The aftershock has stopped. The lights are flickering back on. And somewhere on the fourteenth floor of a tower in downtown San Francisco, a fire-evacuation crew is staring at a question no one trained them to answer fast: is this building safe to walk back into, or not? For most of human history the honest reply was a shrug and a wait for an inspector. Safehub decided the shrug was unacceptable.
This is a company that sells, of all things, certainty - or at least the closest thing to it that physics will allow. Bolt a small sensor to a building, connect it to the cloud, and within minutes of an earthquake the building files its own damage report. Not the neighborhood's. Not the fault line's. The building's. It is a deceptively modest pitch for a deceptively enormous problem.
Your car carries 67 sensors. The office where you spend a third of your life carries zero.
The observation that started Safehub - CEO Andy Thompson01 / WHO THEY ARE NOWThe building, finally talking back
Safehub today is a San Francisco company of roughly thirty-two people that has quietly wired itself into some of the most recognizable buildings in the world. Amazon, Pfizer, Procter & Gamble, JLL, Hudson Pacific Properties, Keysight, Obayashi, the City of Seattle - the customer list reads like a who's who of organizations that cannot afford to guess after a disaster. What they buy is not a gadget. It is an answer delivered while the answer still matters.
The product is sensors plus software plus, increasingly, insurance. The sensors measure how the ground shook and how the structure responded. The software translates raw motion into a plain-language verdict - safe, inspect, evacuate - and pushes it to a dashboard, a text message, an email. The whole point is speed. After a major quake, the bottleneck has never been engineering knowledge. It has been getting an engineer to the building.
Above the fold: a sensor the size of a lunchbox, doing the job a structural engineer used to do on foot, in heels, three days late.
02 / THE PROBLEM THEY SAWEarthquakes are not democratic
An earthquake does not damage a city evenly. Two towers on the same block, built a decade apart on slightly different soil, can come out of the same quake in wildly different shape. Yet for decades the entire apparatus of disaster response and insurance has treated whole regions as a single dot on a map. The model says the area shook at intensity X, so everyone in the area is assumed to have experienced X. It is tidy. It is also frequently wrong.
That gap between the model's guess and the building's reality has a name in the insurance trade: basis risk. It is the reason parametric earthquake policies could pay out when a building is fine, or stay silent when a building is wrecked. It is the reason a CEO sometimes keeps a perfectly sound factory closed for a week, and sometimes sends people back into one that should not have been reopened. The problem Safehub saw was not a lack of seismologists. It was a lack of resolution.
The model says the neighborhood shook. The building knows whether it cracked. Those are not the same sentence.
The basis-risk problem, in one breath03 / THE FOUNDERS' BETPeople who had run toward the rubble
Safehub was not founded by app developers who read about earthquakes. It was founded by two people who had spent careers in them. Andy Thompson, the CEO, spent sixteen years at the engineering firm Arup leading its catastrophe-risk and insurance work, and personally responded to something like fifty disasters - Chile, Haiti, Japan, India, the United States. His co-founder, Doug Frazier, had previously run EQECAT, at one point the largest catastrophe-risk modeling company on the planet.
So when Thompson noticed his car had 67 sensors and his earthquake-exposed office had none, it was not an idle thought. It was a man who had stood in too many partially collapsed buildings deciding that the wait-and-see era should end. The bet was simple and slightly audacious: that a cheap sensor and a cellular connection could do, in minutes, the triage that used to take days - and that the world's most data-hungry industries would pay for it.
The market, eventually, agreed. In April 2021 Safehub closed a $9 million Series A led by A/O PropTech, with Hannover Digital Investments and JLL Spark joining returning backers including Fusion Fund, Ubiquity Ventures and Promus Ventures. It is the kind of cap table - a proptech fund, a reinsurer, a real-estate giant - that tells you exactly which three industries thought this was real.
Founders, unposed: a structural engineer who lost count of his disaster deployments, and the man who once ran the world's biggest catastrophe model. Neither needed a whiteboard to explain the problem.
The Safehub timeline
// A short history of teaching buildings to speak up
04 / THE PRODUCTThree jobs, one sensor
The hardware is intentionally boring - which, for a thing bolted to a wall and ignored for years, is the highest compliment. It installs without rewiring the building, talks over cellular rather than begging for the building's WiFi, and sits quietly until the ground moves. Then it earns its keep.
Building damage monitoring
The flagship. Real-time, building-specific alerts after a seismic event, delivered through a dashboard, SMS and email. The verdict arrives in minutes, not the days a manual inspection demands.
Portfolio risk management
For owners with hundreds of buildings, the sensors scale into a single view of an entire portfolio - which assets are fine, which need eyes, which need to stay closed. Business continuity stops being a guessing game.
ShakeNet Parametric
The newest twist, built with Liberty Mutual Reinsurance. Instead of paying out on a distant model's estimate, the policy is triggered by high-resolution shaking maps - drawn from Safehub's sensor network and processed through the GEM Foundation's OpenQuake engine. The promise: less basis risk, faster payouts, and coverage that can reach the developing markets traditional models overlook.
Never before have businesses had this level of information at their fingertips, allowing them to quickly assess damage.
Andy Thompson, Co-Founder & CEO05 / THE PROOFCustomers, labs and a reinsurer walk into a building
Skepticism is the correct posture toward any company that sells certainty. So look at who is buying. Enterprises that cannot afford downtime: Amazon, Pfizer, Procter & Gamble, Keysight. Real estate that lives and dies by building risk: JLL, LaSalle, Hudson Pacific Properties. A city that knows its fault lines intimately: Seattle. And an insurance establishment - Liberty Mutual Re, Hannover Re, Chubb - that does not place bets on technology it has not pressure-tested.
The sensors themselves were not graded on a curve. They were strapped to structures on shake tables at UC Berkeley and UC San Diego and made to prove their readings against the real thing. The 2024 Liberty Mutual partnership, and the 2025 launch of ShakeNet Parametric, are the clearest signal yet: a major reinsurer is willing to pay claims on what these sensors report.
Receipts: two universities shook the sensors on purpose, and a reinsurer agreed to pay out on the answers. That is a harder room to win than any pitch competition.
The speed argument
// Estimated time to a building-level damage verdict, by method
Bars are illustrative and show relative order-of-magnitude, not measured averages. The point is not the exact numbers - it is that everything else is measured in units a building owner cannot afford.
Backed by the people who underwrite risk
06 / THE MISSIONMake resilience boringly simple
The stated mission is unglamorous on purpose: make earthquake-damage monitoring simple, affordable and scalable. Read it again and notice what it is really arguing - that safety after a disaster should not be a luxury reserved for the best-instrumented towers in the richest cities. A sensor cheap enough to install across an entire portfolio, or across a developing market, changes who gets to know whether their building is safe.
There is an ambition hiding under the modesty. Safehub would like building-level seismic data to become as ordinary as a smoke detector - something you install, forget, and are quietly grateful for on the one day it speaks.
We can provide the highest-resolution regional shaking maps anywhere in the world, including the markets the old models never bothered to map.
Andy Thompson, on ShakeNet Parametric07 / WHY IT MATTERS TOMORROWThe cost of guessing keeps rising
Cities are denser, buildings are taller, and the financial machinery wrapped around them is more interconnected than ever. A single bad assumption after a major quake - a tower reopened too soon, a sound factory shut for a week, a claim paid on the wrong building - now ripples through supply chains, balance sheets and insurance markets that span continents. The price of guessing has never been higher, and it is still climbing.
Safehub's bet is that the future does not belong to better guesses. It belongs to measurement. If sensor-triggered insurance works at scale, the same approach that drained basis risk from earthquakes could spread to other perils that have always been priced on models rather than facts. The buildings, it turns out, were willing to tell us the truth all along. Somebody just had to give them a voice.
Back on the fourteenth floor. The aftershock has stopped, the lights are steadying, and the evacuation crew's phone buzzes. This time there is no shrug. There is a verdict - this building is sound, that one needs eyes - delivered while it still changes what happens next. The earth still moves whenever it pleases. Safehub simply made sure the building gets the first word.
Find Safehub
// Watch & read: product demos and the founding story are featured on safehub.io, Ubiquity Ventures' Series A profile (read the founder Q&A), and coverage on Stacey on IoT and Artemis.