The artificial intelligence boom has long been dominated by a handful of familiar names. Nvidia makes the chips. Microsoft, Amazon and Google provide the cloud infrastructure. Investors have largely focused on whoever controls the most powerful models.
But as AI adoption accelerates, a new bottleneck is emerging: getting computing power into the hands of businesses that actually want to use it.
A series of recent moves by Megaport Ltd (ASX:MP1) reflects that shift in focus.
The Brisbane-based network connectivity company entered a trading halt on Tuesday before launching a fully underwritten $827.3 million entitlement offer to fund a major expansion into AI infrastructure, including a new global GPU pool and four newly signed contracts worth a combined $458.9 million in total contract value (TCV).
The scale of the raise is striking. So is the ambition.
Rather than competing with hyperscale cloud providers in training large language models, Megaport is targeting what many analysts increasingly view as the next major growth opportunity in AI: inference.
The next AI infrastructure battle
Training is the process of building AI models such as ChatGPT, Claude or Gemini. Inference is what happens afterwards — the moment users actually interact with those models and receive responses. Importantly, these distinct workloads have very different infrastructure requirements.
Training tends to be concentrated inside massive data centres packed with tightly clustered GPUs. Inference can be distributed across multiple locations closer to end users, reducing latency and potentially opening opportunities for a broader range of infrastructure providers.
Megaport believes the growing focus on AI inference plays directly into its strengths.
The company already operates a global network spanning more than 1,100 connected data centres across 31 countries. Through last year's Latitude.sh acquisition, it also gained a growing compute platform that management says is increasingly attracting demand for GPU capacity.
The newly announced capital raising is designed to accelerate the company’s strategy.
Around $350 million will be invested in an on-demand GPU pool, while a further $369.5 million will fund hardware required to service the newly secured AI infrastructure contracts.
Chief executive Michael Reid described AI inference as "one of the biggest infrastructure opportunities of the next decade".
Demand continues to surge
Megaport is not the only company chasing the opportunity.
Across the technology sector, demand for AI infrastructure continues to expand rapidly as enterprises move beyond experimentation and begin deploying AI applications into production environments.
Global data centre demand linked to AI inference is expected to grow at a compound annual rate of about 35% between 2025 and 2030. Megaport said prospective customers are increasingly requesting GPU inventory, while procurement lead times can stretch beyond 12 weeks.
The company has already signed eight compute-related contracts over the past two months with combined TCV of $747.8 million.
Wednesday's announcement adds another four contracts to that pipeline, all with US-based technology providers running AI inference workloads. The agreements are expected to contribute approximately $199 million in annual recurring revenue once fully deployed.
Investors taking notice
The market has already begun rewarding Megaport's AI pivot.
Before entering its current trading halt, the stock had climbed roughly 84% over the past month and more than doubled from levels seen earlier this year as investors responded to the increasingly large AI infrastructure contract wins.
The latest announcement helps explain that enthusiasm.
Megaport's traditional networking business remains healthy, with annual recurring revenue rising 25% year-on-year to $277.7 million. But the company's compute division is growing even faster. Including recently announced strategic contracts, pro forma compute ARR has reached $385.2 million, contributing to total group ARR of $662.9 million.
Those numbers suggest Megaport is evolving from a networking specialist into something broader — a provider of compute, storage and connectivity infrastructure aimed at supporting the next stage of AI adoption.
Whether the company can successfully deploy more than $700 million of new infrastructure capital remains the key question.
But if AI inference becomes as large a market as many expect, Megaport is positioning itself to become a far bigger participant in the AI ecosystem than investors might have imagined even a year ago.