The Welcome.AI 100

Alphabet's AI research lab builds the Gemini frontier models and breakthroughs like AlphaFold, powering products that reach billions of users and underpinning Alphabet's ~$403B revenue and ~$4.4T market cap.

OpenAI builds frontier large language models (GPT series) and ChatGPT, the consumer product that catalyzed the generative-AI era and now reaches hundreds of millions of weekly users and most of the Fortune 500.

Frontier AI lab behind the Claude model family and Claude Code, one of two dominant Western LLM providers whose models underpin enterprise AI and agentic software development at global scale.

Global cloud-and-software giant whose Azure platform, Copilot suite, and ~27% OpenAI stake make it one of the central commercial engines of the AI economy.

AWS is the world's largest cloud infrastructure provider and the backbone for hosting and AI compute (Bedrock, Trainium) across a vast share of the internet and enterprise IT, making it foundational to the AI economy.

Oracle is a foundational enterprise database, applications, and cloud-infrastructure giant whose OCI division has become a top-tier supplier of AI compute via mega-deals like the $300B+ OpenAI/Stargate partnership.

Databricks builds the leading data-and-AI "lakehouse" platform that lets enterprises unify analytics, ML, and generative-AI workloads on their own data, making it foundational infrastructure for enterprise AI adoption.

Salesforce is the dominant enterprise CRM platform now embedding agentic AI (Agentforce) and unified data (Data 360) across the world's largest companies, making it a central conduit for AI adoption in business workflows.

The world's dominant enterprise resource planning (ERP) vendor, now embedding Business AI (Joule copilots and agentic AI) across the operational backbone of most large global enterprises.

NVIDIA designs the GPUs and CUDA software stack that power the overwhelming majority of global AI training and inference, making it the foundational infrastructure layer of the AI economy.

The world's largest contract chip foundry, fabricating ~90%+ of leading-edge AI accelerators (Nvidia, AMD, Broadcom, Apple) and thus serving as the irreplaceable manufacturing chokepoint of the entire AI compute economy.

Broadcom designs custom AI accelerators (XPUs) and the networking silicon for hyperscalers like Google, Meta, OpenAI, and Anthropic, plus VMware infrastructure software, making it the second-largest AI chip enabler after Nvidia.

ServiceNow runs the cloud-based "Now Platform" for enterprise workflow automation, now infusing agentic AI (Now Assist) across IT, HR, security and customer operations for thousands of large enterprises and governments worldwide.

Meta's AI assistant and open-weight Llama model family, distributed across Facebook/Instagram/WhatsApp to over a billion users, making it one of the largest-reach AI products and the leading open-weight LLM ecosystem.

Alibaba's Qwen is the world's most-downloaded open-weight LLM family and the engine of Alibaba Cloud's triple-digit AI revenue growth, making it the dominant force in open-source AI and China's largest private AI-infrastructure builder.

AMD designs the EPYC CPUs and Instinct GPUs that power a large share of the world's data centers and AI training/inference clusters, making it the primary credible challenger to Nvidia in AI compute.

Anysphere builds Cursor, the AI-native code editor that became the fastest B2B software company ever to reach billions in ARR and is reshaping how enterprises write software.

Hugging Face runs the dominant open-source hub for AI models, datasets, and ML tooling, making it the de facto distribution and collaboration layer for the global machine-learning ecosystem.

ByteDance's Doubao is China's most-used AI assistant and largest domestic LLM platform (via Volcano Engine), defining the Chinese consumer and enterprise AI market at massive scale.

Palantir builds AI/data platforms (Gotham, Foundry, AIP) that operationalize machine learning across defense, government, and enterprise, now a ~$300B+ market-cap bellwether for applied enterprise AI.

A free AI clinical decision-support engine ("ChatGPT for doctors") used daily by a majority of US physicians at the point of care, monetized via pharma advertising — economically significant as the dominant medical AI platform reshaping how clinicians access evidence.

ElevenLabs builds AI voice infrastructure (text-to-speech, voice cloning, conversational voice agents) that has become the default audio layer for enterprises and developers, reaching ~$500M ARR in roughly three years.

CoreWeave runs a specialized GPU cloud purpose-built for generative-AI training and inference, and has become core infrastructure for the largest AI labs and hyperscalers driving the AI compute buildout.

Tencent's in-house Hunyuan foundation-model family powers the Yuanbao assistant and WeChat AI across a 1.4B-user super-app, making it one of China's most widely distributed frontier LLMs.

Snowflake provides a cloud data platform (the "Data Cloud") with built-in AI (Cortex) that lets enterprises unify, govern, and run analytics and AI on their data, making it core infrastructure for thousands of large organizations' data and AI workloads.

IBM watsonx is the enterprise AI and governance platform of a ~$254B mega-cap, monetizing generative AI across regulated industries with a gen-AI book of business reportedly exceeding $12B by early 2026.

Scale AI provides the data-labeling, evaluation, and generative-AI platform that has trained many of the world's frontier models and now powers major US defense AI programs, making it foundational infrastructure for the AI supply chain.

Together AI runs an "AI acceleration cloud" offering fast inference, fine-tuning, and GPU clusters for open-source and custom models, making it core infrastructure for the open-model AI economy.

DeepSeek builds open-weight frontier reasoning LLMs (R1/V3) trained at a fraction of Western costs, and is economically significant for resetting global assumptions about AI compute spend and triggering the largest single-day market-cap loss in history.

Kling is Kuaishou's frontier AI video-generation platform, one of the world's most-used and best-monetized text/image-to-video models, reshaping creative and advertising production at scale.

Baseten runs a usage-based AI inference platform that lets companies deploy and scale ML/LLM models in production, positioning it at the center of the fast-growing inference-compute layer.

Glean is an enterprise AI "Work AI" platform (search, assistant, and agents over a permissions-aware company knowledge graph) significant because it has become a leading enterprise-AI deployment layer with $300M ARR and Fortune 500 penetration across 50+ industries.

Groq builds LPU inference chips and the GroqCloud API for ultra-low-latency AI inference, pioneering the inference-acceleration category that Nvidia validated with a ~$20B licensing deal.

Baidu is China's dominant search and AI-cloud company whose ERNIE foundation models power one of the largest non-US frontier-LLM ecosystems, making it economically significant as the anchor of China's AI stack.

Applied Intuition builds simulation, validation, and AI software infrastructure that the majority of global automakers and a growing roster of defense and industrial programs use to develop autonomous and AI-defined vehicles, making it foundational plumbing for the physical-AI economy.

Ramp is an AI-powered corporate spend management and finance automation platform whose agentic AI now runs procurement, AP, and expense workflows for tens of thousands of businesses, making it one of the fastest-growing fintechs reshaping back-office finance.

Lambda is an Nvidia-backed "AI neocloud" renting GPU compute (on-demand instances and 1-Click Clusters) to enterprises and frontier labs, making it a load-bearing supplier of the scarce AI training/inference capacity underpinning the modern AI economy.

High-speed inference cloud for open-source LLMs that lets enterprises run, fine-tune, and scale generative AI, serving as critical infrastructure behind major AI products and processing trillions of tokens daily.

LangChain provides the dominant open-source framework and commercial tooling (LangSmith, LangGraph) for building, observing, and deploying LLM agents, making it foundational infrastructure for the AI application layer.

Stability AI builds open-weight generative media models (Stable Diffusion image, plus video/audio/3D) that catalyzed the open-source generative-imaging ecosystem now embedded across creative, marketing, and enterprise workflows.

Synthesia is an AI avatar video generation platform that lets enterprises create studio-quality talking-head videos from text, displacing traditional video production for corporate training, comms, and L&D at scale.

Harvey builds AI agents for legal and professional-services work (drafting, due diligence, research), now used by half the Am Law 100 and 1,500+ organizations, making it the category-leading legal AI platform.

DeepL builds high-quality neural machine translation and Language AI (text + voice) used by 200,000+ businesses, making cross-language communication an enterprise utility.

Tempus AI applies AI and multimodal clinical/genomic data to precision oncology diagnostics and pharma drug discovery, making it economically significant as a key data-and-AI backbone for cancer care and pharmaceutical R&D.

Perplexity is an AI-native "answer engine" and browser (Comet) that delivers cited, web-grounded answers, positioning itself as the leading challenger to traditional search and a fast-scaling enterprise AI distribution layer.

Replit is a browser-based AI app-building platform whose autonomous coding Agent lets developers and non-developers ship production apps, making it a category-defining force in the fast-growing "vibe coding" market.

Anduril builds AI-driven autonomous defense systems unified by its Lattice software platform, and is economically significant as the leading venture-backed challenger reshaping the US/allied defense industrial base at multi-billion-dollar scale.

MiniMax is a Chinese frontier-AI developer whose open-weight M2 LLMs, Hailuo video models, and consumer apps (Talkie, Hailuo AI) reach 200M+ users globally, making it one of the most internationally distributed Chinese AI labs and a recently-listed public company.

China's dominant speech/voice AI and applied LLM company, embedding its Spark models across education, healthcare, automotive, and government at national scale, making it economically significant as a pillar of China's domestic AI ecosystem.

Chinese frontier-LLM lab whose open-weight Kimi K2 models top global developer platforms and rival OpenAI/Anthropic on coding, making it the most-funded open-weight AI challenger to Western labs.

HeyGen is an AI avatar video generation platform that lets businesses create studio-quality spokesperson videos in 140+ languages without cameras, economically significant for collapsing the cost of corporate video, training, and localized marketing content.

An AI-native IDE and coding assistant (formerly Codeium) now owned by Cognition, economically significant as a top-tier developer tool whose talent and IP were split between Google (~$2.4B) and Cognition (~$250M) in a landmark 2025 acquisition.

Abridge builds ambient AI that converts patient-clinician conversations into structured clinical notes and billing codes, and is economically significant as the category-leading scribe deployed across 200+ US health systems, materially reducing physician burnout and accelerating revenue cycles.

Bloomberg's purpose-built financial AI — the 50B-parameter BloombergGPT plus a production AI layer (Document Search, AI Summary, the ASKB agentic interface) embedded in the Bloomberg Terminal — economically significant because it threads generative AI into the dominant workflow used by global finance professionals.

AI "vibe coding" platform that lets non-developers build full-stack web apps from natural-language prompts, now a category leader scaling revenue faster than almost any software company in history.

xAI builds the Grok family of frontier LLMs distributed through X, an enterprise API, and government channels, making it one of a handful of labs with the capital and compute to define the AI frontier.

G42 is the UAE's sovereign-AI holding company building Gulf-scale cloud/data-center infrastructure (Stargate UAE, Khazna), the Jais Arabic LLM, and vertical AI across government, healthcare, and energy — making it the anchor of national AI capacity for the Gulf and a geopolitically pivotal US-aligned compute hub.

Pinecone is a fully-managed vector database that powers semantic search and retrieval-augmented generation (RAG) for AI applications, making it foundational infrastructure for the GenAI stack.

Tsinghua-spun Chinese frontier-LLM developer (GLM/Z.ai) whose open-weight models top global coding benchmarks and underpin AI for most of China's top internet firms, making it the world's first publicly traded pure-play AGI foundation-model company.

Builds the FLUX family of frontier open-weight and API image-generation models, now the most-downloaded text-to-image systems and embedded in major creative platforms (Meta, Adobe, Canva), making it the leading independent visual-AI foundation lab.

Cohere builds enterprise-grade frontier LLMs (Command family) and the secure North agent platform for privacy-first, on-prem and sovereign AI deployments, making it the leading neutral enterprise/government alternative to hyperscaler model providers.

Cerebras builds wafer-scale AI chips and operates the fastest AI inference cloud, positioning it as a credible alternative to Nvidia GPUs for high-speed LLM serving.

Runway builds frontier generative-AI video/image models (Gen-4) and a professional creative suite, making it a leading independent player consolidating the pro AI-video market as competitors like Sora exit.

Tesla's autonomy franchise (FSD driver-assist/robotaxi plus the Optimus humanoid robot) aims to convert its vehicle fleet and factories into AI-driven revenue assets, making it one of the most capitalized bets on embodied AI.

Shield AI builds Hivemind, an AI autonomy "pilot" software and the V-BAT drone, enabling GPS/comms-denied autonomous military flight — economically significant as a category leader in defense autonomy with multi-billion-dollar valuation and government contracts.

Weaviate is an open-source AI-native vector database powering semantic/vector search and RAG for AI applications, making it foundational infrastructure for the generative-AI retrieval stack.

Suno is the leading text-to-music generation platform, economically significant as the category-defining AI music app reshaping how music is created and forcing the recording industry into landmark AI-licensing deals.

Luma AI builds frontier text-to-video and 3D generation models (Dream Machine / Ray series) that are reshaping film, advertising, and design creative workflows and distributed inside Adobe Firefly and Amazon Bedrock.

Builds safety-focused generative AI voice agents (Polaris) for non-diagnostic patient engagement like post-discharge follow-up, economically significant for relieving the global healthcare labor shortage at scale.

Europe's flagship frontier-LLM lab building open-weight and commercial models (Le Chat, La Plateforme API), economically significant as the leading non-US sovereign-AI challenger to OpenAI/Anthropic.

Sierra builds enterprise-grade conversational AI agents for customer experience, replacing/augmenting customer-service operations for large brands and capturing a high-value enterprise AI category on outcome-based pricing.

Decagon builds enterprise AI agents that autonomously resolve customer-support conversations, automating a major labor cost center across airlines, banks, telecom, and high-growth tech.

Alphabet's autonomous-driving subsidiary runs the leading commercial robotaxi service (Waymo One), serving 500K+ paid weekly rides and defining the economics of self-driving mobility.

Midjourney is a bootstrapped, profitable AI image (and now video) generation service that became the category-defining consumer text-to-image tool, monetizing millions of paying subscribers at ~$500M revenue with no outside capital.

An end-to-end generative-AI drug discovery company whose Pharma.AI platform produced the first AI-discovered, AI-designed drug to show clinical efficacy in a randomized trial, validating AI-driven pharma R&D.

Europe's largest defense-AI company, building software (battlefield management, electronic warfare, autonomy) and AI strike drones now central to NATO/Ukraine procurement and European defense-tech sovereignty.

AI21 Labs builds enterprise-grade large language models (Jurassic, Jamba) and the Maestro agent-orchestration platform, pioneering efficient Mamba-Transformer hybrid architectures for regulated, low-error-tolerance industries.

Abu Dhabi government research institute (ATRC's TII) that builds the open-source Falcon LLM family, making frontier-competitive, efficient, royalty-free models — including the leading Arabic models — broadly accessible and advancing Gulf-region AI sovereignty.

SambaNova builds custom Reconfigurable Dataflow Unit (RDU) AI chips and a full-stack inference cloud, positioning as an NVIDIA challenger for enterprise, government, and sovereign AI deployments.

Builds advanced mobile robots (Spot quadruped, Stretch logistics, Atlas humanoid) and is the reference brand for commercial legged/humanoid robotics, now pairing its hardware with Google DeepMind's Gemini Robotics models.

Chai Discovery builds AI foundation models for molecular structure prediction and de novo antibody/biologic design, aiming to turn drug discovery from screening into engineering and shrinking design-to-validation cycles to weeks.

SenseTime is a Chinese frontier-AI company whose SenseNova foundation models and SenseCore compute infrastructure power generative-AI, autonomous-driving, and smart-city deployments across China, making it one of the most economically significant AI players in the world's second-largest economy.

C3.ai sells an enterprise/agentic AI application platform and prebuilt industry apps to large corporations and the US government, but its economic weight is modest with ~$250M revenue, deep losses, and a shrinking valuation.

Cresta builds generative-AI agents and real-time agent-assist for enterprise contact centers, automating and augmenting high-volume customer-service and sales conversations for Fortune 500 operators.

Builds Digit, a bipedal humanoid robot performing real revenue-generating warehouse and manufacturing labor, making it one of the first companies to commercially deploy humanoids at paying customers.

Builds hardware-agnostic robot foundation models (the π-series VLA models) that let robots perform open-world physical tasks from natural-language commands, positioning to be the "foundation model layer" for general-purpose robotics.

PIF-owned Saudi full-stack AI company building sovereign data centers, Arabic LLMs (ALLaM), and an enterprise AI OS (HUMAIN ONE), economically significant as the anchor of Saudi Arabia's ~$100B national AI bet and a major buyer of NVIDIA/AMD compute.

Builds frontier protein language models (ESM3) that jointly reason over protein sequence, structure, and function to accelerate drug discovery and synthetic biology; economically significant as a leading AI-for-biology lab now absorbed into Chan Zuckerberg Biohub.

Consumer AI-companion and roleplay chatbot platform that pioneered the personified-AI category, economically significant as one of the largest consumer AI apps by usage and as the subject of Google's $2.7B license/acqui-hire.

Poolside builds proprietary frontier coding foundation models (Malibu) deployed securely inside enterprise, defense, and government boundaries, making it economically significant as a heavily-capitalized, Nvidia-backed challenger in the enterprise AI software-development market.

Alphabet/DeepMind spinout using AlphaFold-derived AI (IsoDDE) to design small-molecule drugs, partnered with top pharma to potentially reshape early-stage drug discovery economics.

Recursion runs an AI-driven drug-discovery platform (Recursion OS) that industrializes target-to-clinic biology, making it one of the most prominent public bets on AI transforming pharma R&D.

German sovereign-AI lab building EU-compliant LLMs (Pharia) and an on-prem enterprise OS (PhariaAI) for governments and regulated industries; economically significant as Europe's flagship sovereign-AI bet, now absorbed by Cohere in a ~$20B combined entity.

Mira Murati's frontier-AI startup building customizable models and the Tinker fine-tuning API; economically significant chiefly for its record $12B seed valuation and elite ex-OpenAI talent concentration rather than current revenue.

Wayve builds end-to-end "embodied AI" (AV2.0) self-driving software licensed to automakers and mobility operators, positioning itself as the global software layer for autonomy across robotaxis and consumer vehicles.

Builds the "Skild Brain," a unified omni-bodied robotics foundation model meant to control any robot for any task, with potential to commoditize the intelligence layer of physical automation.

Pika is a venture-backed AI video generation startup turning text and image prompts into short videos, notable for consumer reach and an Adobe Firefly distribution deal, but a mid-tier player in a crowded field.

India's full-stack sovereign AI startup building voice-first LLMs (30B/105B), speech, and translation models for 22 Indic languages, economically significant as the government-anointed builder of India's first indigenous foundational model under the IndiaAI Mission.

World Labs builds "large world models" that generate navigable, persistent, editable 3D environments from text/images (product: Marble), positioning spatial intelligence as foundational infrastructure for robotics simulation, gaming, and design.

Figure AI builds general-purpose humanoid robots (Figure 03) powered by its in-house Helix vision-language-action model, positioning it as the best-capitalized pure-play humanoid company aiming to put robots into factories, logistics, and eventually homes.