Weekly briefing
Get the weekly read on what changed across the AI industry map.
The weekly briefing will roll up launches, category shifts, and editorial notes instead of trying to turn the site itself into a noisy news feed.
Company directory
Compare mode
Compare up to three companies side by side.
Use the compare button on company cards to line up focus, stage, access, products, and competitive context without leaving the directory.
Beijing, China
01.AI matters because it remains one of the recognizable Chinese open-model efforts, even as the competitive field there keeps shifting fast.
Palo Alto, US
1X stands out because it is combining humanoid hardware, home deployment ambitions, and a growing in-house AI stack instead of treating robotics as hardware alone.
Las Vegas, US
Abnormal Security matters because security remains one of the clearest enterprise AI applications where model-driven detection and workflow automation have direct operational value.
Pittsburgh, US
Abridge matters because it is one of the clearest enterprise AI companies converting ambient clinical conversation into large-scale healthcare workflow infrastructure.
San Francisco, US
Adept fits the map as a workflow-layer company trying to make agents useful across the existing enterprise stack, not just inside a chat UI.
Seattle, US
AI2 is one of the most important public-interest labs in the ecosystem because it pairs serious research with open releases and evaluation work that other builders actually use.
Hangzhou, China
Alibaba matters because it combines model development with cloud distribution and consumer-scale surfaces, giving Qwen a path from research into deployment.
Seattle, US
Amazon’s AI effort is best understood as a distribution machine: models, tooling, and agents delivered through AWS rather than a standalone consumer lab story.
San Francisco, US
AMI Labs is an early frontier lab working on world models, planning, and persistent memory, with a thesis that controllable intelligence for the physical world requires more than next-token prediction.
Costa Mesa, US
Anduril matters because it is one of the clearest real-world AI deployment companies, pairing autonomy software with vertically integrated defense hardware.
San Francisco, US
Anthropic has become one of the strongest frontier competitors by pairing Claude model quality with enterprise trust, long-context workflows, and unusually strong coding and agent performance.
San Francisco, US
Anyscale is an important bridge between research frameworks and production AI systems because Ray is already deeply embedded in many real-world workflows.
San Francisco, US
Apollo.io matters because sales workflow automation is one of the most commercially important enterprise AI layers, and Apollo sits close to daily go-to-market execution.
San Francisco, US
Arcee AI is worth tracking because it is trying to combine open-weight model strategy with enterprise distribution rather than stopping at research visibility alone.
Palo Alto, US
Augment Code is notable because it is explicitly selling the idea that superior code quality comes from deeper codebase context, not just a stronger general model.
San Francisco, US
Autoscience Institute is worth tracking because it pushes toward the idea of AI systems that do substantive model-development work, not just assistant tasks for researchers.
New York, US
Axiom Math is pursuing formal mathematical reasoning as a wedge into high-trust technical workflows, aiming to build systems that reason with proofs rather than generic language-model imitation.
San Francisco, US
Baseten is part of the inference layer that matters because production model serving is increasingly its own product category, not just a backend detail.
San Francisco, US
Beautiful.ai matters because presentation creation is one of the clearest productivity categories where AI can improve output quality without requiring users to become designers.
Berkeley, US
BAIR is one of the most influential academic AI labs in the world, especially in robotics, multimodal learning, and agent research that often flows into startups and commercial labs.
Freiburg, Germany
Black Forest Labs quickly became one of the most important image-model companies by pairing the FLUX family with multiple distribution paths across API, open weights, enterprise licensing, and self-serve use.
Waltham, US
Boston Dynamics belongs on the map because physical AI is not only a startup story; it already has durable industrial deployment leaders.
Tel Aviv, Israel
Buildots matters because it applies AI to a high-friction industrial workflow where visual monitoring and schedule variance directly affect project outcomes.
Beijing, China
ByteDance / Doubao matters because it shows what scaled consumer AI deployment looks like when a major internet platform combines models, distribution, and media products.
Beijing, CN
Cambricon matters because it is one of China's most visible domestic AI chip companies and a recurring reference point in the push for local AI compute capacity.
Sydney, Australia
Canva matters because it pairs massive distribution with a practical creative suite, making it one of the most consequential AI adoption layers in design and workplace media creation.
Santa Clara, US
Celestial AI is notable because it attacks AI infrastructure at the interconnect level, betting that photonics can become a critical lever for scaling advanced AI systems.
Sunnyvale, US
Cerebras belongs on the map because compute and inference architecture increasingly shape who can train and serve important models in practice.
San Francisco, US
Claude Code is important enough to track independently because it is no longer just 'Anthropic, but for developers' - it is a distinct coding product with its own workflow, distribution, and mindshare.
New York, US
Clay matters because it is turning data enrichment and outbound operations into an AI-native workflow surface for modern go-to-market teams.
San Francisco, US
Codex deserves its own slot in the map because it competes in the agentic coding product layer, where workflow quality and developer trust matter as much as the underlying model provider.
Toronto, Canada
Cohere stands out as an enterprise-first model company, prioritizing retrieval, secure deployment, and business workflows over broad consumer reach.
Livingston, US
CoreWeave is one of the most important AI infrastructure companies, selling a purpose-built cloud for training and inference to model labs and enterprises that need GPU-heavy capacity fast.
San Francisco, US
Credo AI is building the governance layer for enterprises that need to evaluate, approve, and manage AI systems under policy, legal, and risk constraints before broad deployment.
Denver, US
Crusoe matters because AI infrastructure is now a product and capital-allocation story, not just a hosting detail, especially when energy and data-center strategy are central constraints.
San Francisco, US
Cursor is one of the clearest examples of the new agentic coding layer: a developer product built on top of frontier models, but increasingly differentiated by workflow, UX, and team adoption.
Santa Clara, US
d-Matrix is one of the more serious specialist inference-chip companies, focused on the bottlenecks that show up when serving large models at datacenter scale.
San Francisco, US
Databricks is turning its data platform into a full AI operating layer, pairing Mosaic AI and DBRX with the broader Data Intelligence Platform for enterprise model building and deployment.
Boston, US
DataRobot remains relevant because enterprise AI adoption still depends heavily on platforms that promise control, governance, and operational reliability rather than frontier-model novelty alone.
Cologne, Germany
DeepL matters because translation remains one of the clearest enduring AI use cases, and DeepL has become one of the strongest standalone language-product companies.
Hangzhou, China
DeepSeek is one of the most important open-model challengers, pairing aggressive research releases with an API business that keeps pressure on closed-model pricing and benchmark leadership.
San Francisco, US
Descript belongs on the map because it turned AI editing into a practical creator workflow rather than a separate experimental tool.
San Francisco, US
Edra is interesting as an enterprise knowledge-layer company, especially if the map wants to capture firms turning internal know-how into AI-native workflows.
Santa Clara, US
Eightfold AI matters because recruiting and talent operations remain one of the most natural enterprise AI workflow categories, and Eightfold is one of the best-known pure-play companies there.
New York, US
ElevenLabs has become a core voice infrastructure layer for creator, media, and product teams, expanding from speech synthesis into conversational agents, dubbing, and transcription.
San Francisco, US
ELSA matters because it represents a focused, durable AI use case where speech feedback is the product rather than a supporting feature.
Saratoga, US
Eridu is one of the newer infrastructure startups attacking the networking layer behind large-model systems rather than only chips or hosting.
Menlo Park, US
Etched stands out because it is making a very opinionated bet: that dedicating silicon to transformer workloads can beat the economics of general GPUs for frontier inference.
San Francisco, US
Factory represents the next logical step after AI coding assistants: turning the tool into an orchestrated software-production system for teams.
Sunnyvale, US
Figure is one of the clearest attempts to merge humanoid hardware, embodied models, and industrial deployment into one vertically integrated robotics company.
San Francisco, US
Fireflies.ai matters because meetings remain one of the most obvious workflow categories where transcription and summarization create immediate value for teams.
Redwood City, US
Fireworks AI is one of the more important independent inference platforms, helping developers run, customize, and scale model workloads without building the full serving stack themselves.
San Francisco, US
Gamma matters because it is one of the clearest examples of AI-native document creation becoming a practical everyday workflow, not just a demo category.
Palo Alto, US
Genspark is worth tracking because it is one of the newer entrants trying to turn search, research, and multi-model orchestration into a single AI workspace product.
Boston, US
Ginkgo Bioworks is trying to turn its large biological foundry and data engine into a durable AI advantage for biological design, drug discovery, and engineered-organism workflows.
San Francisco, US
GitHub Copilot matters because distribution is part of the product: it can turn model capability into default workflow inside the tools many software teams already use every day.
Palo Alto, US
Glean is one of the clearest enterprise AI companies on the board because it turns internal company knowledge into a live search and agent layer.
London, UK
Google DeepMind combines world-class research depth with mass distribution through Google products, making it one of the most consequential labs for frontier multimodal systems, scientific AI, and robotics.
Bristol, UK
Graphcore remains one of the most recognizable non-NVIDIA AI chip companies, with a long-running effort to offer a different processor architecture for machine intelligence.
San Jose, US
Groq is positioning itself as the speed-first inference stack for production AI, using its custom LPU architecture and GroqCloud to compete on latency and cost rather than training breadth.
Paris, France
H Company stands out because it is trying to bring frontier-style agentic systems directly into enterprise workflows rather than stopping at research demos.
San Francisco, US
Harvey matters because it is one of the clearest examples of an AI-native vertical platform breaking into a high-value professional workflow.
Los Angeles, US
HeyGen matters because business video generation is becoming a real workflow category, not just a novelty media demo.
Palo Alto, US
Hippocratic AI is worth tracking because it is trying to become a safety-first healthcare agent platform rather than a general-purpose model vendor.
New York, US
Hugging Face matters not just as a company but as distribution infrastructure for the open-model ecosystem, spanning model hosting, inference, datasets, and tooling.
Armonk, US
IBM is positioning Granite and watsonx as a governed enterprise AI stack for large organizations that care as much about control, compliance, and hybrid deployment as raw benchmark leadership.
Palo Alto, US
Inception Labs is notable because it is not just another LLM company; it is using diffusion language models to compete on speed, coding, and inference economics.
Hong Kong
Insilico Medicine matters because it is one of the most prominent efforts to turn AI for science into a repeatable drug-discovery business.
Dublin, Ireland
Intercom is worth tracking because Fin gives it a strong position in the race to make customer support one of the first large-scale enterprise agent categories.
London, UK
Isomorphic Labs matters because it represents one of the strongest attempts to turn frontier AI research into a drug-discovery company with real pharma leverage.
Austin, US
Jasper matters because it was one of the first breakout AI application companies and remains a useful signal for how enterprise content tooling is consolidating.
San Francisco, US
Kana looks relevant because it is applying the agentic-software pattern to a concrete business function instead of shipping yet another general chatbot.
Orlando, US
Kore.ai is a serious enterprise platform company rather than a consumer AI app, which makes it relevant anywhere the map tries to represent production AI inside large companies.
San Francisco, US
Krea is one of the more credible independent AI-creative products because it is trying to be a full workflow layer for artists and visual teams, not just a single image model demo.
Paris, France
Kyutai is a nonprofit research lab pushing open speech and multimodal systems, with an explicit mission to democratize advanced AI through open science.
San Francisco, US
Lambda sits in the middle of the AI compute boom, packaging GPU cloud, clusters, and AI factory infrastructure for teams that need training and inference capacity without building it from scratch.
San Francisco, US
LangChain sits in the workflow layer between raw models and working agents, giving teams orchestration, observability, and production controls.
Dubai, AE
LEAP 71 belongs on the map if physical AI includes the upstream design stack for real-world machines, not just robots and autonomy systems.
New York, US
Lemonade matters because it is one of the clearest examples of AI being used as an operating system for a regulated consumer business rather than only as a model layer.
Palo Alto, US
Lepton AI matters because global GPU coordination is becoming a product in its own right, especially for teams that need portability across providers.
San Francisco, US
Letta matters because memory and state management are becoming core infrastructure questions for agentic systems, not just product features.
Boston, US
Levangie Labs is trying to move agents beyond chat loops and into persistent cognitive systems that can operate autonomously over long-running work.
Jerusalem, Israel
Lightricks sits at the intersection of creator software and generative media infrastructure, using products like LTX Studio and the LTXV model line to push deeper into AI filmmaking workflows.
Cambridge, US
Liquid AI is commercializing compact model architectures built for efficient deployment, positioning itself around lower-latency, smaller-footprint systems for enterprise and edge settings.
San Francisco, US
Logical Intelligence is an early research lab focused on energy-based reasoning systems for settings where verification and correctness matter more than generic chatbot fluency.
Palo Alto, US
Luma AI is moving from standout video generation into a broader creative-agent strategy, combining multimodal research with products for image, video, audio, and production workflows.
San Francisco, US
Magic belongs on the map as a model-first coding company, not just another developer tool, because it is trying to push the underlying research frontier.
Singapore
Manus matters because it is one of the more visible examples of the current push from chat interfaces toward autonomous task execution.
Menlo Park, US
Meta AI is unusual on the map because it combines consumer-scale product surfaces with one of the strongest open-weight strategies in frontier AI.
Redmond, US
Microsoft’s AI position comes from product and cloud distribution at scale, not just model development, which makes it one of the most durable deployment players on the board.
Distributed
Midjourney is still one of the strongest creative AI brands because it built a highly engaged community around image generation and is now expanding that identity into broader media tooling.
Montreal, Canada
Mila remains one of the deepest academic AI talent and research hubs, shaping both foundational research and the broader commercial ecosystem from Montreal.
Shanghai, China
MiniMax belongs on the board because it combines frontier multimodal ambitions with consumer product reach, especially in markets outside the usual U.S.-centric map defaults.
Paris, France
Mistral AI sits at an important intersection of frontier research, enterprise adoption, and European strategic positioning in AI.
New York, US
Modal sits in the infrastructure layer for teams that want to run serious AI workloads without owning the full orchestration stack themselves.
San Francisco Bay Area, US
Modular matters because it is trying to collapse fragmented AI infrastructure into one developer-facing platform instead of competing at only one layer like hosting or chips.
Beijing, China
Moonshot AI is one of the highest-visibility Chinese frontier model companies, using the Kimi product line to compete on consumer reach, long context, and increasingly broader reasoning workflows.
Amsterdam, Netherlands
Nebius matters as part of the emerging AI cloud layer, where infrastructure providers are trying to win on specialization rather than general-purpose cloud breadth.
New York, US
Nomic is a good example of how AI product strategy is fragmenting into domain-specific systems with deep workflow context rather than one-size-fits-all assistants.
New York, US
Nous Research matters because it sits at the intersection of open model development, frontier-style experimentation, and a highly engaged builder community.
Santa Clara, US
Nvidia belongs on the map as more than a chip company because it increasingly ships the reference stack for model serving, robotics, and enterprise AI deployment.
San Francisco, US
OpenAI remains the reference point for much of the market because it combines frontier model development with mass user distribution through ChatGPT, strong API adoption, and fast-moving coding and media products.
Unknown
OpenClaw is worth watching because open agent stacks are becoming an important counterweight to closed consumer agent products, even though the category is still early and noisy.
Mountain View, US
Otter.ai is best understood as a workflow product rather than just a transcription tool, using meeting capture, summaries, and action tracking to own a slice of organizational memory.
Scottsdale, US
Paradox is worth tracking because recruiting is one of the clearest enterprise-agent categories, and Olivia has been a durable product in that space.
Paris, France
Pathway is unusual because it now presents itself as both a frontier-model company and an infrastructure layer for continual, time-aware AI systems rather than a conventional enterprise AI wrapper.
San Francisco, US
Perplexity has become one of the strongest AI-native search and research products by turning web-grounded answers into a consumer habit, then expanding that behavior into deeper research and API surfaces.
San Francisco, US
Physical Intelligence is one of the clearest examples of frontier model thinking moving into robotics, with a thesis centered on transferable skills across hardware.
Palo Alto, US
Pika is a consumer-first video company that matters because it keeps lowering the barrier to shareable AI video creation without requiring pro tooling.
San Francisco / Paris
Poolside sits at the seam between frontier-model ambition and enterprise software delivery, using software engineering as its beachhead.
Hangzhou, China
Qwen matters because Alibaba has turned it into both an open-model family and a cloud distribution channel, making it one of the strongest bridges between open weights and enterprise deployment in Asia.
Salt Lake City, US
Recursion is one of the clearest examples of AI-native biotech at scale, using proprietary biological data and machine learning systems to accelerate drug discovery and platform partnerships.
New York, US
Reflection AI is an early frontier lab trying to combine high-end research ambitions with an open-intelligence framing rather than a closed product-only strategy.
San Francisco, US
Reka is one of the more interesting frontier companies because it pushes a modular, real-world framing instead of positioning itself purely around benchmark theater.
San Francisco, US
Replicate matters because it lowers the friction between open-source models and real application use, turning model experimentation into API consumption.
Foster City, US
Replit is notable because it connects coding assistance to runtime, collaboration, and deployment, which makes it more than just a model wrapper.
Wilmington, US
Runpod belongs on the map because low-friction GPU infrastructure has become an important competitive layer for developers and smaller model teams.
New York, US
Runway sits near the center of AI video tooling, combining frontier generative media research with a polished product stack for creators, studios, and API users.
Palo Alto, US
SSI matters less as a product company today than as a signal of where frontier research capital and talent are concentrating.
Tokyo, Japan
Sakana AI is one of the more intellectually distinctive research companies on the board, exploring evolutionary and modular approaches instead of simply scaling the standard frontier recipe.
Palo Alto, US
SambaNova sits in the same critical infrastructure conversation as other AI compute challengers, where hardware and serving stacks are becoming strategic product categories.
Bengaluru, India
Sarvam AI matters because it is building a sovereign AI stack tailored to India rather than merely repackaging global frontier models.
San Francisco, US
Scale AI sits in the infrastructure layer between frontier models and enterprise deployment, turning data, evaluations, and workflow tooling into a product surface.
San Diego, US
Shield AI is one of the clearest examples of real-world AI deployment in defense autonomy, pairing hardware systems with its Hivemind autonomy stack.
San Francisco, US
Sierra matters because it is trying to define AI-native customer experience infrastructure, not just bolt automation onto legacy support software.
San Jose, US
SiMa.ai sits at the intersection of edge AI hardware and physical AI deployment, which makes it more relevant than generic accelerator startups for real-world autonomous systems.
Pittsburgh, US
Skild AI is chasing a shared robot-brain thesis, arguing that one adaptable model layer can generalize across many physical embodiments and deployment environments.
Bozeman, US
Snowflake is trying to turn its data cloud into an AI execution layer, combining Cortex AI and the Arctic model family to keep enterprise AI workloads inside the broader Snowflake platform.
San Francisco, US
Speak matters because language learning is one of the clearest consumer AI categories where voice interaction and personalization actually improve the core product experience.
London, UK
Stability AI still matters because its open-media brand and installed user base give it influence beyond its current organizational turbulence, especially in image and video generation.
Stanford, US
Stanford CRFM is one of the field's most important academic centers for evaluating and governing foundation models, with HELM serving as a flagship benchmark effort.
Shanghai, China
StepFun is worth tracking because it is part of the fast-changing Chinese frontier-model field, where multimodal and agentic positioning can shift quickly.
Cambridge, US
Suno helped define the consumer AI music wave by turning text-to-song generation into a product people return to, not just a novelty demo.
San Francisco, US
Supermemory is part of the emerging memory layer for AI products: the systems that make agents and applications feel persistent, context-aware, and production-ready.
London, UK
Synthesia is a major media-model company because it turned avatar video from a demo category into a real enterprise workflow product.
Toronto, CA
Taalas is interesting because it is making one of the clearest hardware-native arguments against general-purpose GPU economics for inference.
Santa Clara, US
Tenstorrent belongs on the map because compute architecture increasingly shapes who can train and serve important AI systems in practice.
San Francisco, US
Thinking Machines Lab is emerging as a frontier lab with a real product angle, not just a talent story, through customizable AI systems and research tooling.
San Francisco, US
Together AI matters because it turns open-model usage into a production cloud layer, not just a research or hobby workflow.
London, UK
Tractable matters because it represents a durable enterprise AI pattern: narrow, high-value automation in an industry where speed and accuracy directly affect operations.
San Francisco, US
Typeface matters because it is trying to turn generative media into an enterprise-safe, brand-controlled content workflow rather than a standalone creative novelty.
New York, US
Udio is worth tracking because music is one of the clearest consumer AI categories, and Udio is one of the few companies directly competing for that frontier.
Hangzhou, China
Unitree is important because it combines aggressive product cadence, broad visibility, and real retail availability in legged and humanoid robotics.
Cambridge, US
Unreasonable Labs is notable because it is aiming squarely at AI for science and knowledge discovery, not just lab copilots or generic RAG systems.
San Francisco, US
v0 sits at the edge of coding tools and app builders, which is exactly why it deserves a place on the map: it captures the shift from coding assistant to software-construction product.
Los Angeles, US
Vast.ai is a useful map entry because it captures the long tail of GPU access: not every serious AI workload runs on a hyperscaler or a premium AI cloud.
San Francisco, US
Vercel is extending its developer platform into AI app creation and deployment, using v0, the AI SDK, and agent infrastructure to own more of the path from prototype to production.
Toronto, Canada
Waabi stands out because it combines frontier AI talent with a simulation-first autonomy strategy aimed squarely at commercial trucking.
Mountain View, US
Waymo is the clearest example that physical AI already has real deployment at scale, not just prototypes and demos.
San Francisco, US
Weights & Biases remains a core developer platform for training, evaluation, and model operations, which makes it important even as more AI companies try to absorb those workflows into broader stacks.
Mountain View, US
Windsurf is part of the new coding-tool wave where the differentiator is not the base model alone, but the quality of the software-production loop wrapped around it.
New York, US
Wiz matters because AI-era infrastructure adoption is increasing the need for multi-cloud security, posture management, and runtime visibility at enterprise scale.
San Francisco, US
World Labs is one of the most visible spatial-intelligence companies, aiming to build models that can perceive, generate, and interact with the 3D world instead of operating only over text or flat media.
San Francisco, US
Writer is building an enterprise AI stack around its own Palmyra models, agent tooling, and workflow products rather than acting only as a wrapper around third-party APIs.
San Francisco, US
xAI is using Grok, the xAI API, and distribution through X to compete as a consumer-facing frontier lab with unusually tight product feedback loops.
Palo Alto, US
You.com has become more important as an infrastructure company than as a consumer search challenger, which is why it belongs on the board even if the consumer product gets less attention now.
San Francisco, US
Yutori Scouts is worth tracking because it represents the emerging category of always-on personal and research agents that watch for changes instead of only responding on demand.
Beijing, China
Z.ai remains one of the more important Chinese frontier model efforts, using the GLM family and a broad assistant surface to compete across chat, reasoning, coding, and multimodal workflows.