We’ve secured $594M to build the global AI Inference Cloud

Megaport, our parent company, has announced a capital raise of A$827.3M, about $594M in US dollars.

That capital has one job: let us build a globally distributed AI inference cloud.

If you've been following us on social media lately, you already know we closed four major AI infrastructure contracts with US-based technology providers. Roughly A$369.5M of the raise goes into delivering them: latest-generation bare metal GPU and CPU instances, plus the networking and storage to run them.

But fulfilling those contracts is only half of it. The remaining funds will go toward something we've wanted to build for a long time: our own on-demand GPU pool for next-generation AI inference workloads.

We are using a significant portion of the capital raised to purchase 2,048 NVIDIA B300 GPUs and make them available on our platform for on-demand use by this autumn, and we couldn't be more excited about it.

Latitude.sh started from a simple frustration: infrastructure was too slow, too opaque, and too hostile to the people actually building on it.

Tech teams deserved better than waiting several weeks for hardware, decoding pricing pages, or getting punished with egress fees for moving their own data. So we built the opposite. Bare metal you provision in seconds. Instances with bandwidth allowances. Compute pricing that an engineer can reason about without a spreadsheet.

Every product we've shipped has been a version of the same question: what's one thing in compute infra that's harder than it needs to be, and how do we take it away?

That's the through-line. We never really sold servers. We've been selling back time, clarity, performance, and control to the people doing the hard work of building the next big thing in tech. And now we get to do exactly that for AI.

For years, the hard problem in AI lived in training: enormous models, enormous campuses, capacity concentrated in a handful of places. But the teams we serve aren't training frontier models. They're shipping AI into real products, and for them, the problem has quietly shifted to inference: running a trained model to answer several requests, fast, safely, and close to their end users.

Training rewards scale in a single location. Inference rewards being everywhere your users are, and that's a distribution problem. Which is exactly why Latitude.sh and Megaport make sense together.

Megaport's network already interconnects more than 1,100 data centers across 31 countries; add latest-gen GPU capacity to that footprint, close to where requests actually originate, and you get inference that runs next to your users instead of three regions away.

In other words, what makes a developer's AI solution feel fast and affordable is the same thing we've cared about from day one: removing the distance, friction, and fees that punish you for doing what you need to do to ship your next breakthrough.

We've spent years making bare metal fast, automated, transparent, and genuinely good to build on. Now we get to do that for the workload that will define the next decade of software, and for the engineers behind it.

Alongside Megaport, we're set to build the global AI inference cloud. And now we've secured the capital to make it happen.

To our customers: thank you for trusting us with your mission-critical applications. We're grateful to play even a small part in bringing so many incredible ideas to life. The future is bright!