.svg)
We're thrilled to have led Alloy's US$8m (A$11.5m) Seed round, with pre-seed backers Blackbird and Airtree both returning, alongside leaders and engineers from OpenAI, Anthropic, Tesla, Waymo, Halter and Carbon Robotics, and several of Alloy's own customers. Only Canva, Airwallex and Lorikeet have brought Australia's three largest funds together before, each in later rounds. Alloy is the first to do it at Seed.
Every time a robot operates, it records everything it "sees" and "feels" through its sensors: cameras, telemetry, networking, GPS, etc. A single test, following a hardware change for example, produces hundreds of GBs of data. When a robot fails, an engineer can spend days, sometimes weeks, working out why. Often the same issue has come up before. The answer is in the data, just buried.
Historically, this workload was challenging but manageable. Three shifts have changed that. First, robots have moved from controlled environments (factory floors) to open environments (farms, roads, warzones), which create novel scenarios that demand far more validation. Second, the data produced per test has grown significantly, a consequence of open environments requiring richer sensors, and of sensors becoming cheaper and higher resolution. Third, as companies operate networked fleets, the comms data between robots scales exponentially. The result is a bottleneck inside validation teams, who can't keep up with demands from the business.
And the pool of companies feeling this pain is growing fast. Foundation models have made it possible for robots to generalise across tasks, learning from data rather than rules, unlocking categories that were previously too unpredictable to automate. Hardware costs have collapsed: LiDAR sensors that cost ~US$75k five years ago now sell for under US$1k. Robotics patents grew 5x in the decade to 2024, and VCs are funding 10x as many robotics companies as they were ten years ago.
Introducing Alloy
Alloy builds AI agents that analyse robot data to identify the root causes of failures. The platform connects a fleet's logs, telemetry, video and sensor data with the engineering context around it, including Slack, Jira and the other systems teams already use. Alloy's agents continuously scan for anomalies, regressions and recurring patterns. Engineers can investigate specific problems in natural language, while the agents proactively surface issues they might otherwise have missed. And through its native MCP server, Alloy connects fleet data to coding agents like Codex and Claude Code, giving engineers the full context behind every mission rather than a pile of disconnected raw files.
The results are dramatic. At Advanced Navigation, a full day of analysis after each field test now takes less than ten minutes. "The conversation has completely shifted," says Jai Castle, the company's product validation manager. "Instead of 'Can we get this done in time?', it's 'What else can we go after?'" At US drone startup DroneForge, an engineer suspected the wrong component was failing. Alloy showed both state estimators were fine and pointed to the real fault, before the misdiagnosis compounded.
Alloy is already used by teams across navigation, defence, drones, agriculture, maritime, humanoids, construction and medical robotics. Some run entire fleets through it. Others are designing their next robots around it from day one. The ambition is to become the data and analysis layer for this new generation of robotics companies, in the way Databricks did for cloud data, or Segment for web analytics.
The Team
Alloy is, at its core, an investment in solo founder and CEO, Joe Harris. Joe graduated from the same university scholarship program as Atlassian's founders, and, inspired by them, joined Atlassian as a software engineer. He then joined Eucalyptus as one of its first engineers, where he was rapidly promoted to Chief Commercial Officer, leading growth, marketing and product through the company's rise to its eventual A$1.6b acquisition.
When Joe raised his pre-seed in early 2025, we passed. We didn’t know Joe well at the time, and the round had moved quickly. That decision has nagged at us ever since. As we watched Joe go deeper into robotics, and as founders and investors we trust kept telling us he was one of the smartest people they knew, our conviction grew. We don't get many second chances at decisions we regret. We're grateful Joe gave us one.
Joe didn't come from robotics, and we spent real time on that question. What won us over was the depth he'd built in under a year, evident every time we pushed him on the product, the market or the customer. Pair that with a rare combination of technical foundations and commercial instincts, and we think Joe is exactly the founder to win this category.
What's next for Alloy?
The funding will support engineering hires, expansion in the United States, and further development of Alloy's models and agent platform. The longer-term vision is bigger: becoming the default home of a robotics company's operational data, so that every run a fleet completes makes the next robot better. Robotics is one of the hardest markets to build in, and the teams that win will be the ones that learn fastest from their own data.
We couldn't be more excited to partner with Joe and the Alloy team.
