Trajectory just raised $45 million to solve the one problem every mid-size company has with AI. Not the cost of compute. Not data privacy headaches. The bottleneck is simpler and more expensive than both: there aren’t enough machine learning engineers to go around, and most businesses can’t get near building their own models because of it.
Why This Is Happening Now
AI model training has always been a specialist’s game. To build a custom model, you needed a data pipeline, a training loop, hyperparameter tuning, and a team of people who understood all three. That’s not a product manager. That’s not an operations director. That’s a small, expensive team most companies can’t afford or find.
According to a 2025 Gartner report, 68% of companies that want to build custom AI models cite talent shortages as their primary barrier, ahead of cost and data quality. The demand for ML engineers has outpaced supply for four straight years. According to LinkedIn’s 2025 Jobs on the Rise report, machine learning engineer roles saw 42% year over year demand growth, with median salaries above $180,000.
Trajectory’s platform takes the code out of the equation. Their tool lets product teams, business analysts, and domain experts train and fine tune models through a visual interface. You bring the data. They handle the engineering.
The Real Money Play Here
Most people will see this as a tech story. It’s a money story.
Here’s the split that matters. Companies that rent AI call OpenAI or Anthropic’s API for every task. They pay per token, forever, for a general model that’s identical to what their competitors use. There’s no edge. There’s no ownership. There’s just a recurring bill.
Companies that own AI have something different. They have a model trained on their own proprietary data. Their customer purchase history. Their product defect logs. Their support tickets going back a decade. A model trained on that data knows things no general purpose model ever will.
According to a 2025 McKinsey Global Survey on AI Adoption, companies with internally built AI models reported 2.4 times higher return on their AI investments compared to companies that relied entirely on third party models. The reason isn’t complexity. It’s specificity. Custom models solve the exact problem you have.
The AutoML market, tools that automate the technical parts of model training, is growing fast because of exactly this pressure. According to Grand View Research, the global AutoML market is projected to reach $14.8 billion by 2030, growing at a compound annual rate above 40%. Trajectory is positioning itself at the center of that growth by going one step further: removing technical requirements entirely, not just simplifying them.
The competitive field is real. H2O.ai, DataRobot, and Hugging Face’s AutoTrain all compete in adjacent spaces. But most of those tools still assume users understand what they’re doing at some technical level. Trajectory is betting on the next layer down: the operator who understands the business problem deeply but has never written a training loop.
What This Means for Your Business
If you run a company and you’re still paying per API call for every AI task, you’re building on rented land. Here’s what I would do with the information Trajectory’s raise just put in the room.
Start with a data audit. The only thing that makes a custom model worth building is proprietary data nobody else has. Customer behavior logs, transaction histories, product performance records, internal support tickets. If you have any of those sitting in your database, you have the raw material for a model that could beat any general purpose tool on your specific problem.
Pick one narrow problem to solve first. Not “make our marketing better.” Something specific: churn prediction, document classification, fraud flagging, lead scoring. Small scope, clear outcome, measurable result. That’s how you build internal confidence before going bigger.
Think about the capital side too. Building AI capabilities in house has upfront infrastructure and tooling costs. If you’re evaluating how to fund a technology buildout without wrecking your cash flow, SuperMoney’s loan comparison tool can help you find competitive financing rates across multiple lenders at once. The goal is ownership without debt that kills your margin.
The window to build a proprietary AI advantage is open right now. It won’t stay open at this price point.
The Bottom Line
Trajectory’s $45 million isn’t just a bet on one company. It’s a signal about who gets to own AI. For the past three years, custom AI model ownership has been locked behind a wall of engineering talent most businesses can’t access. That wall is coming down. The companies that move now will own models their competitors can’t replicate. The ones that wait will still be paying rent when the market shifts again.
Frequently Asked Questions
What does Trajectory do with the $45M it raised?
Trajectory is using the capital to expand its no code AI model training platform, grow its go to market team, and invest in infrastructure that reduces training time for business users. The core product lets nontechnical teams build and deploy custom AI models without writing code.
Why does AI model training matter for small and mid size businesses?
Custom AI models trained on proprietary business data outperform general models on specific tasks. According to McKinsey, companies with internally built AI models see more than double the ROI compared to businesses that depend entirely on third party APIs. The gap compounds over time as the custom model improves.
How expensive is it to train a custom AI model in 2026?
Costs have dropped sharply. According to research from Andreessen Horowitz, the cost to train AI models at a given performance level has fallen roughly 10 times every 18 months since 2020. Platforms like Trajectory push costs down further by removing the engineering labor component, which is often the biggest single expense.
What data do you need to start training your own AI model?
You need clean, labeled data specific to your use case. Customer records, transaction logs, support tickets, product data, and operational reports all work as training sources. The more specific and proprietary your data, the stronger the advantage a custom model gives you over competitors using the same general purpose tools.
Who are Trajectory’s main competitors in the AI model training space?
The closest competitors include H2O.ai, DataRobot, and Hugging Face’s AutoTrain. Most of those platforms still require some technical knowledge to use effectively. Trajectory is targeting business users with no ML background, which is a different and arguably harder product to build at scale.
