London-based artificial intelligence startup Mantic has raised $25 million after its AI forecasting system outperformed human forecasters in a major prediction competition, highlighting growing investor interest in using artificial intelligence to predict real-world events.
Artificial intelligence has already transformed how people write, design, code and search for information. Now, one startup is betting that AI can become equally powerful at answering a much harder question: what happens next?
London-based AI startup Mantic has raised $25 million in seed funding after its artificial intelligence forecasting technology demonstrated strong results against human forecasters in the 2026 Metaculus Cup.
The funding round was led by Radical Ventures, with participation from Microsoft’s venture capital arm M12, Thinking Machines Lab and Balderton Capital, according to Reuters. Mantic has not publicly disclosed its valuation. (Reuters)
The investment comes as the technology industry explores new applications for artificial intelligence beyond traditional generative AI tools.
For Mantic, that next frontier is AI forecasting.
What is Mantic AI?
Mantic is a London-based artificial intelligence company founded in 2024 by Toby Shevlane and Ben Day.
The company is developing AI systems designed to make predictions about uncertain real-world events involving areas including business, geopolitics, technology, government policy and culture.
Mantic says its systems focus primarily on events between approximately one week and one year into the future. (Mantic)
Instead of simply answering questions using information that already exists, Mantic’s technology attempts to estimate the probability that a future event will happen.
That distinction is important.
A conventional chatbot might explain why an election matters, for example. A forecasting system could instead be asked to estimate the likelihood that a particular candidate will win.
The AI must research available evidence, weigh competing signals, deal with uncertainty and assign probabilities to different outcomes.
This area is sometimes described as judgmental forecasting.
How Mantic AI beat human forecasters
Mantic gained significant attention following its performance in the 2026 Metaculus Cup, a forecasting competition involving predictions about future political, economic and cultural developments.
Metaculus is an online forecasting platform that evaluates predictions by comparing the probability assigned to an event with what eventually happens.
Rather than asking contestants to simply choose between “yes” or “no”, forecasting competitions reward participants for assigning well-calibrated probabilities.
Mantic’s system performed better than the human participants highlighted in the Reuters report, demonstrating how rapidly AI-based forecasting technology has developed. (Reuters)
It is a notable development because Mantic itself previously described expert human forecasters as the benchmark that machines were attempting to reach.
The company’s earlier results had already shown progress. Mantic says its system had previously ranked fourth among 539 human forecasters in the Metaculus Cup. (Mantic)
Its more recent performance suggests that the gap between automated forecasting systems and highly skilled human forecasters has continued to narrow.
The advantage: avoiding the crowd
One of the challenges in forecasting is avoiding what can be described as herd behaviour — allowing the prevailing consensus to have too much influence over a prediction.
Mantic’s performance demonstrated that an AI system can sometimes reach a different conclusion from the human consensus and ultimately prove more accurate.
Reuters cited forecasts involving subjects ranging from Colombian politics to popular music as examples where Mantic’s system differed from the prevailing human view and performed well. (Reuters)
This matters because good forecasting is not simply about collecting as much information as possible.
Forecasters must determine which information is useful, distinguish genuine signals from noise and continuously update their probability estimates when circumstances change.
AI systems could potentially perform that process across enormous numbers of questions simultaneously.
Why investors put $25 million into Mantic
Mantic’s $25 million funding round reflects the commercial potential investors see in AI-powered forecasting.
Businesses make major decisions based on forecasts every day.
Investment firms attempt to anticipate market developments. Multinational companies monitor geopolitical risks. Supply-chain managers assess possible disruptions. Governments evaluate economic, security and policy scenarios.
Even relatively small improvements in forecasting accuracy could therefore be valuable.
Reuters reported that companies and government agencies around the world have shown interest in Mantic’s technology, with some organisations already deploying its forecasting systems. The company has not publicly identified those customers. (Reuters)
Financial institutions could be particularly interested.
Trading companies and hedge funds constantly make decisions involving probabilities, whether they are estimating interest-rate changes, election outcomes, commodity movements or geopolitical events.
An AI system capable of conducting research and generating updated forecasts at scale could become another source of intelligence for those organisations.
Mantic’s technology is not a crystal ball
Despite its strong performance, Mantic’s technology should not be interpreted as an AI capable of knowing the future with certainty.
Forecasting is fundamentally based on probabilities.
A forecast assigning an event a 70% chance of occurring still leaves a 30% chance that something else happens.
Unexpected political developments, natural disasters, technological breakthroughs or new information can rapidly change an outlook.
The significance of Mantic’s results is therefore not that artificial intelligence can perfectly predict future events.
Rather, the results suggest that AI is becoming increasingly competitive at a type of reasoning previously associated with specialist human forecasters.
AI forecasting could become the next major AI race
Most of the recent artificial intelligence boom has centred on generative AI — systems capable of creating text, images, video and computer code.
Forecasting represents a different opportunity.
Instead of asking AI only to generate information, organisations could increasingly ask AI to estimate the likelihood of future outcomes before making important decisions.
Mantic says its goal is to bring automation to judgmental forecasting in the same way computing helped transform fields such as weather forecasting.
The company previously raised $4 million in pre-seed funding, led by Episode 1, with backing that included trading firm DRW and investors connected to the wider AI research industry. (Mantic)
Its new $25 million round gives the startup significantly more capital to improve its technology and expand its use.
Mantic is not operating in an empty field either.
Metaculus has been benchmarking AI forecasting systems against strong human forecasters, reflecting wider efforts to measure whether machines can consistently improve on human judgment. Its AI forecasting benchmark series has included dozens of competing systems. (Metaculus)
The bigger question is now whether success in forecasting tournaments can translate into consistently useful predictions in real-world business and government decision-making.
For investors backing Mantic, the opportunity is clear: if artificial intelligence can learn not only to analyse what has already happened but also provide better estimates of what could happen next, AI forecasting could become one of the technology industry’s next major commercial frontiers.
Discover more from LN247
Subscribe to get the latest posts sent to your email.
