Most people associate the word “AI” with OpenAI, Amazon, or Meta. Companies built on billion-dollar valuations, venture capital, and the Silicon Valley mythos. And with every right. That’s where the term earned its reputation. But the AI world isn’t reserved for giants with endless budgets.
OpenAI
Yes, the very same OpenAI. Why am I including them? Because nine years ago, they were a startup company, a quiet little lab trying to figure out AGI. It was on November 30, 2022, when they launched their first model called ChatGPT. It wasn’t even meant to be what it is today. The team had been watching developers mess around with their API playground and figured, hey, maybe we should make a demo of this. That demo kicked off a growth curve rarely seen in the tech industry. And it’s not that we haven’t seen amazing growth curves in that one.
Tempus AI
Anything that has to do with AI and medicine fascinates me. I truly believe this is one field where AI can have a significant impact on humanity. Tempus AI started back in 2015 when Eric Lefkofsky decided his next move would be healthcare data. The pitch was pretty straightforward. Doctors make treatment decisions based on a fraction of the data they could theoretically access. What if you could actually aggregate all that genomic sequencing, clinical records, and treatment outcome stuff into something useful? Turns out that’s a compelling idea when you’re talking about cancer, where personalized treatment could literally save lives. Fast forward to 2024, and Tempus goes public at a $6.1 billion valuation. Amazing.
Agility Robotics
You may have heard about Digit, the bipedal robot that walks like a real person instead of shuffling around. That one was made by Agility Robotics. The company spun out of Oregon State University back in 2015, which is a pretty unglamorous origin story for a robotics company that’s now deploying machines in Amazon warehouses. See, the thing about humanoid robots is that they’re always “five years away” for about thirty years now. Boston Dynamics makes those viral videos of robots doing parkour, but you don’t exactly see them stacking boxes at your local fulfillment center. Agility took a different approach by building a robot that’s useful now, not one that does backflips for YouTube. By 2024, Agility opened what they’re calling the first factory for humanoid robots at scale, which promises to build thousands of units annually.
Tinfoil
What if you could use powerful AI models without actually letting anyone see your data? That’s what the founders of Tinfoil pitched back in 2024. The founding team comprises a mix of MIT PhDs and former Cloudflare engineers, individuals who’ve built security protocols used by billions of people and worked on NVIDIA’s confidential computing initiatives. They came out of Y Combinator with what they’re calling a “full-stack platform” that runs AI workloads inside secure hardware enclaves. The use case is pretty obvious. Companies want to use ChatGPT for sensitive stuff, but they’re reasonably worried about data leaking. Right now, your options are to sign a data processing agreement and hope for the best, redact all the sensitive bits, or run everything on-premise, which is expensive and annoying. Tinfoil’s pitch is that you get “the privacy of on-prem deployments, while running on the cloud.
Replika
Now this one starts with a tragedy. Eugenia Kuyda built the first version after her best friend Roman Mazurenko died, creating a chatbot trained on his text messages so she could keep talking to him. That deeply personal origin story turned into an AI companion app that now has over 10 million users worldwide. The app got huge during the COVID lockdowns when lonely people downloaded it by the thousands. Users reported feeling genuine social support from Replika, with some saying they used it similarly to therapy. The company designed it around Carl Rogers’ therapeutic approach. Lots of positive feedback and emotional support. Since then, there’s been a huge influx of similar apps and services. Some lean more into a therapeutic spectrum of things, while others provide more of an emotional support. And yes, there are even AI girlfriend chatting platforms out there. So, from the grief project to 10 million people having relationships with AI chatbots? That’s either the future of human connection or a very troubling sign about loneliness in the modern world. Probably both.
Synthesia
Synthesia was launched in 2017 by a team of AI researchers from UCL, Stanford, and Cambridge with a pitch that sounds dystopian until you think about it for a minute: what if you could make videos without cameras, actors, or film crews?
The tech works by mimicking speech and facial movements based on recordings of real people with their consent. You get 230+ avatars that can speak in 140+ languages, and companies use this stuff for the most boring corporate purposes imaginable. Training videos, internal communications, and onboarding materials. Basically, all the content that nobody wants to film but everyone has to make anyway. The weird part is that Synthesia isn’t trying to disrupt Hollywood or replace YouTubers. Instead, they’re going after corporate training videos and HR onboarding. An interesting niche, or should I say, a rather boring one. The company hit unicorn status in June 2023 at a $1 billion valuation. In April 2025, Adobe tried to acquire them for around $3 billion, and Synthesia turned it down over pricing disagreements.
Scale AI
Alexandr Wang dropped out of MIT in 2016 because he noticed something obvious in retrospect: everyone wanted to build AI, but nobody had good labeled data to train their models on. Someone had to tell the computer which pictures contain cats and which contain dogs, and that work was being done by nobody. Wang co-founded Scale AI with Lucy Guo, and the pitch was simple: we’ll handle all the tedious data labeling work so you can focus on building your AI models. They started by building a lidar labeling tool for autonomous vehicles, which required extremely precise standards. The company grew by outsourcing the actual labeling work to contractors. In 2022, Scale won a $250 million contract with federal agencies, and they’ve been working on military projects, including something called Donovan, an LLM they deployed on classified networks.
From MIT dropout to billionaire by convincing people to label images for self-driving cars? The whole AI boom runs on data annotation, and Scale figured that out before anyone else.
Databricks
Databricks came out of UC Berkeley in 2013. Here’s the problem they were solving: companies had data warehouses for structured stuff (think spreadsheets and databases) and data lakes for everything else (logs, images, random files).
Databricks came up with the “data lakehouse” concept. A portmanteau that sounds like marketing speak but actually describes something useful. It combines the cheap storage of data lakes with the management features of data warehouses, so you can run both analytics and machine learning on the same data without copying it around. The company built Delta Lake, an open-source layer on top of cheap storage that adds transactions and reliability.
The valuation trajectory has been pretty wild. In February 2021, Databricks raised $1 billion at a $28 billion valuation. By August 2021, they’d raised another $1.6 billion, at $38 billion. In December 2024, they announced $10 billion in funding at a $62 billion valuation. Then by August 2025, they’d raised another $1 billion in a Series K round, pushing their valuation over $100 billion. From an academic project to $100 billion company in twelve years by convincing enterprises that maybe they don’t need seventeen different systems to store their data? That’s one way to do it.
Source: ameyawdebrah.com/
