At a glance
- Firmus has committed approximately US$300 million for up to 150Tbps on APX East over 25 years, as the cornerstone customer and the first buyer publicly named on the system.
- On current per-pair capacities, 150Tbps fits on eight to ten of the cable's 16 fibre pairs, which puts Firmus between roughly 50 and 62 per cent of the system.
- Firmus is paying about US$30 million to US$37 million per fibre pair.
- Using 9 fibre pairs as the mid-point, a US$281 million build figure has been placed beside Firmus's US$300 million to read the deal as cost plus 7 per cent. The two numbers measure different things, and Bevan Slattery told the Australian Financial Review in January that the plan was US$500 million.
- The anchor tenancy puts APX East on course for contract in force, the milestone at which SUBCO's supply contract takes effect and marine survey and manufacturing begin.
- APX East is Firmus's fourth long-dated supply commitment since June, after 600MW of South Australian power, the Koolunga firming offtake and the Bernacchi-1 cable.
- Every share, price and scenario below is a Certified Strategic projection with its inputs stated, and none has been confirmed by Firmus or SUBCO.
Firmus is the cornerstone customer on a 16 fibre pair system
Firmus, the Sydney-headquartered company building AI factories across Australia and Asia, committed in its 3 September announcement approximately US$300 million for up to 150Tbps of dedicated capacity over 25 years on APX East, SUBCO's 16 fibre pair, roughly 13,000 kilometre express cable from Sydney to California, due in service in the fourth quarter of 2028. Certified Strategic covered the agreement on 3 September. Firmus is the cornerstone customer, and the first buyer either company has named on the system.
Bevan Slattery sized the requirement at 75 to 150Tbps for a 3GW Australian AI factory pipeline in January, and one neocloud has taken the top of that range as its own ceiling. How much of the cable that represents decides whether Firmus bought a large tenancy or most of the building, and SUBCO has published no total design capacity to settle it.
The US$281 million comparison prices a different thing
A build figure of US$281 million has been put beside Firmus's US$300 million, and the pair reads as cost plus a 7 per cent margin. This is derived linearly from the announced total project cost of US$500 million, compared with the nine fibre pairs Firmus is committing to.
The published cost is a different number. Slattery told the Australian Financial Review on 19 January that APX East was a US$500 million plan, or A$747 million, and Submarine Networks records the same figure. The A$700 million in that AFR headline is a capital raise Street Talk reported SUBCO was weighing in September 2025, with Barrenjoey the likely adviser, and Light Reading's "reportedly AU$700 million (US$471 million)" converts that raise figure as though it were the build. Slattery told the AFR he would fund the cable with project finance and cash flow from SUBCO's existing systems.
Eight to ten pairs carry 150Tbps
Spread across all 16 pairs, 150Tbps works out at 9.4Tbps each, which is below what a system entering service in the late 2020s carries per pair. Firmus therefore holds a subset of the cable. Certified Strategic's working range for a 13,000 kilometre repeatered pair on current terminal equipment is 15 to 25Tbps, and a reader who takes a different view of the transmission budget can substitute their own figure and rerun the table.
| Capacity per fibre pair | Pairs carrying 150Tbps | Firmus's share of 16 pairs |
|---|---|---|
| 15Tbps | 10 | 62 per cent |
| 20Tbps | 8 | 50 per cent |
Source: Certified Strategic Editorial projection from SUBCO's published pair count, September 2026. Per-pair capacities are CS inputs, not SUBCO figures.
Firmus's ceiling comes out between half and two thirds of APX East, and the system's design capacity between 240 and 320Tbps. A lower per-pair figure lifts Firmus's share and lowers the system total. SUBCO can settle both by publishing the design capacity.
Gunvor, Koolunga and Bernacchi-1 came first
Firmus builds AI factories and contracts long for the inputs it does not build. Its South Australian power runs under a 12-year, 600MW supply agreement with Gunvor, more than half of whose initial firming comes from a long-term offtake on the 200MW Koolunga battery. In June it underwrote Bernacchi-1, the SUBCO branch carrying more than 60Tbps between Tasmania and the mainland from the second quarter of 2027. It leases space in other operators' Australian data centres for its Melbourne compute while building its own Tasmanian campuses.
APX East is the longest of those commitments. Capital Brief reported on 3 September that Firmus is expected to list on the ASX before the end of the year, which puts a 25-year US$300 million obligation on the balance sheet of a company heading for a prospectus. Whether the money bought equity in the system or a prepaid right to capacity changes how it appears there, and the release settles neither.
The remaining pairs are open to the rest of the market
Firmus bought a block, not exclusivity. SUBCO sells capacity wholesale, so the pairs Firmus did not take are available to NEXTDC, CDC, AirTrunk, DCI or their tenants on the same cable, at whatever the price is once the system is funded rather than proposed. What Firmus bought was first position and the cornerstone rate.
The export path reaches every operator. Australia has 26 million people, so an operator pitching a hyperscaler on domestic demand alone is selling into a small market. A direct Australian-owned route to California lets any operator here sell into a global one, and Data Centres Australia and Mandala have valued AI compute exports at up to A$4.1 billion a year by 2030. SUBCO's own framing for the system is reduced reliance on US hyperscalers, which is a statement about who allocates capacity rather than about the engineering.
Supply arriving in 2028 also lands on a route that already has incumbents. Southern Cross NEXT, Hawaiki and Telstra's Endeavour will face a system with unsold pairs, which reaches operators and carriers at their next renewal whether or not they ever buy on APX East. The Sydney landing adds a second effect: SUBCO has said only that it comes ashore north of the existing protection zone, and wherever that is, campuses with short terrestrial fibre to it gain an interconnection advantage.
Thin supply is the risk on the other side. Firmus at two thirds of the cable leaves few pairs on the market, and a rival operator's tenant could end up buying trans-Pacific capacity from a competitor. That is the sharpest argument for another Australian operator taking a position on the remaining pairs before the second price is set.
What is Firmus buying it for?
Firmus needs a fraction of 150Tbps to move its own inference traffic in 2028. Half to two thirds of a trans-Pacific cable is a wholesale position, and Certified Strategic reads it as capacity bought for Firmus's tenants ahead of them, to be sold or bundled with compute rather than consumed.
The customers Firmus is built for make that work. A hyperscaler or AI lab leasing an Australian AI factory to serve users offshore needs the path to the United States as much as the megawatts, and Firmus now controls the largest single block on the first new route. Slattery said in January that international connectivity, ahead of power, land, data centres and chips, would be Australia's longest lead item, and Firmus has bought the lead item and can resell it. The equity question reprices all of this: a co-owner earns on every pair SUBCO sells, and a capacity holder earns only on what it moves.
Why did both companies sign?
A cable costs everything up front and earns nothing for three years, and the project finance Slattery described rests on contracted revenue a lender can see. Firmus supplied it, and SUBCO can now build. Slattery also said in January that a system arriving in 2029 or 2030 would be too late for the AI buildout, so waiting for a hyperscaler to decide was the more expensive option. Selling a large block early to one buyer starts the build.
Firmus is buying something it cannot manufacture later. Capacity for 2028 is allocated in 2026, and a tenant renting an Australian AI factory to serve users overseas needs the path to the United States as much as the power. Firmus bought more than its own traffic requires and can pass the rest to its customers.
Firmus has now fixed its power for 12 years and its international capacity for 25, which gives it a known cost base while competitors queue for grid connections. SUBCO has built sixteen pairs for a route with one customer, no intermediate landings for a foreign jurisdiction to sit in, and single-end power so the system runs through a fault. The pairs Firmus did not take are the upside, and they sell higher once the cable exists.
What to watch
SUBCO's financial close fixes the build cost. A close near US$500 million confirms Slattery's January figure and makes the US$281 million a component of it. The A$700 million raise Street Talk reported in September 2025 is where that surfaces first.
APX East's design capacity fixes Firmus's share. Until SUBCO publishes it, every percentage in this piece rests on a per-pair input rather than a disclosure.
The second named customer sets the second price. A carrier or hyperscaler paying above Firmus's rate confirms the cornerstone discount, and one paying below it says the market values trans-Pacific capacity under what Firmus committed to. A Firmus prospectus would answer the equity question in the same window.