Beyond OpenAI: The companies wiring enterprise AI

the companies – OpenAI, Anthropic, Google DeepMind and Nvidia dominate headlines, but the enterprise AI economy is being built by less visible companies: model challengers in Europe and Japan, database and data-intelligence platforms, agent security and governance startups, a

For years, the AI story has been told like a single race with a few famous runners. OpenAI’s ChatGPT became the fastest consumer application in history to reach 100 million users. Nvidia’s chips power virtually every major AI training run in the world. Anthropic’s Claude is one of the most trusted model families for enterprise AI. And Google DeepMind’s AlphaFold solved a protein-folding problem that stumped scientists for 50 years. earning CEO Demis Hassabis a Nobel Prize in Chemistry in 2024.

But as more companies try to turn AI into something they can sell, operate, and defend, the plot widens. A different ecosystem is assembling the infrastructure. governance layers. and application platforms that make AI commercially viable—often without the same spotlight. When DeepSeek rattled markets last year, the lesson wasn’t only about model efficiency. It was a reminder that serious AI work has moved beyond a handful of famous labs.

What follows is a look at companies building the rest of the stack—where the decisions are harder, the timelines longer, and the value can be just as real.

Silicon Valley has long treated frontier AI as a homegrown invention. Mistral AI is challenging that assumption from Paris. Founded in April 2023 by Arthur Mensch. formerly of Google DeepMind. along with Guillaume Lample and Timothée Lacroix. both formerly of Meta. the company built early momentum on a claim that frontier-quality models did not require the compute budgets of American hyperscalers.

Mistral’s open-weight models were released for anyone to build on, and the business grew fast. Revenue rose from roughly $10 million in 2023 to more than $400 million in annualized recurring revenue by early 2026. In September 2025, Mistral closed a €1.7 billion Series C led by ASML, pushing its valuation to €11.7 billion.

The company has since launched Vibe. described as an agentic platform for research. drafting. and code deployment. and it is exploring its own chip design. Mensch put the emphasis plainly: “Physical infrastructure control matters just as much as underlying model quality for capturing long-term value.”.

Tokyo-based Sakana AI is pushing the frontier further still. Founded in 2023 by David Ha. who led Google Brain’s research team in Japan. and Llion Jones. one of the co-authors of the influential research paper Attention Is All You Need. Sakana is built on the idea that AI systems can evolve like nature does—through collective intelligence and constant adaptation rather than raw scale.

In March 2025. its AI Scientist system became the first AI to have a paper it independently wrote accepted by peer reviewers at a premier machine learning conference. In November. Sakana raised $135 million at a $2.65 billion valuation. with backing from MUFG. Lux Capital. and In-Q-Tel. the CIA’s venture arm. Ha has argued that “innovation, based on constraints”—not unlimited compute—is where the real opportunity lies.

Earlier this week, Sakana moved Fugu and Fugu Ultra out of beta into general availability. The system routes tasks across a pool of frontier AI models instead of training one from scratch. all through a single interface. Its design leans on a recent real-world warning sign: on June 12, the U.S. Commerce Department issued an export control directive that forced Anthropic to shut down Claude Fable 5 and Claude Mythos 5. just three days after launch. As of this writing, neither has been restored.

Sakana pointed to the vulnerability directly in its launch announcement: “Relying on a single company’s APIs for critical infrastructure, finance, or governance is a material vulnerability.” The company added that the risk “is no longer a hypothetical possibility, but a reality.”

In its own benchmark comparisons. Sakana’s system measures Fugu Ultra against Fable 5 and Mythos Preview. using whichever score is higher on a given test. Neither model is part of Fugu’s agent pool, because neither is publicly accessible. The comparison excludes Mythos 5. the more capable model suspended alongside Fable 5. with independent analysis showing it outperforming Fugu Ultra on several of the same benchmarks.

What powers enterprise AI at scale isn’t only the frontier model at the top of the stack. It’s the infrastructure beneath it—the layer that decides whether AI applications can access, reason over, and act on data in ways that produce real business outcomes.

For many enterprises, the problem begins before AI code is written. Organizations adding AI to existing systems are often told to move data out of relational databases—the traditional table-based systems that power most business software—and into new infrastructure. In practice, that can mean multi-year projects. RavenDB CEO Oren Eini says the pace has made those timelines dangerous. “At the current pace of AI, a two-year project is already too late,” he says. “By the time you ship, the rest of the market has already moved on.”.

With RavenDB, Eini says projects that previously stretched past six months are now being completed in weeks. He argues the bottleneck usually isn’t the AI model. It’s the infrastructure around it. Most business databases were never built to let AI search and reason over their data. Even databases that offer a vector index—allowing search by meaning instead of exact keywords—only solve part of the problem. “If your database only provides a vector index, all of that burden falls on you,” he explains.

RavenDB. founded in 2009. has spent the last two years embedding vector search. AI agents. and generative AI capabilities directly into the database layer. The aim is to let businesses build on top of what they already run instead of replacing it. “You shouldn’t have to rebuild your systems to use AI,” Eini says.

Alation is tackling a related challenge one layer up through data intelligence. The platform counts Truist Bank, Sallie Mae, Cisco, and Daimler Truck North America among its more than 500 customers. Its starting point is blunt: many enterprise AI projects fail at the data layer. Gartner estimates that 85% of AI projects fail due to poor data quality or lack of relevant data.

CEO Satyen Sangani frames it as more than a technical problem. “An agent without business context is just a clever chatbot,” he says. “It can answer questions. It can’t drive an outcome that someone in finance, operations, or compliance will sign off on.”

He has repeatedly seen a specific failure mode: pilot purgatory. S&P Global found 42% of companies abandoned most of their AI initiatives in 2025, up from 17% the year before. The culprit, Sangani says, is that organizations often started with the technology and then searched for problems to solve. The ones that break through do the opposite.

“Every agent action, every human correction, every data product becomes knowledge that makes the next decision better,” Sangani says. “The companies that get this right become indispensable, because their value grows the longer customers use them.”

As enterprise AI adoption accelerates, governance has not always kept up. Between 2023 and 2024, the amount of corporate data uploaded or pasted into AI tools rose by 485%. Gartner projects that over 40% of AI-related data breaches by 2027 will stem from unapproved or improper generative AI use.

Reco, an Israeli agent security and AI governance company, was built to address that risk. Its platform maps the full AI agent footprint across enterprise SaaS environments. showing what is running. what it can access. and how it is behaving. CEO Ofer Klein tells Fast Company that across its customer base organizations typically discover roughly ten times more apps and a hundred times more agents than they expected when they first connect.

Security leaders who have learned to manage the sprawl have stopped asking “what AI tools are my employees using?” and moved to a more difficult question: if an agent has simultaneous access to Salesforce data, SharePoint files, and Slack channels, how do you stop a breach before it starts?

Reco’s answer is grounded in visibility—where AI actually lives inside the enterprise. “Autonomy without governance is exposure at scale,” Klein says. “Every major platform your enterprise already runs on is now also an agent platform. That is exactly where we show value.”

Terzo is making a parallel case in a different part of the stack: contracts. invoices. and purchase orders that govern how large organizations spend. According to World Commerce and Contracting. companies lose an average of 9.2% of annual revenue to poor contract management every year. not through fraud. but through missed rebates. lapsed pricing schedules. and supplier agreements nobody was monitoring.

Terzo’s platform connects contracts, invoices, purchase obligations, and ERP data, validating financial activity against contractual terms in real time. CEO Brandon Card says that a Fortune 50 company found more than $100 million in savings after running its supplier relationships through Terzo’s platform.

As AI agents increasingly handle procurement and financial workflows, trust becomes a data problem as much as a model problem. An AI agent operating from inaccurate or unstructured contract data doesn’t just make mistakes. It makes those mistakes at scale. Terzo is building what it calls a financial truth layer to prevent that.

Card’s pitch goes beyond cost savings. “The companies that win in the next decade will not just have AI models,” he says. “They will own the financial intelligence layer that allows autonomous agents to move trillions of dollars safely, accurately, and automatically across the global economy.”

Text has been the center of gravity for years. Most AI progress has happened in text—reading it, generating it, and searching through it. But voice and video are different: messy, real-time, and until recently, rarely stored and repurposed in a way that makes them useful.

The gap has been enormous and mostly invisible. Billions of customer service calls are placed every year, and most are logged and forgotten. Hundreds of millions of surveillance cameras run worldwide, and most of what they capture is never watched. The information exists. The tools to use it have lagged.

Two companies are now trying to close that divide—one in conversation, the other in physical space.

PolyAI. founded in London by Nikola Mrkšić. a Cambridge machine learning PhD and first engineer at VocalIQ. builds enterprise voice AI for organizations managing large volumes of customer interactions. The company counts more than 100 enterprise customers and runs more than 2,000 live deployments across 45 languages in over 25 countries. A Forrester study found its customers achieve 391% ROI with an average $10.3 million in savings. Deloitte recognized PolyAI as the U.K.’s fastest-growing AI company in 2025.

In December 2025, PolyAI raised $86 million in a Series D with Nvidia’s venture arm among the participants. Mrkšić said at the announcement. “PolyAI started with a simple idea: Enterprises should sound human.” He added. “We turned that idea into reality. and it led to something far greater: The emergence of the agentic enterprise.” He has also said publicly that within five years. 90% of contact center work will be automated.

Lumana is making the same shift in the physical world. Its AI platform transforms existing security cameras—including those in retail stores. healthcare facilities. and factories—into real-time intelligence systems without new hardware. Company CEO Sagi Ben Moshe says the system “understands what it sees—the behaviors. context. and anomalies—and does so in real time.” “Instead of recording footage and hoping someone reviews it later. the system understands what it sees. ” he adds. “A camera isn’t just capturing video anymore. It’s generating intelligence.”.

In under 18 months, Lumana scaled to more than 50,000 cameras. The company says its platform can reduce non-actionable physical access control alarms by as much as 95%. More recently. it has expanded beyond detection into agentic AI. with specialized agents that can monitor environments. investigate incidents. verify events. and initiate responses without requiring a human operator to continuously review video feeds.

Adam Scraba, Nvidia’s head of physical AI ecosystem and vision agents, describes Lumana as turning existing infrastructure into “a perceptive, highly scalable virtual workforce of video analytics AI agents that can understand, verify, and summarize what’s happening in the physical world.”

The goal is clear: cameras move from passive costs to active intelligence assets that can inform decisions about customer behavior, staffing, and workplace safety. At scale, that could make the physical world as legible to AI as the digital one.

The mainstream AI value story often runs in a straight line: powerful models emerge. developers build on top. enterprises adopt. and economic value follows. The middle—the infrastructure. governance. security. and operational intelligence that lets adoption happen safely—rarely gets its share of the spotlight.

McKinsey estimates AI could add up to $4.4 trillion annually to the global economy through enterprise productivity gains alone. But those gains do not arrive because a company signed an OpenAI contract. They arrive when data feeding models is governed and trusted. when agents inside enterprise systems are visible and accountable. when contracts are operationalized rather than stored. and when voice and physical environments are converted into actionable intelligence.

That is the work these companies are doing: building the pieces that models alone can’t replace. In earlier technology cycles, the winners weren’t always the brands with the loudest names. They were the ones whose products could not run without the rest of the stack. If history is any guide, the AI cycle may not be different.

enterprise AI AI infrastructure AI governance agent security RavenDB Alation Reco Terzo PolyAI Lumana Mistral AI Sakana AI Fugu Vibe Nvidia OpenAI Anthropic Google DeepMind