9 insights from ‘Private Tech Trailblazers’: Vertical AI becomes the growth engine
SiliconANGLE Cheryl Knight
Vertical AI is where the money’s going now. From robots to banking, the winners are the ones built on their own data and stack.
Based on reporting by SiliconANGLE, Cheryl Knight — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
At Bank of America’s Private Tech Trailblazers Conference in Palo Alto, the clearest theme wasn’t “AI everywhere.” It was narrower than that: the companies drawing attention are the ones using AI on specific jobs, with proprietary data and, more often than not, their own hardware and software underneath. The buzz has moved from generic models to systems built for restaurants, construction sites, hospitals, payments and defense.
Bear Robotics is a good example of how fast that shift is happening. Bren Pierce said the company has about 16,000 autonomous mobile robots in the field, roughly 4,000 more on backlog, and revenue that doubles every year. Its humanoids now sit on the same software base and cloud setup as its mobile robots, while the company leans on foundation models, onboard compute and large language models to turn work that once took six months into days. The missing piece, Pierce said, is still tactile hands at about $30,000 apiece.
Construction is getting the same treatment. All3, the operating name of Address Robotics, is trying to redesign the whole process rather than just automate a single task. It starts with software that takes a building from plot to permit-ready documents, then uses industrial robots to make one-off parts and a mobile robot called Mantis to assemble them on site. The company says labor accounts for 55% to 60% of construction costs, and its first project is a six-story co-living building on an 11-sided plot.
The same pattern shows up in software and services. Bloomreach said its Loomi AI engine, trained on 7 billion consumer profiles, beats out-of-the-box LLMs by five to 10 times, while Airwallex is turning cross-border banking into something customers can use as if it were local in 80 to 100 economies. Hippocratic AI is using 31 models to keep voice agents safe for healthcare, and CloudWalk says agents now handle 99% of customer support while running on its own cluster of hundreds of Nvidia Blackwell GPUs. The message across the conference was blunt: the model layer is getting easier to copy, so the real moat is the data, the workflow and the stack.
My take — AI-written commentary, not fact-checked reporting
This is what AI was always going to look like once the demo glow wore off: less chatbot circus, more plumbing. The companies with real leverage are the ones using AI to squeeze costs, control risk and own more of the stack, which is a lot less sexy and a lot harder to copy. Boring wins, and Silicon Valley still seems mildly offended by that.
Read more about this at: SiliconANGLE