SharonAI Holdings, Inc. (NASDAQ:SHAZ) had a rough start to the week. The stock closed down 10.77% on Monday and kept sliding in after-hours trading following a third-quarter report that missed Wall Street estimates by a wide margin. By Tuesday premarket, shares were mostly flat, up just 0.33% at $20.19.
The numbers were ugly on the surface. Revenue came in at $9.93 million, missing the consensus estimate of $26.94 million. The quarterly loss was $1.14 per share, nearly double the analyst estimate of a 59-cent loss. That's a double miss, and investors reacted accordingly.
But here's the thing about companies building AI infrastructure: the income statement often lags the contract book by quarters, sometimes years. And SharonAI's contract book is doing something dramatic.
NVIDIA's AI Compute Program Is the Centerpiece
SharonAI signed a six-year agreement under NVIDIA Corp. (NVDA)'s AI Compute Program (AICP) back in the second quarter of 2026. The deal covers 72 megawatts of capacity and roughly 40,000 NVIDIA GB300s. Commissioning is expected in the first quarter of 2027, followed by NVIDIA Infrastructure Service Testing in the second quarter. Customer billing should begin in the third quarter, with deployment accelerating through the fourth.
NVIDIA's take-or-pay commitment gives SharonAI meaningful revenue visibility. The company plans to resell capacity at prevailing market rates, keep 100% of NVIDIA's anchor price, and share additional revenue with the chipmaker.
Management is also leaning toward shorter-duration contracts under the program. Why? Because they can generate more than $30 million in annualized revenue per megawatt. Meanwhile, long-term contract pricing has climbed from roughly $15 million to more than $19 million per megawatt annually in 2026. The NVIDIA program alone accounts for $4.88 billion in anchor contract value and is expected to represent more than 75% of SharonAI's secured debt requirements. Financing, the company says, is well advanced.
The Backlog Keeps Piling Up
SharonAI has secured 212 megawatts of AI factory capacity across Australia and New Zealand. That's nearly four times the 54 megawatts reported in February 2026. Of that capacity, 129 megawatts are under multiyear take-or-pay agreements. The company plans to deploy more than 68,000 NVIDIA GPUs across 2026 and 2027.
During the quarter, SharonAI signed a five-year, $1.32 billion agreement with a global AI laboratory. Subsequent deals included a four-year, $170 million contract with a U.S.-based inference provider and a five-year, $1 billion agreement with a global services company.
Total contract value reached $9.6 billion as of Nov. 16. For context, that figure was $1.3 billion in the first quarter, $7.1 billion in the second quarter, and $8.8 billion in the third quarter. The trajectory is steep, to put it mildly.
Revenue Growth and Deployment Timeline
SharonAI's annualized run-rate revenue (ARR) stood at approximately $50 million in September. The company expects ARR to reach $400 million to $450 million by January 2027. That's eight to nine times the September level.
The anticipated increase reflects roughly 30 megawatts, or 23% of contracted capacity, entering the billing phase. The remaining capacity is scheduled to come online in stages throughout 2027. Management aims to achieve project-level contribution margins above 80% at full deployment.
Technology and Strategic Partnerships
SharonAI achieved SOC 2 Type II attestation and completed NVIDIA Infrastructure Service Testing across live deployments. It is also pursuing ISO 27001 certification.
In June, the company signed an agreement with VAST Data covering 600 petabytes of AI storage capacity, enough to support approximately 100,000 GPUs. The two companies also launched Data Enclave, which lets AI models operate within customer-controlled infrastructure.
In September, SharonAI signed a five-year agreement with Rafay to manage and coordinate operations across its AI factories. The platform can support up to 150,000 GPUs, helping accelerate deployments, improve utilization, and standardize operations.
Market Outlook
Management sees Australia as a potential global exporter of AI computing capacity and tokens, citing its land availability, energy resources, power supply, water security, and renewable energy potential. Australia's operational data center capacity is projected to increase from approximately 1.6 gigawatts in 2026 to 3.9 gigawatts by 2030, according to Data Centers Australia.
SharonAI follows a capital-light strategy, primarily deploying its computing infrastructure at existing or developing grid-connected Tier 3 and Tier 4 partner data centers under multiyear contracts. It's a model that keeps capital expenditures in check while still scaling capacity.
For now, the market is focused on the earnings miss. But the contract backlog and the NVIDIA partnership suggest the real story is about what happens when all that capacity starts billing.