SharonAI Holdings Inc Investment Research Report

Industry and competitive landscape. Demand for AI data centers and accelerated compute is itself a growth industry. McKinsey estimated in 2025 that, to meet global AI demand through 2030, cumulative investment needs across the compute value chain could reach 5.2 trillion dollars, of which AI-related data center demand could reach 156 GW; JLL, in its 2026 outlook, said AI already accounted for about a quarter of data center workloads in 2025 and could approach half by 2030. Demand looks strong, but it comes with multiple bottlenecks across power, land, fiber, construction, capital, and chip supply. Reuters has also recently reported that AI data centers are markedly raising pressure on power infrastructure. The industry is indeed a growth industry, but it is by no means a "make money easily" industry.

Competitors and the company's position. On a global scale, SharonAI does not face a few weak peers but three layers of competition: one layer is hyperscalers such as AWS, Azure, Google Cloud, and Oracle; one layer is AI-native clouds such as CoreWeave, Crusoe, and Nebius; and another layer is local data center operators and systems-integration / sovereign-cloud solution providers. SharonAI itself acknowledges in its risk factors that competitors may have larger market share, more capital, better supplier relationships, cheaper power, and more mature sales and technical capabilities. Its relative advantages lie mainly in first-mover status, ecosystem partnerships, and a localization narrative within the specific niche of Australian sovereign AI infrastructure — for example, being an NVIDIA Cloud Partner, jointly launching Australia's first Cisco Secure AI Factory with NVIDIA alongside Cisco, and jointly deploying GPUaaS with NEXTDC. The issue is that these advantages currently look more like "favorable positioning" than a moat that has already converted into high-return capital.

Moat assessment. Brand advantage: weak. SharonAI has some visibility within Australia's AI infrastructure circle, but is far from a global brand moat. Cost advantage: unproven. Current financials show no cost leadership, and Q1 2026 was still a gross loss. Scale advantage: weak. Compared with top global cloud vendors and CoreWeave, the scale gap is enormous. Network effects: essentially none. Switching costs: low to medium. For customers with long-term hosting and compliance needs there is some switching cost, but compute itself is not a strongly locked-in product. Channel advantage: some early signs, drawn from partners such as NVIDIA/Cisco/Lenovo/WWT/NEXTDC, but highly dependent on partners and not an asset the company owns exclusively. License/regulatory barriers: sovereign data and local deployment create some entry threshold, but not enough to keep out well-capitalized large rivals. Data advantage: weak. Culture/operating capability: unknown, because the history is too short. Capital allocation capability: currently not qualified, or at least not yet proven "good." The company has used an IPO, convertible notes, potential revenue sharing, and equipment financing to drive growth, which is not unusual during an expansion phase, but for shareholders it means both the economic returns and the dilution costs are high.

Scores.

The scores above are based on a combined judgment of the company's 10-K/10-Q, partner announcements, and industry research.

Management, Governance, and Capital Allocation

Is management honest, rational, and long-term oriented? Here we must look both at "economic alignment with shareholders" and at "whether disclosure is reliable." On the favorable side, founders James Manning, Andrew Leece, and Nicholas Hughes-Jones still hold meaningful stakes and control absolute voting power through Class B Super Common Stock. The unfavorable side lies in exactly the same place: the three Class B holders together hold all 136,341 shares of Class B, each carrying 160 votes, and per the company's disclosure the three can jointly decide nearly all matters submitted to a shareholder vote. All officers and directors together hold 68.2% of the voting control. This puts outside common shareholders at a clear disadvantage in governance.

What makes me more cautious is disclosure quality. On April 13, 2026 the company filed an 8-K specifically to correct the statement in its 2025 10-K that "NVIDIA is a strategic shareholder of SharonAI," explicitly admitting the statement was wrong and noting that, as of the correction, NVIDIA did not hold any equity in the company. For a company in a financing- and narrative-driven phase, this kind of erroneous disclosure concerning a core ecosystem partner directly weakens my trust in the quality of management's information. Even if it does not necessarily indicate bad intent, it shows that internal controls and disclosure gatekeeping are not solid enough.

Internal controls and financial reporting. In its 2025 10-K, the company admitted that, due to the accounting treatment of complex financial instruments, there was a material weakness, and that its internal control over financial reporting was ineffective as of the end of 2025. By Q1 2026, the company again clearly disclosed that the material weakness still existed; remediation is underway but has not been confirmed complete. For a company that makes heavy use of convertible notes, fair-value measurement, an IPO, private placements, equipment financing, and potential strategic transactions, this is not a small problem. When the business model is already complex, weak internal controls amplify the risk of "not being understandable" and "not being measurable."

Is capital allocation rational? Over the past two years, the core of the company's capital allocation has not been "efficiently returning excess cash to shareholders" but "continually raising money to chase future deployment." At the end of 2025 the company received about 89.22 million dollars in convertible-note cash; in February 2026 it publicly issued 4.167 million shares at 30 dollars per share, raising about 125 million dollars; in April 2026 the company announced another 350 million dollar, 2031-maturity, 6%-coupon convertible senior note, with an initial conversion price of about 48.24 dollars. In theory these funds all serve GPU, network, and data center deployment, but for shareholders they also mean continued financing dependence, more interest burden, and a high probability of further dilution. The company pays no dividend and has no buyback record. Judged by the Buffett-style standard of "per-share intrinsic value growth," its capital allocation currently looks more like "buying scale and buying survival probability" than creating compoundable per-share cash flow.

Incentives and dilution. Under the company's 2024/2025 equity incentive plans, at the end of 2025 there were 569,500 outstanding options/RSUs to be issued, plus 1.2 million shares reserved under the 2025 plan; 2025 recognized 1.762 million dollars of stock-based compensation, and Q1 2026 recognized another 382 thousand dollars. From the perspective of an early-stage company attracting talent, this is not outlandish; but from the ordinary shareholder's perspective, it means future per-share value will be continually thinned, while the company has not yet proven that these incentives have already produced stable operating results.

Management and capital allocation score: 2/5. The reason is not "the founders have no stake" — quite the opposite, their holdings and control are very strong. The real problem is this: with control so strong, common shareholders need higher-quality disclosure, stronger internal controls, and clearer proof of per-share returns, and the company has provided none of these yet.

Financial Quality and Owner Earnings

The most important point first: SharonAI is not yet a "good business proven by its financial statements"; it is an early-stage infrastructure company whose valuation is supported by expectations of future contracts and financing. Its historical data are short. The main entity was established in February 2024, acquired DSS in June 2024, formed the TCDC joint venture in January 2025, and completed a business combination via Roth CH in December 2025, renaming itself SharonAI Holdings. Therefore, asking it to provide five-to-ten years of continuous operating-quality data is factually impossible. We can only use the disclosed data for the two full years 2024 and 2025 and for Q1 2026.

In the table, 2024A and 2025A are sourced from the audited 10-K; 2026Q1 is sourced from the 10-Q. Free cash flow is an approximation derived from operating cash flow minus cash spent on purchases of property and equipment.

Revenue growth and margins. 2025 revenue grew from 438 thousand to 1.567 million dollars, which looks like fast growth on the surface but starts from an extremely low base; meanwhile Q1 2026 revenue was actually only 294 thousand dollars, below the 325 thousand dollars of the prior-year period. Full-year 2025 gross profit was only 101 thousand dollars, and Q1 2026 swung back to a gross loss of 232 thousand dollars. This shows the company has not yet formed a stable, positive-gross-margin model. If a business today still needs future deployment to explain tomorrow's gross margin, it is far from a "validated good business."

Is the profit real cash profit or accounting profit? 2025 net loss was 39.815 million dollars, yet operating cash flow outflow was only 2.639 million dollars, an enormous gap. The difference comes mainly from several non-cash items: including a 26.031 million dollar loss on the change in fair value of convertible notes, 5.936 million dollars of debt issuance costs, 2.052 million dollars of depreciation and 1.65 million dollars of intangible amortization, and 1.762 million dollars of stock-based compensation, among others. Q1 2026 is the same: a net loss of 20.012 million dollars, including a 70.228 million dollar loss on the fair value of convertible notes, alongside a 65.920 million dollar gain on the sale of the TCDC investment. In other words, accounting profit is severely distorted and net income cannot be used directly as a valuation basis; but this does not mean the company "is actually very profitable," because when operating cash flow and capital expenditure are viewed together, the real cash result is still negative.

Balance sheet and survivability. As of March 31, 2026, the company had 164.29 million dollars of cash on the books, which looks much improved versus the end of 2025; but at the same time, the fair-value liability of convertible notes reached 199.36 million dollars, with total liabilities of 225.26 million dollars. At the end of 2025, shareholders' equity was once negative 10.148 million dollars; in Q1 2026 it turned positive to 88.631 million dollars thanks to the IPO capital raise. This shows the company's near-term ability to "stay alive" comes mainly from financing rather than from operating self-sufficiency. More importantly, as of March 31, 2026 the company also disclosed substantial commitments not yet on the balance sheet: about 92 million dollars of WWT equipment purchase commitments, 70 MW of remaining contracted data center capacity, and an estimated 720 million dollars of capital expenditure expected for the ESDS contract. For long-term shareholders, such off-balance-sheet commitments deserve more attention than a static cash balance.

Working capital, receivables/payables, and capex intensity. In 2025, accounts receivable and other receivables did not deteriorate substantially, indicating that the current revenue scale is too small for working capital to be a major risk source; the real risk lies in the match between capital expenditure and financing. In 2025, cash spent on purchases of property and equipment was about 10.951 million dollars, already far above revenue over the same period; and the ESDS contract alone is expected to require 720 million dollars of capital investment. To use Buffett's phrasing, this is not a "put in a little less capital and take out a little more cash" business, but a "you must keep feeding it capital for it to have a chance to grow" business.

Owner earnings assessment. Looking at it in the most conservative way, closest to the Buffett framework, 2025 owner earnings should not start from net income but should be based on operating cash flow: 2025 operating cash flow was -2.639 million dollars; after further deducting at least the minimum maintenance capital expenditure related to keeping existing equipment running, owner earnings remain negative. Because the company has not yet disclosed a reliable split between "maintenance capex and growth capex," we cannot rigorously break down the 10.951 million dollars of capital expenditure; but regardless, 2025 owner earnings were negative and Q1 2026 owner earnings were very likely still negative, and this conclusion does not depend on a precise split. For a long-term investment target, this point alone is enough to greatly discount the "bull case."

Therefore, I give a conservative owner-earnings estimate: On a 2025 basis, owner earnings are roughly in the range of -5 million to -14 million dollars. The upper bound of the range assumes much of 2025 capex was growth-related and deducts only minimum maintenance capex; the lower bound is approximated directly using the free-cash-flow convention. Whether you take the upper or lower bound, current owner earnings are negative, so the question of "how many times owner earnings the current valuation represents" is not applicable / economically meaningless on a static basis. This is a very important, and very unwelcome, conclusion of this report.

Valuation, Margin of Safety, and Opportunity Cost

As of May 20, 2026, SHAZ was quoted at 52.48 dollars, with a market cap of about 733 million dollars.

The current valuation must first confront a "share-count illusion" problem. On the surface, a 733 million dollar market cap does not seem extravagant; but once potential dilution is counted in, the economic valuation is higher. The company discloses: as of May 13, 2026, Class A shares outstanding were about 16.6079 million shares; the December 2025 convertible notes can convert into up to 8.251 million shares; the 350 million dollar convertible note announced in April 2026, based on an initial conversion price of 48.24 dollars, corresponds to roughly 7.25 million shares; adding the roughly 569,500 equity incentives already in place at the end of 2025, the roughly fully diluted share count is about 32.6 million to 32.8 million shares. This is an inferred figure and does not include the future potential additional issuance of 8.5 million shares the company has flagged, nor the additional shares reserved under new plans. Therefore, at today's share price, the economic, fully diluted equity value is roughly close to 1.7 billion dollars, significantly above the 733 million dollar market cap shown on the face of the statements.

Valuation method one: owner-earnings discounting. Because current owner earnings are negative, only a highly assumption-driven forward valuation is possible. There are two factual anchors: first, the two large contracts the company has publicly disclosed total a TCV of about 2.2 billion dollars over five years, equivalent to annualized revenue of about 440 million dollars at full execution; second, the ESDS contract alone is expected to require about 720 million dollars of capex, indicating that cash-flow realization will not be easy. On this basis, I build three scenarios and deliberately keep the premises conservative: Conservative scenario: only part of the contracts land on schedule, mature-stage annual revenue of 200 million to 250 million dollars, an owner-earnings margin of 5%, a discount rate of 15%, and a terminal growth rate of 2%; the corresponding per-share intrinsic value is roughly 3 to 8 dollars. Neutral scenario: the two announced large orders are broadly fulfilled, mature-stage annual revenue of 300 million to 400 million dollars, an owner-earnings margin of 8% to 10%, a discount rate of 12%, and a terminal growth rate of 3%; the corresponding per-share intrinsic value is roughly 8 to 18 dollars. Optimistic scenario: contracts are highly fulfilled, deployment goes smoothly, and the cost of capital is controllable, mature-stage annual revenue of 440 million dollars or more, an owner-earnings margin of 12% to 15%, a discount rate of 10%, and a terminal growth rate of 4%; the corresponding per-share intrinsic value is roughly 18 to 30 dollars. These ranges all use fully diluted shares as the denominator and do not fully incorporate potential dilution such as the future additional issuance of 8.5 million shares; so they are not harsh. By this framework, the current 52.48 dollars is above the upper end of my judgment for the optimistic scenario.

Valuation method two: relative valuation. Static P/E, EV/EBITDA, and P/FCF are almost meaningless for SHAZ, because the company is still loss-making with negative free cash flow. More useful is comparing the valuation constraints of "stronger players in the same field" and "mature asset platforms." One of the closest strong rivals in the public market is CoreWeave: as of May 20, 2026, CRWV had a market cap of about 52.5999 billion dollars; CoreWeave's Q1 2026 revenue was 2.078 billion dollars, adjusted EBITDA was 1.157 billion dollars, and as of March 31, 2026 its revenue backlog reached 99.4 billion dollars. By contrast, SHAZ's full-year 2025 revenue was only 1.567 million dollars and Q1 2026 revenue was only 294 thousand dollars, while the historical revenue multiple implied by the current share price is already absurdly high. Even if you do not use a static revenue multiple but instead imagine SharonAI's future annualized revenue from its two announced large orders at 440 million dollars, viewed against a fully diluted economic equity value of about 1.7 billion dollars, it still implies highly optimistic expectations for successful execution, smooth financing, and a turn to positive margins. A small platform that has not yet proven its gross-margin model should face the industry leader at a discount, not a premium.

Valuation method three: asset/liquidation. As of Q1 2026, the company's shareholders' equity was 88.631 million dollars, but this includes 18.044 million dollars of goodwill and 6.906 million dollars of right-of-use assets; looking only at the more conservative tangible-net-asset support, it is far below the current market cap of about 733 million dollars. More critically, the company also had 199.36 million dollars of convertible-note liabilities over the same period. In other words, the current share price has almost no "asset floor" protection; what shareholders are really buying is the ability to fulfill future contracts, not existing net assets. For a value investor, this means that if execution slips, the downside is very large.

Margin-of-safety conclusion. I give the following three valuation ranges: Conservative intrinsic value range: 3 to 8 dollars per share. Fair intrinsic value range: 8 to 18 dollars per share. Optimistic intrinsic value range: 18 to 30 dollars per share. At the current 52.48 dollars, the share price carries a large premium to each of the ranges above; from a value-investing perspective, the margin of safety is not merely "unclear" — it is essentially absent. In terms of my own framework, if I must give a price band: the ideal buy range is roughly 5 to 10 dollars; the acceptable holding-price range is roughly 10 to 18 dollars; the clearly overvalued range is roughly above 30 dollars. These ranges are not trading forecasts but business-value judgments based on the full dilution, execution risk, and capital intensity described above.

Comparison with indices and the risk-free rate. The U.S. Treasury website shows that the U.S. 10-year Treasury yield on May 19, 2026 was about 4.67%. For an investor with a balanced risk appetite, SHAZ must offer a long-term return, after risk adjustment, clearly above 4.67% to justify tying up capital; but at the current price, the return I assign in the neutral scenario is not attractive. Compared with a broadly diversified equity basket like the S&P 500, SHAZ carries extremely high single-company execution, financing, supply-chain, and governance risk, and I do not see enough "odds compensation." If your portfolio could hold only five assets, it does not qualify for selection right now.

Risks, Checklist, and Final Judgment

The most important risks. From a "permanent loss of capital" standpoint, the core risks are not short-term volatility but the following categories: First, execution risk: whether the company can complete GPU delivery, data center readiness, network bring-up, and customer acceptance on schedule for its large contracts. The company itself has flagged in its risk factors that hardware delivery delays, unavailable data center capacity, or commissioning overruns could all lead to performance failure. Second, capex and financing risk: the ESDS contract alone is expected to require 720 million dollars of capex, while as of Q1 2026 the company has only temporarily replenished funds through the IPO and convertible notes; if subsequent deployment runs over budget or funding conditions worsen, shareholders may face continued leverage or continued issuance. Third, supply-chain risk: the company is clearly dependent on NVIDIA GPUs, with no long-term agreement that locks in price and supply quantity. If upstream prioritizes larger customers, SharonAI could directly lose delivery capability. Fourth, customer concentration risk: in 2025, 82% of revenue came from three customers, and high concentration is still expected going forward. Any change in demand from a single large customer could alter the company's financial outlook. Fifth, data center and power concentration risk: the company discloses that its infrastructure is currently substantially concentrated within the NEXTDC system. As long as a single data center / power supply / regulatory node has a problem, the company is affected almost across the board. Sixth, accounting and governance risk: as of Q1 2026, the material weakness in internal controls related to complex financial instruments is still unremediated; layered with the correction of the erroneous "NVIDIA is a shareholder" disclosure, this shows disclosure quality is not yet good enough for me to comfortably make a large allocation. Seventh, dilution risk: the 2025 convertible notes, the April 2026 convertible notes, and the equity incentives all mean future per-share value faces substantial dilution; the company has also flagged an additional potential issuance of 8.5 million shares.

The strongest counterargument. If I were a bear, I would put it this way: SharonAI is currently selling the market the dream of an "Australian sovereign AI compute platform," but what today's real statements actually show is extremely low revenue, negative gross margin, negative free cash flow, complex financing, and heavy dilution; the so-called publicly disclosed TCV on the order of 2.2 billion dollars has not yet converted into verifiable operating results, and fulfilling those contracts in fact requires the company to commit enormous capital expenditure first. Put differently, this may not be an undervalued high-quality business, but an overvalued early-stage infrastructure story stock.

Which facts would overturn the current judgment? Three sets of facts matter most. First, seeing the announced large orders genuinely convert into revenue for at least four to six consecutive quarters, with gross margin continuously turning positive and expanding. Second, without seriously harming per-share value, the company proving it can complete deployment and generate positive operating cash flow and owner earnings close to positive. Third, the material weakness in internal controls being formally remediated, with management showing greater consistency and transparency in subsequent disclosures. If these three appear, I would be willing to raise the rating from "watch"; if they fail to appear for a long time, or if reverse signals appear — such as contract delays, customer cancellations, continued large dilution, supply/data center delays, or gross margin staying negative for the long term — then I should acknowledge that the earlier optimistic judgment was wrong.

Investment checklist.

This checklist is based on the preceding combined judgment of the business, moat, financials, governance, and valuation.

Final judgment.

【Final Rating】 Watch

【One-sentence investment thesis】 SharonAI sits in a very hot field with a very large contract narrative, but as of now it remains an early-stage AI infrastructure company that has not yet proven cash-flow quality and is already priced for high expectations and high dilution risk.

【Core bull case】 The AI data center and GPU cloud field the company sits in has strong demand; the company has some niche advantage in its "Australian sovereign AI infrastructure" positioning; it has secured ecosystem partnerships with NVIDIA/Cisco/NEXTDC/VAST/Lenovo/WWT and others; the disclosed public contract TCV is significant in scale; and its financing ability is not weak in the near term, having at least temporarily won market attention and capital support.

【Core bear case】 Real revenue and cash flow are far smaller than the narrative scale; positive gross margin and positive free cash flow are still unproven; customers, supply chain, data centers, and power are all highly concentrated; control is highly concentrated yet internal controls still carry a material weakness; the potential dilution from equity and convertible notes is very large, and the current price provides no margin of safety.

【Key assumptions】 For the investment to hold, several must be true at once: the announced large orders convert into revenue on schedule; GPU and data center capacity are delivered as planned; the mature-stage owner-earnings margin reaches at least mid-to-high single digits; future financing does not significantly dilute per-share value; and the quality of internal controls and disclosure improves markedly.

【Fair buy price】 By the conservative-to-neutral valuation framework in this report, I believe a more reasonable observational buy range should be 5 to 10 dollars; if the company can prove improvement in gross margin, contract fulfillment, and cash flow over the next few quarters, the valuation range could be revised upward, but the current 52.48 dollars does not present a buy point in the value-investing sense.

【Target holding period】 If future operating quality is validated, it suits observation and long-term holding over five to ten years or more; but before validation, it does not suit ignoring current valuation and governance risk just because of the words "long term."

【Expected annualized return】 Based on the current price, in the conservative scenario I expect the long-term annualized return to be roughly -15% to -10%; in the neutral scenario roughly -8% to -2%; and in the optimistic scenario 3% to 8%. This set of ranges is not a short-term forecast but a long-term return projection based on full dilution, execution risk, and the valuation ranges.

【Maximum loss risk】 If contracts are delayed or canceled, GPU/data center delivery falters, subsequent refinancing conditions worsen, and existing or new convertible notes continue to suppress common equity, a loss of more than 80%, or even close to permanent loss, is not unimaginable, because the company currently lacks a sufficiently thick asset floor and a validated cash-flow floor.

【Tracking metrics】 What is most worth tracking going forward is not the share price but: the cadence of quarterly revenue recognition; gross margin and operating cash flow; the go-live timing of the announced large orders and the realization of customer prepayments/guarantees; the amount of capex realized and the cost of financing; the actual number of GPUs deployed and available MW; changes in Class A share count and convertible-note terms; the progress of remediating the internal-control weakness; major-customer concentration; and whether the depth of partnerships with NVIDIA/NEXTDC/Cisco and others substantively strengthens.

【Signals that trigger reassessment】 If any of the following occur, reassess immediately: delays in contract delivery milestones; gross margin failing to turn positive for several consecutive quarters; new large, high-cost financing or low-priced issuance; internal-control problems persisting or new corrections/restatements appearing; an adverse change in the NEXTDC / key supply-partner relationship; or a reduction in orders from a single large customer.

【Final recommendation】 Put plainly, this is not a stock suited to a "Buffett-style long-term business owner" framework for a heavy buy at the current price. It looks more like a heavy-capital, strong-narrative, strong-execution-dependent infrastructure startup that still needs continued validation. For a long-term investor with a balanced risk appetite, my recommendation is restrained: put it on the watch list, not in the core position. Wait for the financial results to catch up with the story before discussing price; do not let the story stand in for cash flow.

Information limitations. The company's complete auditable operating history is short, and many future-return judgments can only be built on disclosed contracts, financing terms, and industry-experience assumptions; in addition, some of the company's 2026 follow-on financing/project progress remains, in the public materials used here, in a state of "signed or announced, but not yet fully validated by longer-term financial reports." Therefore, this report treats intrinsic value with a markedly conservative approach that leans toward the fully diluted convention.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.