At a glance

  • Sharon AI raised its expected near-term deployment to 100MW by early 2027, up from 70MW a quarter ago.

  • Total contracted value across Canva, GMI Cloud, ESDS and an unnamed Asia-Pacific customer is now above US$2.2 billion.

  • Sharon AI’s existing 50MW lease at NEXTDC M3 Melbourne is around 22 per cent of M3’s 225MW Tier IV envelope. That is a large single-tenant share.

  • The additional 30 to 50MW plausibly lands across some mix of further M3 expansion, NEXTDC’s S3 Sydney, the planned M4 Melbourne campus and the new KL1 facility in Kuala Lumpur.

  • M3 holds Certified Strategic status under the Hosting Certification Framework, which makes Sharon AI’s Supercluster eligible for federal sovereign AI workloads.

What changed in the 15 May Q1 results

Three numbers moved in Sharon AI’s first quarter as a Nasdaq-listed company.

The capacity target. The company now expects 100MW of data centre capacity deployed by early 2027, raised from 70MW a quarter ago. That is a 43 per cent uplift inside three months.

The contract book. Total contracted value sits above US$2.2 billion. Canva is confirmed under a master services agreement. GMI Cloud, ESDS Software Solutions (US$1.25 billion over five years) and an unnamed “global technology company with major Asia-Pacific presence” (US$950 million over five years) are the other named contracts.

The revenue outlook. Sharon AI introduced an exit-2026 revenue run rate of at least US$470 million, with ESDS billing from September 2026 and the unnamed Asia-Pacific deal billing across Q3 and Q4 2026.

On 18 May, Compass Point raised its SHAZ price target from US$50 to US$90 and kept its Buy rating.

The maths at NEXTDC M3 today

NEXTDC’s M3 facility in West Footscray, Melbourne is a 225MW Tier IV campus. It is the largest data centre in NEXTDC’s Australian portfolio.

Sharon AI’s existing 50MW expansion agreement at M3, announced in November 2025, sits inside that envelope. 50MW of 225MW is roughly 22 per cent of the facility’s total power. That share is unusual for a single tenant at hyperscale class. Most large facilities have anchor tenants under 15 per cent.

The Supercluster live today inside M3 uses 1,016 NVIDIA H200 GPUs, connected with NVIDIA Quantum-2 InfiniBand networking. Sharon AI has said the same footprint can grow to more than 20,000 NVIDIA B200, B300 or GB300 GPUs as the 50MW is filled out.

That growth path keeps Sharon AI at M3 well into 2027. But it does not, on its own, reach the new 100MW target. Roughly 30 to 50MW of additional capacity has to land somewhere else, or at M3 in a second tranche, by early 2027.

Where the extra 30 to 50MW plausibly goes

Four candidate sites carry the load.

M3 further expansion. A second tranche at M3 keeps network fabric, cooling and operational team shared, which is the cleanest economics. The constraint is whether NEXTDC has remaining M3 capacity to allocate against pipeline customers. Lifting Sharon AI from 22 per cent of M3 to closer to 40 per cent would make it a near-defining anchor tenant.

S3 Sydney. S3 is an 80MW Tier IV facility in Artarmon, less than 10km from the Sydney CBD. Operational today, right city for a sovereign customer base. 20 to 30MW at S3 would be a clean second site. The 80MW envelope limits how much of the 100MW target S3 can absorb.

M4 Melbourne. M4 is NEXTDC’s A$2 billion AI factory campus at Fishermans Bend, Port Melbourne, approved by the Victorian Government in 2026. Targeted IT capacity is 150 to 162MW. M4 is designed for NVIDIA Blackwell and Rubin Ultra GPUs at rack densities above 1,000kW, which is the workload class Sharon AI is selling. The catch is timing. M4 is in planning. Phased delivery would need to start in 2026 for capacity to be live and tenanted by early 2027. Best technical fit, highest schedule risk.

KL1 Kuala Lumpur. NEXTDC opened KL1 on 14 May 2026, the same day Sharon AI announced its US$950 million Asia-Pacific contract. KL1 targets 65MW total. Phase 1 is 15MW, with 10MW already contracted ahead of opening. KL1 cannot host 30 to 50MW inside Phase 1, but a Phase 2 commitment in 2027 is plausible. See our analysis of NEXTDC’s Asia data centre pipeline after KL1.

The plausible split is a mix. M3 absorbs a second tranche. S3 takes a smaller allocation. M4 holds the longer-dated capacity once it enters service. KL1 carries optionality for the unnamed 14 May customer.

A disclosure correction worth noting

Sharon AI’s annual 10-K, filed on 31 March 2026, called NVIDIA a strategic shareholder of the company. On 13 April 2026, Sharon AI filed an 8-K correcting that. NVIDIA holds no equity in Sharon AI and is not a shareholder. The correction sits under Item 8.01, the SEC’s lighter “voluntary disclosure” category. The 8-K was filed the same day Sharon AI announced its US$1.25 billion ESDS contract. Sharon AI is in the NVIDIA Cloud Partner program. That is a customer link, not a shareholding. Sharon AI’s ASX prospectus, due in June, will need to use the corrected version.

H200 today, Blackwell-class tomorrow

The Supercluster live at M3 runs NVIDIA H200 GPUs, which draw roughly 700W each. The next-generation Blackwell chips (B200, B300, GB300) draw closer to 1.4kW each, and GB300 system densities push rack power above 100kW.

Swapping the same number of GPUs in the same building requires roughly double the power and significantly more cooling. M3 was designed before Blackwell densities were standard. M4 is being designed for them. The migration inside Sharon AI’s footprint is partly a question of which sites can take the heat.

The ESDS and unnamed Asia-Pacific contracts both specify B300-class GPUs in their deal terms. The Blackwell migration is what supports the US$2.2 billion contract book.

The HCF angle: sovereign AI is on the table

NEXTDC’s M3 facility holds Certified Strategic status under the Australian Government’s Hosting Certification Framework. That means M3 meets the federal requirements for hosting sensitive government workloads.

Sharon AI’s Supercluster sitting inside M3 means Sharon AI’s GPU capacity is hosted in a facility eligible for sovereign AI workloads. Most coverage treats the sovereign AI category as ResetData’s lane. The Sharon AI footprint is in the same lane by hosting facility, even if the customer book has not yet turned that eligibility into a public government contract.

For Australian federal agencies running AI projects under the National AI Plan or the new data centre expectations, the choice of provider is no longer ResetData alone. Sharon AI’s Supercluster is qualified by hosting facility, available now, and at a scale unavailable to other domestic neoclouds.

What to watch

  • Site disclosure. Sharon AI’s ASX prospectus, expected in June 2026, should disclose which NEXTDC sites the 100MW target is allocated to. That document will close the analytical question this article opens.

  • The unnamed 14 May customer. The US$950 million counterparty has not been named. Its workload location will partly determine whether KL1 is in the 100MW maths or sits alongside it as an Asia-Pacific overflow.

  • M4 first-phase timing. Any update on M4’s phased delivery schedule directly affects whether M4 is part of the early-2027 picture or a 2028 story.

  • HCF-tied government contracts. Any disclosed sovereign AI workload at Sharon AI’s M3 footprint would mark a meaningful expansion of the federal compute supply base beyond ResetData.