Why it matters
The bottleneck for most AI builders is not model quality but affordable, available GPU capacity, and a wave of neoclouds is forming specifically to arbitrage that gap. Firmus is the latest, and the structure of its Nvidia agreement, where Nvidia supplies the silicon, takes product revenue, shares in the cloud revenue, and is already an equity investor, shows how Nvidia is extending its reach down the stack into the cloud layer rather than only selling chips.
The deal
Firmus Technologies, an Australian AI-infrastructure firm, will buy Nvidia infrastructure and resell Nvidia-powered cloud services to AI Native customers and others, according to Reuters. The partnership covers 170,000 GPUs deployed from the first quarter of 2027 through the start of 2028, located in Batam, Indonesia. Firmus said it expects up to $30 billion in revenue over the first six years of the deal, based on customer commitments. Co-chief executive Tim Rosenfield told Reuters the aim is to close the compute gap for smaller firms relative to large companies. Nvidia participated in Firmus prior capital raisings, making it an existing investor; Firmus said in April it had raised $1.35 billion over the prior six months for a $5.5 billion post-money valuation and has engaged banks on a potential IPO.
Practitioner read
Siting 170,000 GPUs in Batam, just across from Singapore, targets Southeast Asian demand and cheaper power and land while staying close to a major connectivity hub, a pattern repeating across the region. The revenue-share model is the detail to watch: it gives Nvidia recurring upside beyond hardware sales and ties a neocloud's economics tightly to its supplier, which can constrain pricing flexibility even as the pitch is lower-cost access. The $30 billion figure is a projection resting on customer commitments and 2027-2028 delivery, so execution and utilization risk are real. For buyers, more dedicated GPU supply outside the hyperscalers is a near-term positive for availability and price.
Watch next
- •Whether the 2027 delivery schedule and customer commitments hold as Firmus pursues an IPO.
- •How Nvidia's revenue-share neocloud model spreads to other regional operators.
Key Points
- 1Firmus and Nvidia signed a deal to deploy 170,000 GPUs in Batam, Indonesia, from 2027 to 2028.
- 2Nvidia earns product revenue plus a cloud-revenue share and is already a Firmus equity investor.
- 3Firmus projects up to $30 billion in revenue over six years, adding non-hyperscaler GPU supply for smaller AI firms.
Scoring Rationale
A 170,000-GPU neocloud buildout with a major projected revenue base materially expands non-hyperscaler compute supply. The Nvidia revenue-share-plus-equity structure is a notable model showing Nvidia extending reach down the cloud stack. Impact is tempered by 2027-2028 delivery timeline and execution risk on the $30B projection.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
