Fastino Raises $17.5M to Make AI Training Affordable with Gaming GPUs
What if training powerful AI models didn’t require multi-million dollar data centers and the most expensive chips on the market? That’s exactly the question Fastino, a startup based in San Francisco, is answering. And now, they’ve got $17.5 million in new funding—led by Khosla Ventures—to prove they’re onto something big.
Let’s break down how Fastino plans to transform the AI industry, why using gaming GPUs is such a game-changer, and what this could mean for startups, developers, and the future of artificial intelligence.
Wait… AI models trained with gaming GPUs?
Yes, you read that right.
At the heart of Fastino’s approach is the idea that you don’t need super expensive chips like NVIDIA’s H100 GPUs—which can cost tens of thousands of dollars—to train machine learning models. Instead, they’re using consumer-grade chips, the kind of graphics cards you’d usually find in a gaming PC.
Why does this matter? Because training large AI models today typically requires enormous amounts of computing power and cost. That’s something only tech giants like Google, OpenAI, and Meta can generally afford. Fastino, however, is flipping the script by showing that even more affordable hardware can be put to work—effectively—at scale.
What makes Fastino’s tech special?
Using gaming GPUs isn’t a new idea on its own. But what makes Fastino stand out is how efficiently they’re using them. They’ve built their own software stack and training platform that squeezes every ounce of performance out of these low-cost GPUs.
Their secret sauce lies in a clever mix of efficient algorithms, system-level optimization, and creative memory handling. In simple terms, Fastino’s tech figures out how to keep the training going fast without blowing past the hardware’s limitations.
Here’s what Fastino brings to the table:
- Lower Costs: Training models on gaming GPUs can be up to 10 times cheaper compared to enterprise hardware.
- Increased Access: Smaller startups and researchers can now build and deploy AI with fewer financial barriers.
- Eco-Friendly Approach: Less power and simpler setups mean a reduced carbon footprint for AI training.
Who’s backing Fastino?
This $17.5 million seed round was led by Khosla Ventures, a major name in tech venture capital. They’re known for backing some of the most disruptive companies in AI and biotech. Other investors include several angels and operators who’ve seen firsthand the challenges of AI infrastructure costs.
Why are big investors betting on Fastino? Because reducing the cost of AI training is not just a nice-to-have—it’s one of the biggest bottlenecks holding back innovation in the field.
Fastino’s founders know their stuff
Fastino was founded by Glen Takahashi (CEO) and Ali Kheyrollahi (CTO), both computer engineers with a deep background in systems architecture and AI training frameworks.
Before coming together to build Fastino, they had firsthand experience navigating the headaches of current AI infrastructure: high costs, long wait times for chip access, and rigid platforms that lock you into expensive ecosystems.
That frustration turned into an idea—and now it’s a fast-growing company with the funding to scale.
How does this change the AI landscape?
If you’ve ever tried launching an AI project, you know how quickly cloud compute costs add up. On top of that, snagging access to high-power GPUs can come with long waitlists. It’s a little like trying to build a skyscraper but only being allowed to use one crane—shared with hundreds of others.
Fastino is essentially giving every team their own set of tools—affordable, efficient, and ready to go.
Imagine a world where:
- Startups can train their LLM (large language models) without raising millions first
- Independent researchers explore new AI models from their home offices
- Universities run experiments on clusters of gaming GPUs instead of waiting in line for supercomputers
This is the future Fastino is betting on. And with this new funding, they’re planning to grow their team, improve their software, and expand access to more early customers.
Where does this go next?
According to CEO Glen Takahashi, Fastino’s goal isn’t to just build tools—they plan to offer full model training runs as a service. That means developers could log into their platform, upload a model or dataset, and let Fastino handle the rest.
This opens the door for “AI model training as a service,” similar to how cloud storage or website hosting became mainstream. You won’t need to be a hardware expert or budget for expensive equipment. Just plug in and go.
Could Fastino disrupt Nvidia?
Let’s be clear—Nvidia still powers the majority of serious AI training workloads, and their hardware is not going away anytime soon. But Fastino is carving out a role for itself in an underserved and fast-growing market: startups, small teams, and global developers trying to build cutting-edge AI without breaking the bank.
Think of it like this: Nvidia is selling luxury cars. Fastino is making bicycles that can keep up in the race—at a fraction of the cost. And in today’s tough financial climate, who wouldn’t want a cheaper, faster way to get from point A to point B?
Final thoughts: Why should you care?
AI is changing the world. But for it to reach its full potential, it has to be accessible to more people than just the Big Tech elites.
By making AI training faster, cheaper, and more accessible, Fastino is democratizing innovation. Whether you’re a startup founder, a curious student, or a seasoned machine learning engineer, tools like this could be the key to launching your next big idea.
So, if you’re working on building something in AI—or just considering it—keep an eye on Fastino. The way we train machine learning models might be about to change forever.
SEO Keywords (used throughout):
- AI training platform
- gaming GPUs for machine learning
- affordable AI model training
- AI startup funding
- machine learning hardware alternatives
Have ideas about where AI is heading? Drop a comment and let’s chat—because the future of artificial intelligence might just start with a gaming GPU.