Who's Building AI?

OpenAI is an AI research and deployment company building toward AGI for the benefit of humanity. Now a Public Benefit Corporation (PBC). Developer of the GPT and o-series models, ChatGPT (1B+ WAU), GPT Image 2, and the Codex coding agent. Its 2026 consolidation retired several side products — Sora, the Atlas browser, and Operator — folding browsing and coding into ChatGPT Work. The $852B valuation funding round (Mar 2026) is the largest private fundraise in history.

Anthropic is an AI safety company building the Claude model family. With $965B near-trillion valuation (2026), $45B ARR, and a confidential IPO filing (Jun 2026), Anthropic has become the fastest-growing AI company in history. Claude Fable 5.1 (Sep 2026) is its most capable generally available model; the current Opus, Sonnet and Haiku are 5.5 (Sep-Oct 2026). Amazon has committed $33B+ total, including $25B in Apr 2026 for 5GW of compute. Claude Code hit $2.5B annualized ARR.

Google DeepMind is the unified AI research arm of Alphabet, combining Google Brain and DeepMind. Pioneers in reinforcement learning, protein folding, and multimodal AI.

Meta AI is Meta's artificial intelligence research division. Champions open-source AI with the Llama model family and conducts fundamental research in NLP, computer vision, and robotics.

Microsoft has made AI central to its business strategy through a multi-billion dollar partnership with OpenAI. Integrates AI across Azure, Office 365, Bing, and GitHub with Copilot products.

Hugging Face is the leading open-source AI platform, hosting 500,000+ models and 100,000+ datasets. The Transformers library powers much of the modern NLP ecosystem.

Mistral AI is a European AI startup building highly efficient open and commercial language models. Their models achieve competitive performance at significantly smaller parameter counts.

Stability AI is the company behind Stable Diffusion, the open-source image generation model that democratized AI art creation. Also develops models for audio, video, and code.

Cohere provides enterprise-focused NLP APIs for search, classification, summarization, and generation. Strong in retrieval augmented generation (RAG) and enterprise deployments with on-prem options.

Runway is a creative AI company building next-generation video generation and editing tools. Their Gen-2 and Gen-3 models enable text-to-video and video-to-video transformations of cinematic quality.

ElevenLabs develops the most realistic AI voice synthesis technology available. Their models capture emotional nuance, accent, and prosody at a level indistinguishable from humans.

Perplexity AI is building a new kind of search engine that combines LLMs with real-time web search to provide cited, accurate answers. Used by researchers, journalists, and knowledge workers.

Scale AI provides data labeling and synthetic data infrastructure that powers AI training for leading frontier labs and enterprises. The 'picks and shovels' of the AI gold rush.

Together AI provides fast, affordable inference and training infrastructure for open-source AI models. Hit $1B annualized run rate Feb 2026. Targeting $7.5B valuation in new ~$1B funding round (Mar 2026). Own data center in Maryland live since Jul 2025.

Inflection AI was co-founded by Mustafa Suleyman and developed Pi, a personal AI focused on emotional intelligence and supportive conversation. Most of the team joined Microsoft AI in 2024.

xAI is now the AI division of SpaceX after the largest merger in history (Feb 2026, $1.25T combined entity). Develops the Grok model family with real-time X data access. Grok 4 (Jul 2025) and Grok 4.3 (Apr 2026) are current frontier models. Colossus supercluster (220K+ GPUs) leased entirely to Anthropic in May 2026 after xAI migrated to Colossus 2.

Midjourney is an independent research lab that produces an AI image synthesis service with a distinctive, highly aesthetic output style. Operated primarily through Discord, it has become the most widely recognized AI art generation tool among creatives.

Pika Labs is an AI video generation startup that enables anyone to create and edit video content from text prompts. Their Pika 1.0 model democratized AI video creation with an accessible interface and quality results.

Synthesia is an enterprise AI video platform that enables organizations to create professional videos with AI-generated presenters in 130+ languages without cameras, studios, or actors.

HeyGen is an AI video platform enabling businesses to create personalized videos with realistic AI avatars and voice cloning. Their video translation feature can dub content into multiple languages while preserving the speaker's voice and lip movements.

Groq is a chip startup that has developed the Language Processing Unit (LPU), custom silicon designed specifically for LLM inference. Their hardware delivers inference speeds 10-25x faster than GPU-based alternatives.

Cerebras Systems builds the world's largest computer chips - the Wafer Scale Engine (WSE) - designed specifically for AI training and inference.

AI21 Labs is an Israeli AI research company developing large language models and enterprise NLP tools. Creator of Jamba, the first production-grade hybrid SSM-Transformer model.

Character.AI is a conversational AI platform for AI personas of celebrities and fictional characters. Underwent a major restructuring in Aug 2024 when Google completed a 'reverse acquihire' - founders Noam Shazeer and Daniel De Freitas (+ 24 researchers) returned to Google DeepMind. Character.AI continues operating independently but faces steep competition and declining user growth.

Replicate is a cloud platform that makes it easy to run machine learning models with a single API call. Hosts thousands of open-source models with pay-per-prediction pricing.

Weights & Biases (W&B) is the leading MLOps platform for tracking experiments, visualizing model performance, and managing the machine learning lifecycle.

Pinecone is the leading managed vector database company, providing production-ready infrastructure for similarity search at scale.

Aleph Alpha is Germany's leading AI company developing sovereign AI infrastructure for European enterprises and governments.

Moonshot AI is a Chinese AI startup developing the Kimi family of large language models with an emphasis on extremely long context windows.

Zhipu AI is a leading Chinese AI company and developer of the GLM series of bilingual language models. A spin-off from Tsinghua University's KEG lab.

01.AI is a Chinese AI company founded by Kai-Fu Lee, developing the Yi series of open-source large language models globally competitive in performance.

SambaNova Systems designs AI accelerator hardware and full-stack AI platforms for enterprise and national lab deployments.

Adept AI is building AI agents that can take actions in software on behalf of users. Their Fuyu-Heavy model and ACT agent are designed to interact with existing software interfaces.

DeepSeek AI is a Chinese AI research lab founded by quantitative hedge fund High-Flyer Capital that shocked the global AI industry in January 2025 by releasing open-source models (DeepSeek R1, V3) that match GPT-4-class performance at a fraction of the training cost. Their work demonstrated that frontier AI is achievable with far less compute than previously assumed.

Apple is a $3.5T+ technology conglomerate whose AI strategy centres on on-device intelligence, privacy-preserving Personal Intelligence, and deep ecosystem integration across iPhone, Mac, iPad, Apple Watch, and Vision Pro. Apple Intelligence (launched iOS 18.1, Oct 2024) uses on-device models and Private Cloud Compute to deliver AI features without sending data to the cloud.

Amazon Web Services is the world's leading cloud platform and a pivotal force in enterprise AI infrastructure. Through Amazon Bedrock (multi-model AI API), Amazon SageMaker (end-to-end ML platform), Amazon Q (enterprise AI assistant), and Amazon Nova models, AWS enables tens of thousands of enterprises to build and deploy AI at scale. Amazon has also committed $8B+ to Anthropic.

Baidu is China's dominant internet search company and one of the country's leading AI companies, investing heavily since 2010. ERNIE Bot (文心一言) — its LLM and AI assistant — had 200M+ registered users by 2025. Baidu's Qianfan platform provides enterprise AI APIs and its Apollo self-driving division is among China's most advanced autonomous vehicle programmes.

Kuaishou is China's second-largest short video platform and the company behind Kling AI, one of the world's most capable text-to-video generation models. Kling 2.0 Master ranked #1 on EvalCrafter video benchmarks in 2025, supporting up to 3-minute high-fidelity videos. Kuaishou also operates Kwai internationally.

Codeium built the AI coding assistant Codeium and the agentic IDE Windsurf. Known for its generous free tier and fast autocomplete, Codeium grew to 800,000+ developers. A reported ~$3B OpenAI acquisition collapsed on 11 July 2025 when the exclusivity period lapsed; Google then paid around $2.4B in a non-exclusive licensing deal that took co-founders Varun Mohan and Douglas Chen and part of the R&D team to DeepMind, and Cognition AI acquired the remaining business — IP, product, brand, 350+ enterprise customers and about $82M of ARR — on 14 July 2025 on undisclosed terms. Windsurf's Cascade agent autonomously edits code across entire repositories.

Anysphere is the company behind Cursor, the AI-first code editor that became the fastest-growing developer tool in history. Cursor is a fork of VS Code with native multi-model AI capabilities (Claude, GPT-4, Gemini), full codebase context, agentic multi-file editing, and background agents. It reached $3B ARR by April 2026, with 64% of Fortune 500 companies adopting it. The company is in valuation discussions at $50B as of June 2026.

Black Forest Labs is the company behind FLUX, the state-of-the-art text-to-image model family. Founded by the original creators of Stable Diffusion (Robin Rombach and team from CompVis/Stability AI), FLUX.1 Pro and FLUX 1.1 Pro Ultra (4MP) set the benchmark for photorealism and prompt accuracy in 2024-2025. The open-weights FLUX.1 [schnell] and [dev] variants became the most widely adopted open-source image generation models post-Stable Diffusion.

Suno is the leading AI music generation company, enabling anyone to create full songs with vocals, instruments, and production quality from a text prompt. Suno v5.5 (March 2026) delivers radio-quality music across all genres. With $400M raised at a $5.4B valuation (June 2026) and $100M+ ARR, Suno is the dominant consumer AI music platform. The company faces ongoing litigation from major record labels (UMG, Sony) though Warner Music settled.

Cognition AI is the developer of Devin, the world's first AI software engineer agent capable of autonomously completing complex engineering tasks end-to-end. Devin 2.0 (March 2025) dramatically reduced pricing from $500 to $20/month. Cognition acquired Windsurf on 14 July 2025, days after OpenAI's deal collapsed and Google licensed its technology, on undisclosed terms; it came with 350+ enterprise customers and about $82M of ARR, and was later folded into Devin Desktop. Customers include Goldman Sachs, Mercedes-Benz, Citi and Santander.

Vercel is the platform for frontend developers, enabling teams to deploy web applications with zero configuration. In the AI era, Vercel launched v0 (AI UI component generator, September 2023) which became the leading design-to-code tool. Vercel also develops the AI SDK (open-source toolkit for building AI-powered applications with React/Next.js) and Next.js, the most widely deployed React framework. Valued at $3.25B, Vercel powers 3M+ projects.

StackBlitz is the company behind Bolt.new and WebContainers technology, which enables running full Node.js environments directly in the browser. Bolt.new (launched September 2024) became the fastest-growing AI development tool, enabling anyone to build and deploy full-stack web applications from a single text prompt in a browser — no local setup required. Bolt.new reached 1M+ monthly active users by December 2024.

NVIDIA is the dominant designer of GPUs and AI accelerators. Its data-center GPUs (A100, H100, H200, and the Blackwell B200/GB200) together with the CUDA software platform power the majority of the world's AI training and inference, making NVIDIA the foundational infrastructure provider of the modern AI era. Beyond silicon, NVIDIA ships full-stack AI software including NVIDIA AI Enterprise, NIM microservices, TensorRT, and the Omniverse simulation platform.

Writer is an enterprise-grade AI platform purpose-built for large organisations, offering custom LLMs (the Palmyra series), a full-stack AI application builder, RAG over enterprise knowledge, and a governed AI writing assistant. Unlike generic tools, Writer is built for compliance, brand consistency, and enterprise security, serving customers like Salesforce, Verizon, Intuit, and L'Oréal.

Harvey is an AI platform built specifically for law firms and legal departments, using fine-tuned LLMs to assist with legal research, contract analysis, due diligence, litigation drafting, and regulatory compliance. Backed by OpenAI and top law firm investors, Harvey is the leading AI platform for the legal industry, with clients including Allen & Overy, PwC, and the US Air Force legal department.

Twelve Labs builds multimodal video understanding AI that can search, analyze, and extract information from video content semantically — not just via transcription. Its Marengo and Pegasus model families enable developers to build apps that understand what happens in video, who appears, what is said, and the mood or context, all through a developer-friendly API.

Poolside is an AI company focused exclusively on building the world's best AI for software engineering. Using reinforcement learning from code execution (not just human feedback), Poolside trains models that understand software at a deep level. Its Malibu and Point models are available via API and are designed for agentic coding workflows, with Fortune 500 and developer-tool company customers.

CoreWeave is the leading specialised cloud provider for AI/GPU workloads, operating over 250,000 NVIDIA GPUs across 32 data centres. Originally a crypto mining company, it pivoted to GPU cloud in 2019 and rode the AI training boom to become one of the fastest-growing cloud infrastructure companies ever. CoreWeave completed a $23B IPO in March 2025 — the largest US tech IPO since Arm Holdings. Customers include Microsoft, Cohere, OpenAI (via Stargate), and IBM. It acquired Weights & Biases for $1.7B in July 2025 to add ML observability to its stack.

Databricks is the leading unified data analytics and AI platform, combining big data processing (based on Apache Spark) with machine learning, LLM fine-tuning, and governance. Founded by the creators of Apache Spark, Delta Lake, and MLflow, Databricks serves 10,000+ enterprises worldwide. Their open-source DBRX (March 2024) set a new benchmark for open-weight MoE models at launch. At a $62B valuation (Dec 2024), Databricks is one of the most valuable private software companies in the world.

Luma AI is a multimodal AI company best known for Dream Machine, a state-of-the-art text-to-video and image-to-video model, and NeRF-based 3D capture technology. Dream Machine (June 2024) generates high-quality, cinematic videos from text or images in seconds and rapidly became one of the top consumer video AI tools. Luma also develops AI for 3D scene reconstruction (Luma NeRF) used in film, AR/VR, and e-commerce.

Fireworks AI is a high-performance AI inference platform that serves open-source frontier models — including Llama 4, DeepSeek R1, Qwen, Mixtral, and FireFunction — at industry-leading speed and cost. Founded by ex-Google Brain researchers, Fireworks specialises in inference optimisation with custom CUDA kernels, speculative decoding, and multi-GPU orchestration, delivering latency benchmarks that consistently beat AWS, Azure, and Together AI. Its FireFunction V2 model leads tool-calling benchmarks.

Modal is a serverless cloud platform built specifically for AI and ML workloads. Developers define Python functions with decorators and Modal handles GPU provisioning, container building, scaling, and scheduling automatically. There is no infrastructure to manage — an LLM fine-tuning job or a diffusion model inference endpoint can be deployed in minutes. Modal is especially popular for running batch AI jobs, hosting ML model APIs, and fine-tuning open-source models without DevOps overhead.

Ideogram is an AI image generation startup that broke out in 2024 by solving a long-standing problem: accurate, stylistically rich text rendering within generated images. Ideogram 2.0 (Feb 2025) and Ideogram 3.0 (2025) produce photorealistic images and typography-heavy designs (posters, logos, social graphics) with unprecedented legibility. Its Magic Fill and Edit features allow targeted inpainting. Ideogram's API is used by Canva, Adobe, and other creative platforms.

Lightricks is an Israeli AI creative technology company best known for its consumer apps (Facetune, LightLeap, Videoleap) and for LTX Video — one of the fastest open-source text-to-video diffusion models, capable of generating high-quality video in real time. LTX-Video (Nov 2024) was released open-source on Hugging Face and became one of the most-downloaded video generation models. Lightricks also runs the B-Roll AI stock video platform and offers an enterprise AI content generation API.

Lambda Labs is an AI cloud computing company providing on-demand and reserved GPU instances for AI training and inference. Its Lambda Cloud offers H100, A100, and H200 clusters with competitive pricing specifically targeting AI/ML workloads. Lambda also sells Lambda Workstations and Lambda Blade servers for on-premise GPU computing. It is a preferred GPU cloud for AI researchers, startups, and enterprises who need flexible access to NVIDIA accelerators without long-term commitments.

Replit is a cloud-based collaborative coding platform and AI agent that lets anyone build, run, and deploy software from a browser. The Replit Agent can build complete applications from natural language descriptions across 50+ programming languages. With 30M+ registered developers and a $1.16B valuation, Replit is a leader in AI-assisted software development education and rapid prototyping.

Notion is an all-in-one workspace that combines notes, wikis, databases, and project management into a single flexible platform. Notion AI (launched 2023) adds LLM-powered writing, summarization, Q&A across pages, and database autofill. With 30M+ users and a $10B valuation, Notion is the leading productivity and knowledge management platform for teams.

Safe Superintelligence Inc. (SSI) is an AI lab founded by OpenAI co-founder and former chief scientist Ilya Sutskever with a single stated goal: building safe superintelligence. The company has no products and no near-term commercial roadmap, instead pursuing one straight-shot research effort. Despite shipping nothing publicly, SSI reached a $32B valuation in 2025, making it one of the most valuable AI labs by capital-per-employee.

Thinking Machines Lab is an AI research and product company founded by former OpenAI CTO Mira Murati. The company focuses on making AI systems more widely understood, customizable, and capable, with an emphasis on multimodal models and human-AI collaboration. It assembled a high-profile team of ex-OpenAI, Meta, and Mistral researchers and reached a $12B valuation in one of the largest seed rounds in history.

World Labs is a spatial intelligence AI company founded by Stanford professor Fei-Fei Li (the 'godmother of AI' who created ImageNet). The company builds Large World Models (LWMs) that understand and generate interactive, physically grounded 3D worlds from a single image or text prompt. World Labs targets applications in gaming, film, robotics, design, and virtual environments.

MiniMax is a Shanghai-based AI company building foundation models across text, audio, image, and video. It is best known for its Hailuo AI video generator, the MiniMax-01 series of long-context LLMs (up to 4M tokens), and popular consumer apps like Talkie. Backed by Alibaba and Tencent, MiniMax is one of China's leading 'AI tiger' startups and filed for a Hong Kong IPO in 2026.

Sierra is an enterprise conversational AI company co-founded by Bret Taylor (former Salesforce co-CEO and OpenAI board chair) and Clay Bavor (former Google VP). It builds branded AI agents that handle customer service and support for large companies, resolving issues end-to-end across chat and voice. Sierra reached a $10B valuation in 2025, making it one of the most valuable AI agent startups.

Physical Intelligence (Pi / π) is a San Francisco robotics AI company building general-purpose foundation models for robots — software that can control many different robot bodies to perform real-world physical tasks. Its flagship π0 (pi-zero) and π0.5 vision-language-action models let robots fold laundry, clean tables, and generalize to new environments. Backed by Jeff Bezos, OpenAI, and Thrive at a $2.4B+ valuation.

Glean is an enterprise AI company building a 'Work AI' platform that connects to a company's apps (Slack, Google Drive, Jira, Salesforce, and 100+ others) to power secure, permissions-aware search, assistants, and agents across all internal knowledge. Founded by ex-Google search engineers, Glean reached a $7.2B valuation and $100M+ ARR, becoming a leader in enterprise AI search and agents.

Hume AI is a research lab and startup building emotionally intelligent voice AI. Its Empathic Voice Interface (EVI) is a conversational speech model that measures and responds to the emotion, tone, and prosody in a user's voice, aiming to make AI interactions more natural and human-aware. Founded by former Google DeepMind researcher Alan Cowen.

Abridge is a healthcare AI company that uses generative AI to convert patient-clinician conversations into structured clinical notes in real time, integrated directly into electronic health records like Epic. It reduces physician documentation burden and burnout, and is deployed across many of the largest US health systems. Reached a $5.3B valuation in 2025.

Figure AI is a humanoid robotics company building general-purpose autonomous robots for commercial and home use. Its Figure 02 and Figure 03 humanoids combine dexterous hardware with Helix, an in-house vision-language-action (VLA) model that lets the robots see, reason, and act from natural-language instructions. After ending its OpenAI collaboration to build AI in-house, Figure reached a reported $39B valuation in 2025.

Cartesia is a real-time voice AI company building ultra-low-latency speech models based on state space model (SSM) architectures. Its Sonic text-to-speech model produces natural, expressive voices with sub-100ms latency, and the company develops on-device and streaming voice technology for conversational agents. Founded by the researchers behind the Mamba SSM architecture.

Decagon builds AI agents for customer support that autonomously resolve complex customer conversations across chat, email, voice, and SMS. Its Agent Operating Procedures (AOPs) let companies encode support workflows in natural language, and the platform integrates with existing helpdesks to deflect and resolve tickets end-to-end. Decagon is used by companies like Duolingo, Notion, Rippling, and Eventbrite, and reached a $1.5B valuation in 2025.

Lovable is a Swedish AI startup whose platform lets anyone build full-stack web apps and websites from natural-language prompts. Its AI generates React front-ends, connects databases and auth via Supabase, and deploys instantly, making it one of the fastest-growing 'vibe coding' tools. Lovable reached a reported $100M ARR within roughly eight months of launch and raised a large round at a multi-billion-dollar valuation in 2025.

Contextual AI builds enterprise-grade retrieval-augmented generation (RAG) systems for knowledge-intensive workflows. Founded by the researchers who pioneered RAG, its platform delivers grounded, accurate AI over a company's own data with specialized components for retrieval, reranking, and grounded language modeling. It targets regulated, high-stakes industries such as finance, engineering, and telecommunications.

Nous Research is an applied AI research organization focused on open-source large language models, decentralized training, and human-aligned agents. Known for its widely used Hermes model series (fine-tuned open models), it also pioneers distributed training methods like DisTrO and Psyche that let models be trained across the internet rather than in a single data center. Nous has a strong following in the open-source and crypto-adjacent AI communities.

Prime Intellect is a decentralized AI compute and training platform building toward open, distributed artificial intelligence. It aggregates GPU compute from many providers and pioneers globally distributed model training, having produced INTELLECT-1 and INTELLECT-2 — some of the largest models ever trained across geographically separated hardware. Its mission is to make frontier-scale training accessible outside centralized data centers.

Reka AI is a research-driven AI company building efficient multimodal foundation models that understand and reason over text, images, audio, and video. Founded by researchers from DeepMind, Google Brain, and Meta, Reka delivers compact yet capable models (the Reka Core, Flash, and Edge family) designed for enterprise and on-device deployment, with a focus on strong performance at lower compute cost.

Mercor is an AI-powered hiring and human-data marketplace that connects domain experts with AI labs to produce high-quality training and evaluation data. It uses AI to source, vet, and match specialists — from software engineers to doctors and lawyers — for contract work, and has become a major supplier of expert human data used to train and evaluate frontier models. Mercor grew explosively to hundreds of millions in annualized revenue.

Magic AI (Magic.dev) builds frontier AI models for software engineering, with a focus on extremely long context and autonomous coding. Its models are designed to reason over entire codebases at once, and the company drew attention for pioneering ultra-long context windows (up to 100 million tokens) aimed at letting an AI 'see' a full repository. Magic partners with cloud providers to train on large GPU clusters.

Hippocratic AI builds safety-focused large language models for healthcare, powering AI 'agents' that make patient-facing phone calls for non-diagnostic tasks such as chronic-care check-ins, medication reminders, and pre-op instructions. The company emphasizes clinical safety, testing its models with thousands of nurses and physicians, and positions its agents to address healthcare staffing shortages.

Imbue is an AI research company building AI agents that can reason and code, with the long-term goal of enabling people to build their own software and tools. Formerly known as Generally Intelligent, Imbue trains its own large foundation models optimized for reasoning and develops agent systems and developer tooling aimed at robust, verifiable AI that non-experts can direct.

ByteDance runs one of the largest AI operations in the world through its Seed research organisation, and is unusual in reaching enormous consumer scale before Western attention caught up. Its Doubao assistant is China's most used AI chatbot at roughly 155 million weekly active users, and the Seed model family was processing more than 180 trillion tokens a day as of June 2026. The same models power CapCut's generative video features and Dreamina, giving it a distribution advantage few labs have: hundreds of millions of people already inside its apps.

Alibaba's Qwen family has become the most widely downloaded open-weight model line in the world, passing three billion downloads in the six months to August 2026 and overtaking Meta and Google on that measure. More than 460 Qwen models have been open-sourced and the ecosystem has produced over 300,000 derivatives, which makes Alibaba the centre of gravity for open-weight AI in a way its Western equivalents are not. It also ships closed flagships: Qwen 3.8-Max launched 3 August 2026 with 2.4 trillion parameters and a one-million-token context.

Tencent is one of China's two largest internet groups and a frontier-model developer through its Hunyuan line. Hunyuan Hy3, released on 6 July 2026 after an April preview, is a mixture-of-experts model reported at 295B total and 21B active parameters, aimed at autonomous agent tasks, and was wired into Tencent's Yuanbao assistant at launch alongside a cut to developer API pricing.

Huawei is China's principal domestic alternative to NVIDIA for AI training and inference silicon. It began mass shipments of the Ascend 910C in 2025, built on SMIC's enhanced 7nm process, and its CloudMatrix 384 rack system links 384 Ascend 910C chips for roughly 300 BF16 petaFLOPS, which reporting places above NVIDIA's GB200 NVL72 at the rack level.

Palantir is the dominant vendor for AI deployed on sensitive government and enterprise data. Its Artificial Intelligence Platform acts as a governed orchestration layer over third-party language models rather than training its own, and its short-form 'Bootcamp' pilots have largely displaced conventional enterprise sales cycles.

Anduril is the largest of the new defence technology companies, building AI-driven autonomous systems around its Lattice command-and-control software. It raised $5B in May 2026 at a $61B valuation, roughly doubling the $30.5B set in 2025, and reported doubling 2025 revenue to $2.2B.

Waymo operates the only large-scale commercial fully driverless ride-hailing service in the United States, running in Phoenix, San Francisco, Los Angeles, Austin, Atlanta and Miami. It delivered roughly 15 million rides in 2025 and is expanding to more than twenty US cities plus London and Tokyo during 2026.

AMD is NVIDIA's most credible merchant-silicon competitor in AI training and inference. In October 2025 it signed a 6-gigawatt supply partnership with OpenAI, with the first gigawatt of Instinct MI450 accelerators deploying from the second half of 2026, and issued OpenAI a warrant for up to 160 million AMD shares vesting against deployment and share-price milestones.

IBM was positioned highest and furthest right in the Leaders quadrant of the inaugural Gartner Magic Quadrant for AI Governance Platforms, published 16 June 2026, which assessed thirteen vendors. Its watsonx portfolio spans model building, governance and agent orchestration, and it has open-sourced its Granite model family under Apache 2.0.

ServiceNow was named a Leader in the inaugural Gartner Magic Quadrant for AI Governance Platforms in June 2026, alongside IBM and Truyo. Its Now Assist products are scaling quickly: customers spending over $1M in annual contract value on Now Assist grew more than 130% year over year, and Q2 2026 subscription revenue reached $3.88B, up 24.5%.

Helsing is Europe's highest-valued defence technology company, building AI software for battlefield decision-making alongside its own autonomous aircraft, strike drones and underwater vehicles. Its 2026 Series E of $1.8B at an $18B valuation was the largest defence technology round ever raised in Europe.

Skild AI builds a single robotics foundation model intended to control many robot body types without hardware-specific retraining, trained on large volumes of synthetic simulation episodes combined with real-world human manipulation video. Its January 2026 round of $1.4B at a valuation above $14B is among the largest ever for a robotics AI company.

Wayve develops end-to-end learned autonomous driving software designed to be hardware- and geography-agnostic, licensed to carmakers rather than operated as a fleet. Its February 2026 round of $1.2B at an $8.6B valuation brought in Mercedes-Benz, Stellantis and Nissan as its first automaker investors, and it plans a London robotaxi service during 2026.

Sakana AI is Japan's most valuable AI startup, founded by former Google researchers including Transformer co-author Llion Jones. It pursues small, efficient models optimised for Japanese language and context rather than frontier-scale training, and builds self-improving and research-automation systems such as its AI Scientist.

Isomorphic Labs is an Alphabet AI drug-discovery company spun out of Google DeepMind in 2021 and led by Demis Hassabis. It builds on AlphaFold-lineage research to design drug candidates with its AI drug design engine (IsoDDE). In May 2026 it raised a $2.1B Series B led by Thrive Capital.

Manus, built by Butterfly Effect, makes a general-purpose autonomous AI agent. Meta agreed to acquire it for about $2B in December 2025, but on 27 April 2026 China's National Development and Reform Commission ordered the deal withdrawn. Manus announced in August 2026 that it would operate independently again, and in October 2026 it closed a round of more than $500M co-led by Boyu Capital and IDG Capital.

Nebius is an Amsterdam-headquartered, Nasdaq-listed AI cloud (neocloud) provider of GPU infrastructure for model training and inference. It took its current form in 2024, when Yandex N.V. sold its Russian business and renamed itself. In Q2 2026, revenue grew 454% year over year to $582.3M.

Crusoe is a vertically integrated AI infrastructure ('AI factory') company that builds energy-first data centres and runs a GPU cloud. It is developing the planned 1.2 GW Abilene, Texas campus, whose first buildings serve Oracle and OpenAI's Stargate.

DeepL is a Cologne-based AI language company that grew out of the Linguee dictionary and launched DeepL Translator in 2017. Its neural machine translation and DeepL Write products are widely used by enterprises. Its May 2024 round valued it at $2B.