Sharon AI Revenue Up 1,859% to $9.9M as Net Loss Widens to $43.8M
Sharon AI grew Q3 revenue 1,859% to $9.9M, but a $43.8M net loss and a $9.6B contract book reveal a GPU build-out racing to start billing by January 2027.

Why it matters
- Sharon AI reported Q3 2026 revenue of $9.9 million, up 1,859% year over year, with net loss widening to $43.8 million.
- Contract book reached approximately $9.6 billion, including $4.88 billion tied to the six-year NVIDIA AI Compute Program covering ~40,000 GB300 GPUs.
- Investing activities used $477.6 million in the quarter; cash and equivalents stood at $1.34 billion at September 30, 2026.
- Company anticipates ARR of $400–450 million by January 2027, up from ~$50 million in September 2026, pending cluster acceptance.
- Sharon AI has secured ~$2.6 billion of institutional debt and equity since December 2025, including a $356 million Goldman Sachs-backed facility at 9.95%.
Sharon AI posted third-quarter revenue of $9.9 million, up 1,859% from a year earlier, while its net loss widened to $43.8 million from $1.6 million. The results, reported on November 16, 2026 for the quarter ended September 30, 2026, capture the core tension in the Australian GPU cloud provider's business: a contract book of roughly $9.6 billion against recognized revenue that is still measured in the tens of millions.
The gap is the story. Sharon AI is financing and building out far more infrastructure than it is currently billing, and the numbers show exactly where the money goes — and when some of it might come back.
What is actually contracted?
Sharon AI's headline metric is total contract value (TCV): the company's estimate of committed spending under binding customer contracts. As of October 31, 2026, TCV carried a weighted-average remaining term of 5.5 years. Two caveats matter here: TCV excludes non-binding agreements, and it can change through modifications or terminations. It is not recognized revenue.
The announcement includes $2.5 billion contracted since the second quarter of 2026 alone. That figure widens the distance between GAAP revenue and the much larger pool of business the company is committed to delivering.
The single largest item is NVIDIA. The earnings presentation assigns $4.88 billion of TCV to the six-year NVIDIA AI Compute Program, covering approximately 40,000 GB300 GPUs. Partial billing is indicated for the third quarter of 2027. Under the arrangement, Sharon AI can resell that capacity to end customers, retain the anchor price on capacity sold, and share any upside above that price with NVIDIA.
The takeaway: the contract book is large, long-dated and real, but it converts to cash only as clusters pass acceptance.
How close are the GPUs to billing?
The quarter moved actual hardware into place. A global technology customer accepted a first cluster of approximately 1,000 NVIDIA B300 GPUs in August 2026 — about 20% of the GPUs in that deployment. That acceptance is a milestone because billing follows it.
After quarter-end, data centers were energized for clusters of approximately 8,000 B300 and 4,000 GB300 GPUs. Energization is not the finish line. Commissioning, testing and NVIDIA Infrastructure Services verification are still required before customer billing can begin, and customer acceptance must follow.
Sharon AI expects to complete commissioning, testing and NVIDIA verification for the global technology and ESDS deployments in December 2026 and January 2027. Those dates anchor the company's own revenue expectations for early next year.
In short: the GPUs are powered on, but nobody is paying for most of them yet.
Where did the money go?
The operating loss reached $37.5 million. Adjusted EBITDA — a non-GAAP measure excluding interest, tax, depreciation and amortization, share-based compensation and other adjustments — came in as a loss of $25.8 million.
Infrastructure spending dwarfs both figures. Investing activities used $477.6 million during the quarter, including $293.1 million of advance payments for property and equipment. Cash and cash equivalents totaled $1.34 billion at September 30, 2026. That funding base sits alongside a build-out that consumes capital before most contracted capacity bills.
Sharon AI has secured approximately $2.6 billion of institutional debt and equity capital since December 2025, excluding customer deposits. A committed $356 million GPU-backed senior secured facility with Goldman Sachs and private credit funds carries a fixed 9.95% rate, secured by the GPUs, associated cash flows and a temporary parent guarantee. Additional customer deposits are typically held in escrow until cluster acceptance.
The mechanism is worth stating plainly: lenders finance GPUs, customers deposit against future capacity, and revenue starts only when NVIDIA verifies the infrastructure and the customer signs off. Every stage in between burns cash.
The takeaway: Sharon AI is spending at data-center scale while billing at startup scale, and its $1.34 billion cash position has to bridge that gap.
What does the January target test?
The company's own forward-looking numbers frame the next checkpoint. Sharon AI reported annualized run-rate revenue (ARR) of approximately $50 million for September 2026 and anticipates $400 million to $450 million for January 2027 — a roughly ninefold jump. The company expects about 30MW of its 129MW of contracted capacity to be billing at that point. ARR annualizes expected monthly contract revenue; it is not GAAP revenue and not a revenue forecast.
The January target is a specific, checkable claim. It depends on verification and customer acceptance finishing in December 2026 and January 2027 for the global technology and ESDS deployments. If energized clusters clear those gates on schedule, recognized revenue should follow. If they slip, the gap between the $9.6 billion contract book and the income statement stays open longer.
Beyond that, Sharon AI expects 212MW of total capacity across Australia and New Zealand in 2027. Approximately 155MW of that is described as available for sale, comprising uncontracted capacity and NVIDIA program capacity to be resold to end customers.
That unsold capacity cuts both ways. It is upside if demand holds, and it is carrying cost if it doesn't.
The quarter leaves Sharon AI with a large contract book, a $2.6 billion institutional funding base since December 2025 and $1.34 billion in cash — but still modest recognized revenue and widening losses. The immediate test is whether verification and customer acceptance turn the energized clusters into billing on the January schedule, while the larger NVIDIA program extends those delivery and financing demands through 2027.
Source: unite.ai
More from Sophie Lindqvist
Show full bio
Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.
749 articles
Related articles
- Nebius Prepayments Now Cover More Than 100% of Capex as Q3 Revenue Hits $980.8M
- Anthropic and Nscale Reportedly Near a Second $45 Billion Compute Deal
- OpenAI's 2026 revenue projection drops $20bn short of investor guidance
- Firmus Seeks $3 Billion Privately, With Nvidia and Meta Prepayments
- Nvidia Carries $164.5 Billion in Commitments Into CEO Vacuum