Sift, a Californian start-up founded by former SpaceX engineers, can now pull live sensor data straight out of machines that run QNX, the real-time operating system from BlackBerry that sits inside more than 275 million vehicles. Under an integration the two companies announced on 17 September, Sift subscribes to the telemetry that a QNX system already broadcasts, and engineers can query each reading with standard SQL within one second of it leaving the device.
The integration works with QNX OS 8.0 and picks up data that devices already publish over MQTT, the lightweight messaging protocol that much of the IoT world uses to move sensor readings, with custom ingestion paths for other formats. Teams do not need to change their firmware or rebuild their software to use it, which matters for systems where every change to the code on the device can trigger fresh testing and certification.
Sift aims the service at the industries where QNX already runs, from cars, trucks and rail to robots, medical devices, defence systems and industrial automation. "The next decade of hardware will be won by the teams that learn fastest from their machines," said Austin Spiegel, Sift's co-founder and chief executive.
What One Second Of Latency Buys An Engineer
That speed changes how engineers debug machines, because the usual route for data from an embedded system runs through log files that someone copies off the device and analyses hours or days later. With a live feed, an engineer can put thousands of measurements from dozens of subsystems on one timeline while a machine runs, and compare a fault as it happens with the processor load or temperature that came before it.
Romain Saha, who leads strategic alliances at QNX, said the partnership gives customers "a validated, low-friction path from a QNX-powered device to real-time analysis." The same data also lets teams compare a new machine against records from earlier builds or deployments, which is how Sift's founders used similar tools at SpaceX to test and launch rockets.
Why SQL Matters Here
Standard SQL may sound like a minor detail, yet it lowers the bar for who can use the data. Engineers and analysts who already query business databases can work with machine telemetry without learning a specialist tool, and the same queries can feed dashboards, alerts and, increasingly, AI models that look for patterns in how machines behave.
Sift's Climb From Rockets To Factories
Sift started in 2022 in Los Angeles, and its founders, Spiegel and Karthik Gollapudi, built telemetry software at SpaceX before leaving to sell the idea to others. The company raised $42 million in a Series B round led by StepStone in March 2026, with GV among its backers, taking total funding to about $67 million, and TechCrunch reported that some of its customers' vehicles stream data from more than 1.5 million sensors at once.
Its customer list still leans towards space and defence, including United Launch Alliance and satellite maker Astranis, but the company has been pushing into factories, rail and robotics. Spiegel took over as chief executive in May 2026 from Gollapudi, who moved to focus on product direction and industry outreach.
QNX Wants To Be More Than A Car Company
For QNX, the deal fits a push beyond the car dashboard into what the industry now calls physical AI, the machines that sense and act in the real world, a shift we traced in how physical AI turns sensing into action. In April, QNX and NVIDIA paired QNX OS for Safety 8.0 with NVIDIA's IGX Thor computer for robots, medical devices and industrial machines, the same class of hardware whose edge AI benchmarks we examined last week. QNX president John Wall said at the time that "safety and determinism cannot be afterthoughts as systems become more autonomous and software defined."
Cars still pay most of the bills, and they are paying more. BlackBerry reported QNX revenue of $80.3 million for the quarter to the end of August, up 27% on a year earlier, and on 22 September Coretura, the software joint venture of Daimler Truck and Volvo Group, chose the Alloy Kore platform that QNX built with Vector, a deal BlackBerry said adds more than $100 million to QNX's royalty backlog.
Partnerships like the one with Sift make QNX more useful to engineers who have spent years on Linux-based tools, since a real-time operating system wins design slots partly on how easily teams can see what it is doing.
A Crowded Field For Machine Data
Sift will not have that job to itself, because investors have poured money into tools that help teams make sense of machine data. Foxglove, which builds an observability platform for robotics, raised a $40 million Series B in November 2025 and counts NVIDIA, Amazon, Anduril and Wayve among its users, and it created MCAP, an open-source format for logging robot data.
The difference lies partly in where each tool starts: Foxglove grew up around robotics developers and their logs, while Sift grew up around test stands and launch pads where engineers watch hardware in real time. A validated integration with QNX gives Sift a way into the safety-critical systems where QNX is strongest, a group of customers that tends to buy slowly but stay for years.
What Buyers Should Check
The one-second figure comes from the companies, and actual latency will depend on the customer's network, the MQTT broker in between and how much data a machine sends. Buyers will also want to know what it costs to stream and store high-rate telemetry from a fleet rather than a test bench, since neither company has published pricing, and how the data stays secure on its way from the device to Sift's cloud.
The integration sits outside the safety-critical code, reading data that the system already broadcasts, which should keep it clear of recertification but also means teams need a device that publishes the right signals in the first place.
The Machines Are Starting To Report In Real Time
The Sift and QNX tie-up is small on its own, but it points to how companies will build physical AI: machines that stream what they are doing while they do it, and engineering teams that treat that stream like any other database. As robots, trucks and medical devices take on more autonomous tasks, the companies that can see and explain their machines' behaviour fastest will fix faults, pass audits and train better models before their rivals do.
For QNX, every tool that makes its operating system easier to observe helps it compete for design wins beyond cars, and for Sift, a validated path into QNX devices opens a large base of safety-critical machines. The next proof point will be named customers running the integration on fleets rather than test benches, which neither company has disclosed yet.
