Big AI contracts, bigger delivery test
- SharonAI rents access to high-end GPUs instead of building its own data centers.
- The company has shifted away from Filecoin storage and toward GPU cloud services.
- Recent contracts total more than $2.2 billion, including $1.26 billion with ESDS and about $950 million with an APAC technology customer.
- The hard part is now execution: buying enough B200, B300, and GB300 GPUs and turning data center space into live revenue.
- A 2026 Nasdaq listing raised $125 million, but losses and cash needs remain central to the stock.
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.
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.
What SharonAI sells
GPU Cloud services
This is the core business now. SharonAI provides managed access to GPU infrastructure for AI and high-performance computing customers.
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.
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.
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.
Legacy Filecoin storage
SharonAI wound down Filecoin-related activities in 2025. This line is no longer the center of the company’s strategy.
A pivot in the numbers
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.
What could break the story
GPU supply shortfall
High impact · Medium oddsThe 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.
Data center partner bottleneck
High impact · Medium oddsSharonAI 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.
Backlog without margins
High impact · Medium oddsLarge 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.
Cash burn before revenue ramps
High impact · Medium oddsThe 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.
AI export control shock
High impact · Low oddsThe 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.
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.