Finvest
SHAZ AI Infrastructure · Neocloud · GPU compute · Australia · Thesis updated July 15, 2026

Big AI contracts, bigger delivery test

01 Running thesis

Demand is less of a guess

SharonAI has moved from a tiny, concentrated revenue base to a much larger signed-demand story. The company reported a $1.26 billion master services agreement with ESDS, then added an APAC technology customer contract worth about $950 million. Together, those deals push the backlog above $2.2 billion.

That changes the main question. The bull case is no longer only, “Can SharonAI find customers?” It is now, “Can SharonAI deliver?” If the company can secure enough NVIDIA GPUs, light up the 29.6 MW of new capacity, and start billing on time, revenue could scale far faster than a normal data center builder.

The bear case is simple. These are huge contracts for a company that had only $1.6 million of revenue in 2025 and a net loss of $39.8 million. The company must procure scarce GPUs, install them, run them reliably, and fund the buildout before the full contract revenue arrives.

The open question is margin. The filings give the size of the contracts, but not enough detail on delivery timing, GPU cost, utilization, or profit per dollar of revenue. Those details will decide whether the backlog becomes valuable cash flow or just a hard growth promise.

May 2026The story moved from proving demand to proving execution. SharonAI disclosed more than $2.2 billion of new contracts and about 29.6 MW of added data center capacity, while also adding a new AI export control risk.
Mar 2026The initial view was set from the 2025 Form 10-K. SharonAI had pivoted from storage to GPU cloud, but still had small revenue, heavy losses, and high customer concentration.
02 Business model

Cloud without owning the buildings

SharonAI is a neocloud. That means it sells cloud access to specialized computing hardware, mainly GPUs, for AI training, AI inference, research, rendering, and other high-performance workloads. Customers pay for managed compute instead of buying and running the hardware themselves.

The company tries to stay capital-light by placing its equipment in partner data centers, such as NEXTDC, rather than building its own sites from scratch. This can speed up expansion and reduce the time needed to add power, cooling, and network capacity.

The model still needs a lot of money. SharonAI must buy or finance expensive GPUs before customers can use them. It also depends on partner data centers, hardware vendors, and network partners to be ready at the same time.

In 2025, 82% of revenue came from 3 customers. The new ESDS and APAC contracts could diversify that base after deployment, but until billing begins, concentration and timing risk remain.

03 Product portfolio

What SharonAI sells

Growth engine

GPU Cloud services

This is the core business now. SharonAI provides managed access to GPU infrastructure for AI and high-performance computing customers.

Growth engine

SharonAI Cloud

This is the company’s orchestration and automation platform. It helps customers provision GPU compute for AI training, inference, rendering, and other workloads.

Growth engine

Sovereign Australian AI compute

SharonAI offers low-latency compute hosted in Australia. That matters for customers that need data to stay under Australian privacy and data rules.

Option

Supercluster capacity

The company has built and planned larger GPU clusters hosted with partners. This could support large customers, but it depends on power, cooling, and GPU delivery.

Steady

Legacy Filecoin storage

SharonAI wound down Filecoin-related activities in 2025. This line is no longer the center of the company’s strategy.

04 Business segments

A pivot in the numbers

GPU infrastructure services91%growing fast
Legacy Filecoin storage9%declining

The mix is based on 2025 revenue categories in the Form 10-K. GPU infrastructure is now the main revenue line, while Filecoin storage was wound down in the second quarter of 2025.

05 Risk factors

What could break the story

GPU supply shortfall

High impact · Medium odds

The backlog depends on getting enough advanced NVIDIA GPUs, including B200, B300, and GB300 systems. If supply is delayed or allocations go elsewhere, SharonAI may miss service start dates. That would slow revenue and could strain customer trust.

We watchLook for disclosed GPU purchase orders, deployment counts, and customer service start dates tied to ESDS and the APAC contract.

Data center partner bottleneck

High impact · Medium odds

SharonAI does not build most of its own facilities. It depends on partners such as NEXTDC and other data center providers for power, cooling, space, and uptime. The company has signed for about 29.6 MW of additional capacity, but that capacity must still become usable.

We watchTrack updates on the 29.6 MW capacity, start dates beginning in late 2026, and any changes to NEXTDC or other hosting arrangements.

Backlog without margins

High impact · Medium odds

Large contracts do not automatically mean strong profits. SharonAI has not yet shown the margin profile for the ESDS and APAC contracts. GPU financing, data center fees, networking costs, and utilization could decide whether the deals create cash or consume it.

We watchWatch gross margin, cash from operations, and any contract-level color on pricing, utilization, or monthly billing.

Cash burn before revenue ramps

High impact · Medium odds

The company raised $125 million with its Nasdaq listing, and it also reported earlier capital raises. Even so, it lost $39.8 million in 2025 and must fund equipment and deployment before major revenue arrives. More financing could dilute shareholders or add debt.

We watchMonitor cash balance, operating cash flow, new debt or equity raises, and capex for GPU purchases.

AI export control shock

High impact · Low odds

The 2026 Q1 filing added a risk that US or foreign government rules could force SharonAI to limit, suspend, or terminate service to certain customers. This matters because advanced AI chips and AI compute are increasingly tied to trade policy and national security rules.

We watchTrack US AI chip export controls, customer geography disclosures, and any filing language about restricted customers or services.
06 Quick answers

In one breath

What does SharonAI do?

SharonAI sells managed access to GPU cloud infrastructure for AI and high-performance computing. It focuses on Australian, low-latency compute and places its hardware in partner data centers.

Why did the SharonAI thesis change in 2026?

The company signed more than $2.2 billion of long-term contracts, including $1.26 billion with ESDS and about $950 million with an APAC technology customer. That makes demand less of the main issue, while GPU supply, deployment timing, and margins become the main issues.

Is SharonAI profitable?

No. SharonAI reported $1.6 million of revenue and a $39.8 million net loss for 2025, so the company is still in a heavy investment phase.

What should investors watch next?

Watch for GPU procurement, live deployment of the 29.6 MW of added data center capacity, and monthly billing under the ESDS and APAC contracts. Gross margin and cash burn will show whether the backlog is turning into a healthy business.