producing visible contributions to the ecosystem such as papers, benchmarks, or open repos.

Source Material

Source material includes investor update presentations, company websites, analyst day presentations, earnings call transcripts, management interviews, press releases, media coverage, SEC filings, financial reports, industry reports, market research, academic publications, research papers, social media and blog posts from executives and key opinion leaders, conferences, webinars, industry events on AI and technology trends, and feedback from industry experts, analysts, and consultants.

Flybridge uses Aiera.ai as its main data source provider.

Weighting Methodology & Rebalancing Policy

We chose an equally weighted index. Companies like NVIDIA, Google, Meta, and Microsoft have much larger market capitalizations than the others. A market-cap-weighted approach would have let these giants disproportionately drive performance and not represent the overall market.

Flybridge AI Index is capped at a maximum of 20 companies. The selection of these 20 constituents is revisited once per year on the last day of January, and the chosen companies remain fixed for the rest of that year. The only exceptions are when a company that meets our criteria goes public, in which case it may be considered for inclusion 91 days after its IPO, or when a company is delisted, in which case it is removed immediately while its historical data is preserved.

Importantly, rebalancing affects index performance only going forward; if a company is dropped and another is added, past performance remains unchanged to preserve historical continuity.

Index Start Date

We set January 1, 2023, as the Index start date. AI has a history spanning more than five decades, and the 2017 paper "Attention Is All You Need" triggered major advances in generative AI. Still, we view the release of ChatGPT at the end of 2022 as the true inflection point. Its launch brought generative AI into the mainstream, captured public and business attention, and showcased its broad potential, driving rapid development and adoption across industries.

BACK TO INDEX

Flybridge AI Index Methodology

Criteria For Inclusion

Flybridge team select the top 20 U.S.-listed public companies that, in our view, have the highest impact on AI. A company qualifies by meeting several of the following criteria:

Core model development: Builds and trains its own foundation models.
Ecosystem influence: Demonstrably shapes the AI stack through open source projects, model hubs, SDKs, and widely used APIs that other builders adopt.
Critical infrastructure: Provides scarce or enabling layers of the stack, such as GPUs/NPUs.
Data infrastructure for AI: Operates platforms where customer data is stored, processed, and activated specifically for AI workloads.
Revenue impact: Derives a material share of revenue or growth from AI, or has a credible 12–24 month path to AI monetization at a scale that meaningfully impacts financials. AI also represents a significant market expansion potential for their business.
Strategic investment: Commits significant R&D and CapEx to AI, and expands capabilities through acquisitions or partnerships.
Talent and research density: Concentrates top AI researchers and engineers, producing visible contributions to the ecosystem such as papers, benchmarks, or open repos.

Source Material

Flybridge uses Aiera.ai as its main data source provider.

Weighting Methodology & Rebalancing Policy

Importantly, rebalancing affects index performance only going forward; if a company is dropped and another is added, past performance remains unchanged to preserve historical continuity.

Index Start Date

BACK TO INDEX

Insights

Company Analysis within the Index

Companies & Rationale

Included primarily for its Firefly generative AI model, trained on licensed and Adobe Stock data to ensure IP-safe outputs. Firefly is integrated across Photoshop, Illustrator, and other Creative Cloud tools. We recognize the disruption risk Adobe faces from new entrants and will monitor it closely as we reassess the index next year.
Alibaba has become a leading contributor to open source foundation models, notably through its Qwen series, a suite of high performing language, multimodal, coding, math, audio, and vision models.
AMD currently sits as the clear number two or three GPU provider behind NVIDIA and Google, with its MI300 series showing competitive performance in certain workloads. While NVIDIA’s CUDA ecosystem still dominates, AMD’s open-source ROCm platform has been steadily improving, particularly for inference workloads, where the industry’s focus is shifting.

With AI workloads tilting toward inference and enterprises seeking alternatives to NVIDIA’s closed ecosystem, AMD is well positioned to expand its role in the market
Arm underpins much of the global AI ecosystem by designing the energy-efficient CPUs that power smartphones, edge devices, and increasingly data center accelerators. Its licensing model gives it unmatched reach, embedding Arm IP into billions of devices annually, from mobile chips to AI-enabled IoT hardware. With new AI-focused initiatives, including specialized architectures for inference and edge processing, Arm is positioned as a critical enabler of both mass-market AI adoption and large-scale infrastructure build-outs.
Pure-play infrastructure company solving one of AI’s hardest bottlenecks: moving data efficiently between GPUs, memory, and networks at hyperscale. Its PCIe, CXL, and Ethernet connectivity solutions are purpose-built for AI workloads, enabling cloud providers to unlock the full performance of accelerators in training and inference clusters
Broadcom plays a central role in AI infrastructure by supplying both custom accelerators and high-performance networking chips that reduce compute and data bottlenecks at scale. Its technology underpins the efficiency and reliability of hyperscale AI clusters, making it a key enabler of model training and deployment.
Cloudflare delivers the infrastructure tools that make AI more accessible, performant, and secure. It enables inference at the edge, integrates storage and vector-based tools for developers, and offers an MCP layer that lets AI agents interact with tools and data in a controlled way. Cloudflare also protects original content by blocking AI scrapers.
Provides infrastructure-as-a-service tailored for industries like AI, machine learning, and CGI. The expectation is that they will go public in the first half of next year (2026).
Meta remains a leading proponent of open source, with Llama 3 serving as a major catalyst. Although the Llama 4 model fell short of expectations, they continue to drive forward in open source model development. Meta also has one of the highest concentrations of AI talent, now under the leadership of Alexander Wang and Nat Friedman. They created PyTorch, which has become the default deep learning framework across industry and academia
Duolingo shows how AI can reshape both scale and scope of consumer products. Course development that once took 12 years to reach 100 languages now produces over 100 new courses in a single year, with almost all new launches fully automated. AI has also allowed Duolingo to expand beyond languages into new areas like chess, while premium features such as the conversational voice tutor in Duolingo Max are already converting into meaningful subscription revenue.
Google has been foundational to modern AI, from pioneering the “Attention Is All You Need” paper to building world-class research hubs like DeepMind. Its Gemini models rank among the top performers and continue to drive advances in multimodal AI. While its core ads business faces disruption from AI-driven search, Google is actively adapting with AI-powered overviews and remains one of the highest-talent, highest-impact players in the field.
HubSpot has moved quickly to embed AI into its CRM, adding features like Copilot and Breeze Agents for sales, support, and content. Among application-layer incumbents, it’s reportedly moving the fastest, likely because it's still founder-led, with Dharmesh as active CTO and Brian highly involved. A sign of this is Agent.ai, a separate platform they launched for users to create and share agents, which has gained significant traction.
Microsoft has played a key role in AI by investing heavily, acquiring talent, and integrating AI into its products. As a major early backer of OpenAI, it secured exclusive access to GPT models on Azure, making it a primary distribution channel for advanced AI. Microsoft also acquired Inflection AI’s team, led by Mustafa Suleyman, to strengthen its in-house capabilities. It quickly embedded AI into core products like GitHub Copilot and Microsoft 365 Copilot, bringing generative AI to millions of developers and enterprise users.
MongoDB is a core database for enterprises, storing large volumes of operational and application data. With the release of vector search, it now supports hybrid search that combines traditional keyword queries with semantic search over embeddings. The acquisition of Voyage, which brought in-house embedding models, further strengthens its role as an emerging player in AI memory and retrieval.
A central player in AI development, the company supplies GPUs essential for AI processing and applications. It has become the biggest early winner in the generative AI wave, as its advanced GPUs provide the computational power to train and run large models efficiently.
Since ChatGPT launched, the company's market capitalization has risen from $400 billion to $4.3 trillion.
Palantir is one of the earliest application-layer winners in AI, leveraging its ontology and deep customer integrations to layer LLM orchestration on top of existing data systems. Its “consulting-to-software” approach, long criticized, is now a strength in AI adoption.
Leading provider of on-device AI, with its Snapdragon chips enabling phones, PCs, and cars to run increasingly large models locally. The new Snapdragon X Elite can handle LLMs with over 13B parameters directly on-device. With broad OEM adoption in laptops, continued phone upgrades, and long-term automotive design wins, Qualcomm is set to play a central role as AI shifts from the cloud to devices.
Snowflake has become a pivotal platform for enterprise AI, evolving from its roots in structured data to now support unstructured data through Cortex, integrated ML workflows, and targeted acquisitions. Its ease of data sharing, governed environment, and ability to bring models directly to where the data resides make it well positioned as enterprises expand their AI adoption.
TSMC is indispensable to AI because it manufactures nearly all of the world’s most advanced chips, including those from NVIDIA, AMD, and Apple. With over 60% foundry market share and unmatched ability to scale new nodes like 5nm and 3nm into high-yield production.

It effectively holds a monopoly. Samsung and Intel remain far behind, making TSMC the single most critical link in the AI hardware supply chain,

POTENTIAL FUTURE ADDITIONS

Databricks: has extended its strong position in data infrastructure into AI by integrating model training and deployment directly into its platform. The company open-sourced DBRX, a large language model designed for enterprise use, and has positioned its Data Intelligence Platform as a way for customers to unify analytics and AI workflows.
Cerebras: A pioneer in AI hardware, designs and manufactures one of the world's largest and most powerful AI processors, the Wafer-Scale Engine (WSE). Its superior on-chip memory enables higher tokens per second and low latency. As companies adopt more open-source models and expand workloads from training to inference, the company is expected to remain a relevant player. It is already working with large customers like Perplexity. Expect IPO in 2026 - 2027.

BACK TO INDEX

Thoughts on the private markets

Each Spring, Flybridge GP Jeff Bussgang publishes The Rocket Ship Startup List — a curated list of exciting, fast-growing private companies that are well-capitalized and bringing on talent at rapid speeds. This year there are 43 AI companies on the list.

While the list focuses on where talented graduates should set their sights for employment, this analysis is a proxy for what to watch in the private markets as opportunities develop at an unprecedented pace.

Many of the companies on the list have already become household names such as OpenAI, Anthropic, and Humane. Yet others are achieving incredible scale without the same level of brand recognition and some hailing beyond Silicon Valley like Runway ML, Hugging Face, and Qloo.

These private companies aren’t included in the index but we are following their rise.

Thoughts on the private markets

Each Spring, Flybridge GP Jeff Bussgang publishes The Rocket Ship Startup List — a curated list of exciting, fast-growing private companies that are well-capitalized and bringing on talent at rapid speeds. This year there are 43 AI companies on the list.

These private companies aren’t included in the index but we are following their rise.