IMPLICATOR
💰 Fresh Funding
Geordie raises $30M to secure and govern enterprise AI agents
Visit Geordie AI →Geordie AI raised a $30 million Series A led by Balderton Capital, the London and New York startup told Fortune in a May 28 exclusive, money to scale an independent security layer for enterprise AI agents. The round runs about 4.6 times its $6.5 million seed from September 2025 and, by Fortune's calculation off UK Companies House filings, values Geordie at roughly $180 million post-money.
Crosspoint Capital joined as a new backer, alongside follow-on money from General Catalyst and Ten Eleven Ventures, bringing total funding to $36.5 million. Geordie's platform discovers every AI agent running across a company, on laptops, in the cloud, inside SaaS platforms and the codebase, then maps the data each can reach and flags risk in real time. CEO Henry Comfort calls the role "the Switzerland of the future" and air traffic control for enterprise agents, and points to a market gap to match, noting that 82% of enterprises already run AI agents while only 44% have policies to secure them.
Comfort sells Geordie as the neutral layer, yet he told Fortune the principal risk is whether "incumbents eventually use their distribution advantage to great effect in this space," and incumbents like OpenAI are already shipping systems that manage rival vendors' agents. Geordie's founders came out of Darktrace and Snyk, the kind of established vendor with that same distribution muscle. The clearest proof so far comes from a single deployment. AI biotech Owkin ran Geordie, found three times more agents than it knew about, and put mitigated risk exposure at $12 million to $13 million by its own methodology, which Fortune notes is not independently audited. Geordie employs 37 people and expects about 50 within three months, the headcount the new money is meant to buy.
💰 Fresh Funding
Tensormesh raises $20M to cut AI inference costs with KV caching
Visit Tensormesh →AMD Ventures, CoreWeave and Nvidia's NVentures all backed Tensormesh in a $20 million round the San Francisco startup disclosed on May 27, a seed extension that brings its total funding to $24.5 million. The money lands about seven months after its $4.5 million seed and alongside the general-availability launch of Tensormesh Inference, the platform built to cut what it costs to run AI models.
Valley Capital Partners and Laude Ventures also joined, and the release names no lead. The product stores and reuses the intermediate KV cache that large language models build while reading a prompt, so GPUs stop recomputing system prompts, conversation history and tool definitions on every request. Tensormesh prices cached input tokens at $0 across serverless deployments, the company says, "not as a promotional rate, but as a permanent part of how Tensormesh prices its platform."
The company calls itself "the first company to bring [KV caching] to market as a fully productized, enterprise-grade platform" and frames the cached data as "a whole new class of data" it is "uniquely positioned to define." But the technique is not proprietary to it, and its headline claim of up to 10x cuts in latency and GPU spend matches the figure TechCrunch reported for the open-source LMCache utility in October 2025, by which point a 10x inference-cost cut was already a familiar claim. LMCache, the project Tensormesh commercializes, now plugs into vLLM, Nvidia Dynamo, AWS SageMaker and Oracle OCI, putting the same caching layer inside the rival stacks Tensormesh is pitching against. NVentures backed it even though that same open-source layer already runs inside Nvidia's own Dynamo stack. The $20 million now has to prove enterprises will pay for hosting and support rather than run the open-source version themselves.
💰 Fresh Funding
Orbital Industries raises $50M to turn AI-designed materials into data-center hardware
Visit Orbital Industries →Orbital Industries, the London and San Francisco startup recently rebranded from Orbital Materials, raised a $50 million Series B led by Plural, Fortune reported in a May 28 exclusive, with Nvidia's NVentures, Radical Ventures, Compound and Fly Ventures joining. The round lifts the company's total raised to $71 million, more than triple the roughly $21 million it had gathered across a 2023 seed and a February 2024 Series A.
No valuation was disclosed. Founded by former DeepMind researcher Jonathan Godwin and running on about 50 people, Orbital uses an in-house AI model called Orb to simulate the quantum-mechanical behavior of atoms, then sells the hardware it discovers rather than licensing the science to incumbents like BASF or PPG. Its first two products are a PFAS-free cooling fluid for high-density AI GPUs and a modular data-center unit, Nova Array, that Orbital says can move from purchase order to power-on in 24 weeks against up to three years for conventional builds. Orb, published open-source on GitHub, runs about 10 times faster than the nearest alternative, the company says.
"Frontier AI gives us PhD-level expertise across every discipline... so what used to take a decade, we can now do in months," Godwin said. Resilience Media reports the cooling technology is "currently in development," Nova Array has "no signed customers yet," only "active commercial conversations," and the 24-week timeline is "not atypical" for modular systems. Godwin concedes to Fortune that after years of AI applied to drug discovery, "no AI-discovered drug candidate has yet to make it through clinical trials and on to the market." The $50 million now funds commercial deployment of products that have not yet commercially deployed, and the first test is whether a paying customer signs for the cooling fluid.
Get Implicator.ai in your inbox
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
Pace, a New York startup building AI agents for insurers' back-office work, raised a $46 million Series B co-led by Thrive Capital and Sequoia Capital, the company said in a May 27 release. The round lands about four months after its $10 million Sequoia-led Series A and, Forbes reported, values Pace at $375 million, a figure in Anna Tong's byline rather than the company's own disclosure.
Pace's agents read across documents, work inside insurers' internal applications and place phone calls to handle submission intake, policy servicing, claims and data entry, the kind of work that has made insurers a prime target for agentic automation. The company says those agents have autonomously completed more than 250,000 insurance workflows since launch, cut claim cycle times by 30% at customer Ryze Claim Solutions, and automate thousands of hours of manual work at Prudential.
Founder and CEO Jamie Cuffe, who grew up around Lloyd's of London, frames the mission as closing what Pace calls the $9 trillion global insurance protection gap. In the January round he was blunter about the mechanism, saying "all of this work that was being outsourced offshore can now be outsourced to AI," and Forbes described the agents as handling the dull work insurers have long shipped to offshore operators.
Thrive partner Philip Clark framed the bet around the founder, calling Cuffe "one of those people where you go in biased to saying yes." Forbes marks a company of about 28 employees at $375 million, roughly $13 million per head about a year into shipping, and the 250,000-workflow figure counts tasks rather than disclosed revenue. The Series B has to turn its Ryze and Prudential pilots into a recurring book before that number looks earned.
Daloopa raised $47 million in Series C funding led by Brighton Park Capital, the New York company said in a May 28 release, to expand the platform that turns public filings into structured financial data for AI systems. The round runs roughly 2.6 times its $18 million 2024 Series B and lifts cumulative funding above $100 million, though the company disclosed no valuation.
Squarepoint Capital, Touring Capital and Nexus Venture Partners joined the round. Daloopa's customers started as hedge funds, but it says over half of its new opportunities now come from asset managers, banks and other institutions, and its data increasingly feeds the AI agents doing investment research.
The funding pitch leans on Daloopa's own FinRetrieval benchmark, which put three frontier agent systems through 500 finance questions. Anthropic's Claude hit 91% accuracy when it could query Daloopa's structured data through MCP, against 20% on web search alone, the gap the company tops out at "71 percentage points."
Daloopa's own research blog walks that back. "For production finance work, 90% accuracy is not fully dependable or delegatable," it says, and web-only accuracy across the test "varies widely (20-71%)," so the lift shrinks against a strong baseline. Lead-round participant Squarepoint is a quantitative investment firm, and Daloopa notes that for some backers "firsthand experience with the technology has also led to investment conviction," so some of the 160 financial institutions it counts as customers are also investors.
"It's no longer enough for models to simply generate answers; they must be accurate and fully traceable," CEO Thomas Li said. Coverage now spans more than 5,500 public companies, a base Daloopa keeps widening, and the company says it doubled revenue over the past year without putting a number on it.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: [email protected]
