SharonAI's $8.8 Billion Contract Book Meets Its First Real Revenue Test After 59% Stock Slide

SharonAI enters its September-quarter report with one of the widest gaps in the AI infrastructure space between what it has signed and what it has actually delivered. Last quarter, management unveiled a six-year, $4.9 billion compute collaboration with NVIDIA, three more multi-year take-or-pay contracts, and a contracted book that roughly quadrupled to about $8.8 billion. The market has since cut the stock by nearly 59%, so this report is less about proving demand exists and more about proving the company can turn paper into power, GPUs and invoices.

The Street is looking for revenue of about $18.3 million, nearly ten times the $1.9 million reported last quarter, alongside a loss of $1.47 per share, much wider than the prior quarter's $0.27 loss. That combination tells the story of a business at an inflection point: revenue is expected to start appearing, but costs tied to the build are expected to grow faster for now. Management has not issued formal guidance for the quarter, but it repeatedly framed the fourth quarter of 2026 as the period when material revenue begins, with hardware deliveries and data center readiness as the swing factors. That makes the September quarter a bridge period, and the consensus revenue figure effectively assumes some early clusters, such as the first B300 deployment handed to a customer last quarter, are already billing.

The most useful evidence will be operational rather than financial. Investors should want confirmation that the 212 megawatts of secured capacity targeted by the end of 2027 is moving on schedule, that ready-for-service dates with third-party data center partners and Super Micro are holding, and that the goal of more than 64,000 NVIDIA GPUs by mid-2027 remains intact. Any slippage in delivery timing would directly challenge the fourth-quarter ramp, which is the hinge of the entire narrative. Conversely, a clean update showing clusters energized and customers onboarded, particularly on the NVIDIA GB300 build, would validate last quarter's confidence.

Financing is the second major test. SharonAI has raised roughly $2.2 billion since December, including an oversubscribed $1.6 billion round and a $350 million convertible note, but the large debt facilities needed to fund GPU purchases were still at the term-sheet stage. Closing that debt on reasonable terms would remove a key overhang; further delays or dilutive alternatives would likely explain some of the stock's steep decline. The finance function is also in transition, with Tim Broadfoot stepping back and former Macquarie executive Anuj Gul taking over as CFO, so the clarity of capital disclosures this time carries extra weight.

Pricing deserves a look too. Management highlighted a record of more than $4 per GPU-hour on B300 capacity, but the NVIDIA arrangement guarantees only a floor of roughly $817 million a year, with upside tied to placing higher-rate customers. Any early sign of re-contracting at premium rates would support the bull case that the floor is a minimum rather than the realistic outcome.

Sentiment has softened from deeply bearish to moderately bearish since the last report, even as the shares fell. At $21.81, the stock trades less than half its 200-day moving average of $47.91 and sits not far above its post-earnings low of $19.01, a long way from the $80 peak. That suggests expectations have been reset sharply lower, which lowers the bar but does not remove the skepticism. The central question is simple: can SharonAI show that its deliveries, data centers and debt financing are aligned closely enough to make the fourth-quarter revenue ramp credible? If so, the contracted book starts to look like a pipeline of cash flow. If not, the market will keep treating it as a promise.

This earnings preview was originally published on Earnings Whispers.

This article is for informational purposes only and is not investment advice. Earnings Whispers and its staff may hold positions in the securities mentioned. It is AI-assisted, based on proprietary Earnings Whispers data and company reports, and edited by Earnings Whispers staff.