Xiao-I Corporation (NASDAQ: AIXI) — A Going-Concern AI Shell, Reverse-Split to Survival and Pumped on a Patent It Can’t Bank

⚡ Claude’s Take

This block is the author’s own independent opinion and general information only — not investment advice. The analytical body that follows (sections 1–15) deliberately carries no recommendation and no price target.

Verdict: AVOID — not investable as an equity; not a short you can responsibly carry either (borrow is scarce/expensive, recall risk and ~200% idiosyncratic volatility make the squeeze risk lethal). For all but a disciplined options-defined speculator, the right position size is zero. Any “valuation zone” is a statement about option value, not intrinsic value: I would not pay for the equity above a deep-distress, sub-$2 (ADS) call-option level, and even there it is a lottery ticket, not an investment.

Xiao-I is not a cheap AI company; it is a near-insolvent VIE shell wearing an AI costume. The facts that matter are not the conversational-AI product slides — they are these: revenue collapsed −82.5% in FY2025 ($70.3M → $12.3M); the company booked a $101.8M net loss of which roughly $59.5M was a single write-off of receivables, supplier advances, and a loan to a 26%-owned related party; book equity is negative $104.9M; total cash is $2.4M against $47M of debt; the auditor and management both state substantial doubt about going-concern; the only financing the company can raise is death-spiral convertible notes (conversion at 85% of the lowest 10-day VWAP); it has executed a cumulative ~1-for-180 reverse split since its 2023 IPO purely to stay listed; the founder handed himself >50% of the votes via super-voting preferred issued for no cash; the CEO installed in January 2026 is a former liquor-and-pharma salesman; and two US securities class actions allege the IPO itself overstated the company’s AI capabilities and customers. The “moat” thesis is disproven by the financials — a real moat does not let 82% of revenue and the receivables behind it evaporate in twelve months.

The recent price action is the tell. After a May-2026 1-for-20 reverse split reset the quote, the ADS ran from ~$11.73 to $15.47 (May 22) on a favorable Apple-patent headline and pure AI-momentum, then collapsed −60% to $6.26 within four weeks. This is not a value opportunity emerging from a washout; it is the down-leg of a reverse-split-driven pump in a security with ~200% annualized idiosyncratic volatility and near-zero factor explanatory power (R² < 1%) — i.e., a casino chip, not a stock. The frame is neither “contrarian value” nor “quality compounder”; it is a falling knife attached to a going concern. Conviction: high (on the AVOID). What would flip me constructive: a real, audited recovery in collectible revenue plus a non-dilutive recapitalization that ends the death-spiral financing — or an actual, cash infringement award from Apple (not another validity ruling). What confirms the bear: the next financing prices off the lowest VWAP again, a third reverse split, or any going-concern/delisting escalation. Tag: “the patent’s worth a billion, the company can’t make payroll.”

📈 Stock Price Action — Five-Year Event Map

Factual price history, not a recommendation. Price moves are FACT; attributed drivers are INTERPRETATION. Reverse splits make multi-year comparisons treacherous — figures below are stated as quoted at the time, with split-adjusted reference points flagged.

Xiao-I IPO’d in March 2023 at $6.80 per ADS, was caught in the April-2023 meme-squeeze of small Chinese ADRs (a multi-hundred-dollar spike, ~$1,870 split-adjusted to today’s share base), and has since round-tripped through a near-total collapse punctuated by two reverse splits used to defend the $1.00 Nasdaq minimum bid. As of 2026-06-18 the ADS trades $6.26, after a May-2026 1-for-20 reverse split, a pump to $15.47 (May 22), and a −60% four-week decline. The stock has lost well over 99% of its inflation/split-adjusted value since the 2023 squeeze.

The five-year arc is a textbook distressed-microcap pattern: a hyped IPO, a squeeze, a multi-year collapse, and serial reverse splits to defend the listing while dilutive financing grinds the share base. The latest leg (events 6–8) is the one that matters for anyone looking at the quote today — a reverse-split-and-pump that has already given back most of its gains.

1. Executive Summary

Xiao-I Corporation is a Shanghai-based “cognitive intelligence” software company — conversational AI, a “Hua Zang” large language model, smart-customer-service hardware, and a portfolio of smart-city/intelligent-construction/metaverse offerings — listed on Nasdaq as an ADR through a Cayman Islands holding company that controls its PRC operating entity via a variable-interest-entity (VIE) contractual structure. On paper it is an AI company in the world’s second-largest AI market. In financial reality it is a sub-scale, money-losing enterprise in apparent terminal decline, kept alive by expensive, highly dilutive financing.

The FY2025 results (20-F filed May 15, 2026) are not a soft patch; they are a structural break. Revenue fell 82.5%, from $70.3M to $12.3M. The company recorded a $101.8M net loss, the bulk of which (~$59.5M) is a single non-cash write-off of accounts receivable, advances to suppliers, and a $13.5M interest-free loan to a 26%-owned affiliate — i.e., much of the asset base generated by the prior years’ “revenue” proved uncollectible. Management’s own framing — that it deliberately exited “low-margin cloud-platform reselling” and “customized development” — is consistent with a more troubling reading: that a large share of the historical top line was low-quality, possibly round-tripped, revenue. Two US securities class actions allege precisely this: that the 2023 IPO prospectus overstated the company’s AI capabilities and customer contracts.

The balance sheet is the verdict. Book equity is negative $104.9M; total assets are $15.9M against $120.8M of liabilities; cash is $2.4M against ~$47M of debt and $91.4M of negative working capital. The auditor — a small Malaysian firm appointed in 2026, the company’s second small offshore auditor in three years — and management both flag substantial doubt about the company’s ability to continue as a going concern, as they did the prior year. The accumulated deficit is $226.6M.

Capital allocation is value-destructive by construction. The only post-IPO financing the company has accessed is a series of death-spiral convertible notes — conversion priced at the lower of a fixed price or 85% of the lowest 10-day VWAP minus $0.05, with 8% original-issue discounts, 18% default rates, and “pre-delivery” ADS issued to facilitate lender dumping. To keep the resulting dilution from breaching Nasdaq’s $1.00 minimum bid, the company has executed two reverse splits (1-for-9 in 2024, 1-for-20 in 2026; ~1-for-180 cumulative). Meanwhile the founder issued himself 3.7M super-voting preferred shares (20 votes each) for no disclosed cash, securing majority voting control and “controlled company” governance exemptions, even as public ADS holders are diluted into oblivion. The company simultaneously owes the founder $7.6M.

There is no demonstrable moat. The financials disprove durability: a real competitive advantage does not permit an 82% revenue collapse and the simultaneous write-off of the receivables behind it. The Hua Zang LLM is a rounding error ($7.8M of model-as-a-service revenue) competing against Baidu, Alibaba, ByteDance, DeepSeek, and iFlytek — incumbents with vastly greater capital, compute, and distribution. The one genuinely interesting asset is the company’s long-running patent litigation against Apple over Siri-related “chatbot” patents: the patent’s validity has been repeatedly upheld by Chinese courts, and Xiao-I claims RMB 10 billion (~$1.4B) in damages — but there is no infringement ruling, no cash has ever been received, and IP licensing is not a current revenue line. It is a contingent lottery ticket, not a value driver.

This report renders no recommendation and sets no price target (see the explicitly fenced Claude’s Take above for a personal view). The institutional conclusion is narrower and factual: Xiao-I exhibits the complete checklist of a distressed, serially-dilutive, governance-impaired Chinese microcap — collapsing revenue, going-concern doubt, negative equity, toxic financing, reverse-split survival mechanics, founder control entrenchment, management and auditor turnover, and active securities-fraud litigation. The embedded expectation in any positive equity case is not “AI growth” but “survival plus a windfall.” Both are speculative.

2. Business Overview

Xiao-I describes itself as a pioneer of “cognitive intelligence” in China, founded in 2001 (the Nasdaq holding company was incorporated more recently in the Cayman Islands; the Shanghai operating entity, Shanghai Xiao-i Robot Technology, is the historical business). Its stated technology stack integrates natural-language processing, speech processing, computer vision, machine learning, affective computing, and data intelligence. From that base it markets a sprawling product list: conversational-AI platforms (iBot Pro), smart voice customer service, knowledge-management systems, robotic process automation, an in-house large language model branded “Hua Zang,” and vertical “Hua Zang series” ecosystems for finance, urban public service, enterprise, architecture, and healthcare. It has also dabbled in metaverse/virtual-human/AR-VR products, AI “smart glasses” (launched H2-2024), and AI customer-service hardware (“tAIkbox”).

Revenue model and segmentation. The 20-F groups revenue into two product lines — MaaS (Model-as-a-Service), built on the Hua Zang LLM (subscription/API/local deployment), and non-MaaS (assessment, design, development, deployment of non-model products) — and into five delivery categories. The category detail tells the story of the collapse:

By product line, MaaS revenue was $19.2M (FY23) / $51.3M (FY24) / $7.8M (FY25); non-MaaS was $39.97M / $18.99M / $4.48M.

Recurring vs. non-recurring. Only maintenance & support ($4.8M, 39% of FY25 revenue) is genuinely recurring — and it is the only category that grew year-over-year. Management characterizes the cloud platform as “subscription-based with high renewal rates,” but that line fell 89% in a single year, which is fatal to any stickiness claim. The remainder is project- and resale-based. The practical conclusion: Xiao-I’s revenue is overwhelmingly transactional, lumpy, and — as the FY2025 write-offs demonstrate — of questionable collectibility.

Customers and geography. The company claims deep penetration of Chinese financial institutions (“80% of the top-10 China banks, 60% of the top-10 insurers”) and government/smart-city clients. Customer concentration is high and rising: the top customer was 31.9% of FY2025 revenue (29.3%/22.4% in FY23/24) and the top five were 65.8% (69.7%/49.9% prior). A single supplier, Beijing Blanstar, was 21.5% of purchases (37.6%/24.4% prior). There is no geographic revenue segmentation; long-lived assets are “substantially located in the PRC.” Overseas ambitions (smart glasses in “North America/Europe”) are aspirational and immaterial.

Corporate/legal structure. Investors in AIXI own ADSs of a Cayman holding company that has no material operations of its own. The holding company owns Hong Kong subsidiaries, which own a PRC wholly-foreign-owned enterprise (WFOE), Zhizhen Technology (Shanghai), which exercises contractual control over the operating VIE, Shanghai Xiao-I, through a set of “VIE Agreements.” ADS holders therefore have no equity ownership of the operating business — only contractual, China-court-untested exposure to it. The VIE and its subsidiaries represented 96–99% of total assets across 2022–2024. Post the May-2026 reverse split, one ADS represents 60 ordinary shares.

Verdict. This is a sub-scale, project-heavy Chinese AI software business with a thin recurring core, dangerous customer concentration, an opaque VIE wrapper, and — as the next sections show — a revenue base that has both collapsed and proven substantially uncollectible. The “broad AI platform” narrative is wide; the durable, collectible, recurring economics are nearly absent.

3. Industry Dynamics

China’s conversational-AI and generative-AI market is genuinely large and fast-growing — but that is precisely why it is a bad place to be sub-scale. The market is dominated by a handful of extraordinarily well-capitalized incumbents: Baidu (Ernie), Alibaba (Qwen), ByteDance (Doubao), Tencent (Hunyuan), iFlytek (Spark), and the disruptive low-cost entrant DeepSeek. Each commands compute budgets, data, distribution, and engineering headcount that dwarf Xiao-I’s by orders of magnitude. Xiao-I’s own competition disclosure names “Apple Siri, Microsoft Cortana and Amazon Echo” plus “many other Chinese competitors,” and concedes competition is “intense.”

Through the lens of the capital-cycle framework (Marathon/Chancellor), Chinese AI is a textbook example of capital flooding into a hot theme: a wave of investment into foundation models and AI applications, driving prices, talent costs, and compute toward levels at which returns are competed away. In such a cycle, scale and cost position determine who survives. The supply-side reality is brutal for a company that, in FY2025, cut R&D spending (deferring model training) and operates with 58 total employees, 27 in R&D. A frontier or even competitive foundation model cannot be sustained on that base. DeepSeek’s emergence has further commoditized the value of mid-tier proprietary models — the exact category Hua Zang occupies — by demonstrating that capable models can be trained cheaply and released openly, collapsing the pricing umbrella for subscale proprietary LLMs.

Regulatory landscape. Chinese AI operators face a thickening regulatory stack: the Cyberspace Administration of China’s generative-AI measures (algorithm filing, content controls, security assessments), the Data Security Law and Personal Information Protection Law, cybersecurity review for data-handling, and — for offshore-listed entities — the CSRC’s overseas-listing filing regime and SAFE restrictions on moving cash out of China via dividends. Layered on top is the VIE-enforceability overhang (the structure has never been validated in a Chinese court and could in principle be deemed contrary to PRC policy) and US-side risks: the Holding Foreign Companies Accountable Act (HFCAA) audit-inspection regime and PFIC tax status for US holders. Xiao-I’s auditors being Singapore/Malaysia-based mitigates direct HFCAA delisting risk relative to a mainland-audited issuer, but the broader China-regulatory and VIE risks are fully present.

Profit pools and value-chain position. The economic profit in Chinese AI accrues to (a) the hyperscale platform owners who bundle models into existing distribution (cloud, search, social, devices) and (b) a few differentiated vertical specialists. Xiao-I sits in neither position: it is a small horizontal/vertical hybrid reselling cloud capacity and doing bespoke integration projects, with a thin maintenance annuity. Its place in the value chain is that of a price-taking systems integrator dressed as a model company.

Verdict: structurally hostile industry for this company. The market’s growth is real, but it is a growth that rewards scale and capital and punishes sub-scale independents. Xiao-I has neither the capital nor the scale to earn an economic return here; the capital cycle is working against it, not for it.

4. Competitive Position

The central question — does Xiao-I have a durable competitive advantage? — is answered most powerfully by its own financial statements. In a single year, revenue fell 82.5%, gross profit fell from $48.0M to $5.5M, and roughly $59.5M of the assets generated by prior revenue were written off as uncollectible. No business with a genuine moat — switching costs, network effects, intangible/brand advantage, scale economics, or a cost advantage — experiences that. A moat, by definition, is the mechanism that prevents exactly this kind of deterioration. Its absence is therefore demonstrated, not merely argued.

The company nonetheless asserts several advantage claims, which warrant pressure-testing:

• “First-mover / pioneer in cognitive intelligence” and a 2001 founding. Age is not a moat. The relevant question is whether early entry produced a durable cost or demand advantage that persists today; the financials say it did not. Whatever lead Xiao-I once had in Chinese NLP has been overrun by far larger players.
• “>500 patents granted/pending” and AI-standards participation. Patent counts rarely translate into pricing power in software, and Xiao-I’s own results show zero licensing revenue from its IP (the Apple matter aside, discussed in §7). Standards participation is a marketing credential, not an economic moat.
• The Hua Zang LLM. A proprietary model is only a moat if it is differentiated and defensible. Hua Zang generated ~$7.8M of MaaS revenue in FY2025 (down from $51.3M of MaaS the prior year), is trained by a 27-person R&D team, and competes against models with hundreds of times the resources. It is a feature, not a fortress.
• Customer relationships with banks/insurers/government. Deep relationships can create switching costs — but the 82% revenue collapse and rising concentration (top customer 31.9%) suggest the opposite: a shrinking set of relationships, not sticky ones. If switching costs were real, revenue would not have fallen this far this fast.

Against the Greenwald & Kahn taxonomy — supply (cost) advantage, demand (captivity) advantage, or economies of scale combined with customer captivity — Xiao-I qualifies for none. It has no cost advantage (it is sub-scale and resells third-party cloud); it has no demonstrable customer captivity (revenue and receivables both evaporated); and it has negative economies of scale relative to the hyperscalers it competes with. The market-share-stability test fails spectacularly: a company losing this much revenue is losing share, not holding it. The ROIC test fails: the company earns deeply negative returns on capital.

Verdict: no durable competitive advantage. This is a crowded market in which Xiao-I is a sub-scale, undifferentiated, and now financially distressed participant. The moat claims do not tie to any financial outcome that would deteriorate without them — because the deterioration has already happened.

5. Growth History and Forward Opportunities

History. Reported revenue grew from $32.5M (FY2021) to $48.2M (FY2022) to $59.2M (FY2023) to $70.3M (FY2024) — and then collapsed to $12.3M (FY2025). The 2021 year was the company’s last profitable year (net income +$3.7M); every year since has been a loss, widening to the catastrophic $101.8M loss in FY2025. Critically, the FY2025 write-offs cast a retrospective shadow over the quality of the 2022–2024 growth: if ~$59.5M of receivables and related-party balances generated by that revenue had to be written off, then a meaningful portion of the reported “growth” never converted to cash. The two securities class actions allege exactly this pattern of overstated commercial traction dating to the IPO.

Management’s explanation for the FY2025 collapse is a “strategic realignment”: deliberately exiting “third-party cloud-platform resale activities that generated revenue at gross margins substantially below those of our core product lines” (cloud platform −89%) and declining “customized development engagements that did not meet our revised margin thresholds” (tech-dev service −91%). Two readings coexist. The charitable one: the company pruned junk revenue to focus on profitable lines. The skeptical one — better supported by the simultaneous write-off of the receivables from those same lines — is that the revenue was low-quality and uncollectible, and the “strategic exit” is a post-hoc framing of a business that fell apart. Either way, the durable franchise that remains is tiny: $4.8M of recurring maintenance plus a few million of model and software revenue.

Forward opportunities. On paper, Xiao-I has the standard menu: monetize the Hua Zang LLM via MaaS, sell AI customer-service into Chinese enterprises and government, expand AI hardware (glasses, tAIkbox), and pursue overseas markets. In practice, every one of these requires capital the company does not have and scale it cannot reach. With $2.4M of cash, going-concern doubt, a 27-person R&D team, and financing available only through dilutive convertible notes, Xiao-I is not positioned to invest into growth — it is positioned to manage decline. The realignment is a contraction, not a platform for expansion: R&D and selling expenses were both cut in FY2025.

The single largest potential “growth” event is non-operational: a cash damages award in the Apple patent litigation (§7). That is a binary, exogenous, and so-far-cashless contingency — not organic growth.

Verdict: low-quality, now-negative growth. The historical growth was substantially uncollectible; the current trajectory is severe contraction; and forward opportunities are unfundable at the company’s current liquidity. This fails the quality-of-growth test on every axis.

6. Financial Quality

Xiao-I’s financials are among the weakest in the coverage universe. The income statement, balance sheet, and cash flow each independently signal distress.

Income statement. FY2025 revenue $12.3M (−82.5%); gross profit $5.5M (gross margin 44.4%, down from 68.3%); operating expenses $102.1M, including SG&A of $77.6M (up from $26.3M) — the SG&A explosion is the bad-debt/credit-loss provision ($45.8M of provision for credit losses alone, vs. $2.1M the prior year) plus severance from the headcount purge. Operating loss $96.6M; net loss $101.2M (−101.8M pretax). For scale: the net loss is more than eight times revenue. The accumulated deficit is $226.6M.

The write-off, in detail. The cash-flow statement adds back $59.47M of “allowance of accounts receivable, advance to suppliers and due from a related party,” plus $2.5M of asset impairment and $5.9M of stock-based compensation. Accounts receivable collapsed from $67.4M (FY24) to $6.3M (FY25); advances to suppliers from $3.2M to $0.04M; long-term investments from $2.5M to $0. Embedded in this is a $13.48M (RMB 98.4M) interest-free loan to Zhizhen Guorui, a 26%-owned related party, fully provisioned in FY2025. This is the analytical heart of the case: the company lent money interest-free to an affiliate, booked revenue it could not collect, and wrote both down in the same year that revenue fell 82%.

Balance sheet — effectively insolvent. Total assets $15.9M; total liabilities $120.8M; total equity −$104.9M (it has been negative every year since at least 2022, but the FY2025 figure is an order of magnitude worse than FY2024’s −$15.8M). Cash $2.3M ($2.4M including restricted). Short-term borrowings $41.6M; total debt ~$46.8M; net debt $44.4M. Current ratio 0.115; negative working capital $91.4M. Accounts payable rose to $39.2M and accruals to ~$29.9M — the company is financing itself in part by not paying suppliers and accruals. This is a balance sheet that, absent continuous external financing, cannot meet its obligations.

Cash flow. Operating cash flow was −$3.7M in FY2025 (−$15.1M FY24, −$15.8M FY23). The smaller FY2025 burn is misleading: it reflects ~$71.7M of non-cash charges (the write-offs plus SBC) and working-capital relief from stretching payables, not an operational improvement. Free cash flow has been negative every year. The business does not self-fund; it is sustained by rolling short-term bank debt and dilutive convertibles.

ROIC/ROE. Both are deeply negative and, with negative book equity, not meaningfully computable as ratios. The honest statement is that the company destroys capital: it has earned a positive operating result in only one of the last five years (FY2021).

Unit economics / quality flags. Gross margin held in the mid-40s only because the lowest-margin reselling was cut; on the remaining base, the company still loses many multiples of revenue at the operating line. Stock-based compensation ($5.9M) is large relative to revenue. The auditor change (to a small Malaysian firm appointed in 2026, the second small offshore auditor in three years) and previously-disclosed material weaknesses in internal control over financial reporting compound the reliability concerns. For a Chinese VIE with a related-party write-off and active securities-fraud litigation, the quality-of-earnings discount should be severe.

Verdict: economics do not improve with scale — they have never demonstrated positive returns at any scale, and the company is now effectively insolvent on a book basis and dependent on dilutive financing to survive.

7. Capital Allocation

Capital allocation is where a distressed company either preserves or destroys what remains of shareholder value. Xiao-I’s record is unambiguously the latter, and the structure of its financing is the single most important thing for any prospective ADS holder to understand.

Financing = death-spiral convertibles. Since the March-2023 IPO (5.7M ADS at $6.80, ~$38.8M gross), every capital raise has been a toxic convertible note. Six such deals were executed from 2024 to 2026, with consistent structural features: conversion at the lower of a fixed price or 85% of the lowest 10-day VWAP minus $0.05; ~8% original-issue discounts; coupons stepping to 18% on default; prepayment at 110%; MFN and anti-dilution ratchets; change-of-control redemption at a 25% premium; and “pre-delivery” ADS issued to the lender at par to facilitate immediate selling. The fixed conversion price visibly cascaded downward as the stock fell — $7.201 (Jan-2025) → $3.04428 (Jun-2025) — confirming the death-spiral dynamic: as the price drops, the conversion price drops, more shares are issued, and the price drops further. There has been no straight equity, no ATM, no conventional PIPE — only these structures, sold to unnamed “institutional investors,” frequently with no placement agent.

Dilution and reverse splits. Ordinary shares outstanding rose from 31.9M (Dec-2024) to 55.2M (Dec-2025), and exceeded 164M (non-affiliate float alone) by April 2026 — the relentless output of note conversions priced off depressed VWAPs. To keep the ADS above Nasdaq’s $1.00 minimum bid through this dilution, the company executed a 1-for-9 reverse ADS split (Aug 2024) and a 1-for-20 reverse ADS split (May 2026) — a cumulative ~1-for-180 since the IPO. The pattern — reverse split, issue more dilutive paper, reverse split again — is the defining feature of the capital structure.

Governance and alignment — severe red flags. In December 2023 the company issued 3,700,000 super-voting preferred shares (20 votes each) to ZunTian Holding Limited, a BVI vehicle wholly owned by founder Hui Yuan, for no disclosed cash consideration. This handed Yuan >50% of the voting power (initially >79%), triggered “controlled company” status, and let the company opt out of certain independent-governance requirements. Public ADS holders are thus diluted by toxic notes while the founder cemented control for free. The company also owes the founder $7.6M (now interest-bearing at 6.8%), and granted him an option on ~1.45M ADSs. There are no dividends, no buybacks (other than a token repurchase right on pre-delivery ADS), and no value-accretive M&A.

Management turnover. In January 2026 the company installed a new CEO, Mingqu Lin, whose background is sales at a liquor company and a pharmaceutical distributor plus real-estate engineering — no AI or public-company experience. The CFO is strikingly junior. Independent directors include a steel trader and a railway-bureau labor-union chairman. Whatever this signals about strategy, it does not signal a team equipped to turn around a frontier-AI competitor.

Verdict: capital allocation is value-destructive and conflicted. The financing mechanism systematically transfers value from public holders to note investors and entrenches control in the founder. This is the single strongest reason the institutional read is so negative.

8. Changes and Headwinds — Last Two Years

The trailing two years contain almost nothing that strengthens the thesis and a great deal that weakens it.

• Revenue collapse and $59.5M write-off (FY2025). The defining event: an 82.5% revenue decline and the write-off of receivables, supplier advances, and a related-party loan — discussed throughout.
• Going-concern doubt, two consecutive years. Both the FY2024 and FY2025 audits/management discussions flag substantial doubt; management explicitly concedes its remediation plan “cannot alleviate the substantial doubt.”
• Two reverse splits. 1-for-9 (Aug 2024) and 1-for-20 (May 2026), both to defend the $1.00 minimum bid against dilution.
• Chronic Nasdaq non-compliance. Minimum-bid deficiency (Jul 2024, cured Sep 2024); minimum-bid and minimum-market-value-of-publicly-held-shares deficiencies (Dec 2025); MVPHS regained Apr 2026; minimum bid regained May 2026. The Nasdaq rule eliminating second compliance periods after a reverse split (effective Jan 2025) makes any future breach acutely dangerous.
• Founder control entrenchment. Super-voting preferred issued Dec 2023; “controlled company” exemptions adopted.
• Management and board overhaul (late 2025–early 2026). New CEO (Jan 2026), junior CFO, multiple director resignations and appointments of administratively-oriented replacements.
• Auditor change (Apr 2026). From Assentsure PAC (Singapore) to CHI-LLTC (Malaysia) — the second small offshore auditor; previously-disclosed material weaknesses in internal controls.
• Securities class actions. NY State Supreme Court (Jun 2024) and SDNY (Oct 2024), both alleging the 2023 IPO disclosures overstated AI capabilities and customer contracts; motions to dismiss pending.
• Serial toxic financing. Six convertible-note deals 2024–2026 (see §7).
• Apple patent — favorable validity ruling (Apr 2026). The China Supreme People’s Court rejected Apple’s appeal to invalidate certain Xiao-I VIE patents. A genuine legal milestone — but no infringement ruling, no damages, no cash. This is the one item the bull case leans on; it is a contingent, cashless event.

Verdict: the last two years strongly weaken the thesis. The only arguable positive — the Apple validity ruling — is non-operational and uncashed, and is dwarfed by the collapse, the write-off, the going-concern doubt, the dilution, and the governance deterioration.

9. Risk Analysis (Risk Matrix)

Catastrophic-loss / total-loss assessment. The probability of a permanent, near-total impairment of the equity is materially elevated relative to a normal company. The combination of going-concern doubt, negative book equity, $2.4M of cash, a death-spiral financing structure, and chronic delisting risk means that a path to zero (or near-zero) for the common ADS is plausible, not remote. Even survival scenarios involve such severe dilution that current holders may retain only a small fraction of any recovery. This is the defining risk: the realistic downside is not “−30%,” it is “−100%.”

10. Valuation Discussion (Embedded Expectations)

Conventional multiple-based valuation is largely meaningless for Xiao-I because the denominators are broken: earnings and EBITDA are deeply negative, book equity is negative, and the revenue base just fell 82% and is of contested quality. We therefore frame valuation as embedded expectations and option value, not as a multiple to be re-rated.

What the multiples say (with heavy caveats). Enterprise value is small — on the order of $40M (ROIC’s FY2025 estimate; the precise figure is in flux because the share count rises continuously via conversions and the post-split ADS count is unstable). On the collapsed FY2025 revenue of $12.3M, EV/sales is ~3.3x; on the suspect FY2024 base of $70.3M it would be ~1x. P/E, P/B, EV/EBITDA are all negative and uninformative. Own-history valuation-percentile datasets return no data for AIXI (too small/illiquid to be covered), so there is no clean own-history percentile to anchor to. The honest reading is that the equity is not “cheap” or “expensive” in any normal sense — it is an option on survival plus a patent windfall, sitting on top of ~$47M of debt and negative book equity.

Embedded expectations. For the current ADS price to be justified on fundamentals (setting aside the patent), one must believe the company will (a) survive the going-concern threat, (b) refinance or extinguish ~$47M of debt without wiping out the equity, © arrest and reverse an 82% revenue decline in a market dominated by far larger competitors, and (d) do so with $2.4M of cash, a 58-person workforce, and a financing channel that only offers dilutive convertibles. That is a demanding conjunction of low-probability events. The market is, in effect, pricing a survival-and-recovery option plus a patent lottery ticket — not a going concern with predictable cash flows.

Scenario sketch (qualitative).

• Bear (high probability): continued dilution and/or delisting; revenue stagnates or declines further; the equity trends toward a small fraction of current value or zero. The death-spiral mechanics make this the path of least resistance.
• Base (moderate probability): the company limps along, surviving via repeated dilutive financings and reverse splits; the ADS oscillates violently on AI/patent headlines while the underlying per-ADS claim on the business shrinks with each conversion. Holders experience high volatility and steady value erosion even absent a terminal event.
• Bull (low probability): a genuine catalyst — a cash Apple infringement award, a non-dilutive recapitalization, or an unexpected revenue recovery — re-rates the equity sharply. The asymmetry here is real but the probability is low and the timing uncontrollable; the Apple matter has been litigated since 2012 with no cash to show.

The patent, valued honestly. The RMB 10B (~$1.4B) claim is a headline, not an asset. Chinese courts have upheld the patent’s validity repeatedly, but there is no infringement ruling and no damages award; even a favorable infringement verdict would face appeals, and any award would likely be a fraction of the claim. Assigning it a non-trivial expected value requires assumptions the evidence does not yet support. It is best treated as a free option embedded in the equity — not as a basis for a target.

Verdict: there is no defensible fundamental valuation that supports the equity as an investment; the price reflects option value on survival and a patent windfall, both speculative. Consistent with firm policy, this report sets no price target and makes no recommendation (see Claude’s Take for a fenced personal view).

11. Variant Perception

Consensus / market belief. There is no meaningful sell-side consensus on AIXI — it is an uncovered microcap. The de facto market belief, revealed by the trading, is that AIXI is a vehicle for AI-narrative and patent-headline speculation: a low-float, high-volatility ADS that spikes on Apple-patent news and AI momentum and is traded, not invested in. The recent reverse-split-and-pump to $15.47 and subsequent −60% collapse is the behavioral signature of this belief.

Strongest bull case. (1) The Apple patent validity has been repeatedly upheld, and a future cash infringement award — even a fraction of the RMB 10B claim — would dwarf the company’s market value, creating enormous asymmetry. (2) China’s AI market is huge and policy-supported; a sub-scale survivor could, in principle, find a defensible vertical niche. (3) The equity is so small and so beaten down that any positive catalyst produces violent upside (as May 2026 showed). (4) The “strategic realignment” could, charitably, leave a smaller but profitable core.

Strongest bear case. (1) The company is effectively insolvent (negative equity, $2.4M cash, going-concern doubt) and survives only via death-spiral financing that guarantees ongoing dilution. (2) The 82% revenue collapse plus a $59.5M write-off — including a related-party loan — points to historically low-quality (possibly overstated) revenue, the subject of two securities-fraud suits. (3) Governance is conflicted: the founder controls votes via free super-voting preferred while public holders are diluted. (4) The competitive position is hopeless against Baidu/Alibaba/ByteDance/DeepSeek/iFlytek with 27 R&D staff. (5) The patent is cashless after 14 years of litigation. (6) The realistic downside is permanent near-total loss.

The 3–5 assumptions that matter most.

1. Survival vs. insolvency/delisting — can the company keep financing itself without a terminal event?
2. Dilution trajectory — how much of any recovery accrues to current holders after death-spiral conversions and further reverse splits?
3. Revenue reality — was the historical revenue real and collectible, or substantially overstated (as the write-off and lawsuits suggest)?
4. Apple patent → cash — does validity ever convert into an actual, collected damages payment?
5. Governance — will the controlling founder act in public holders’ interest, given the structure?

Factor-positioning read. A quantitative factor model shows AIXI with near-zero factor explanatory power (R² < 1% across all models) and ~200% annualized idiosyncratic volatility — it is not driven by Momentum, Value, Quality, or any systematic factor; it is pure idiosyncratic noise. Its “related stocks” per the model are leveraged ETFs (e.g., a 1.75x MSTR ETF), underscoring that it trades like a casino instrument, not an equity with a coherent factor identity. The leaderboard returns null (insufficient stable history after the reverse splits). This corroborates the variant-perception conclusion: AIXI is priced by speculation and flows, not by fundamentals or factor exposures. Where consensus may be “offsides” is in the willingness of momentum/headline buyers to treat a going-concern shell as an AI growth story — but fading that is a short, and the short is uncarryable given borrow scarcity and squeeze risk.

Verdict: the variant perception is not “the market is too bearish on a cheap AI company.” It is that the market periodically prices a going-concern shell as a speculative call option, and the durable economic value underneath is minimal and shrinking.

12. Fact vs. Interpretation Table

13. Open Questions

1. Who were the counterparties to the $59.5M of written-off receivables and supplier advances? Were they related parties or undisclosed affiliates, and does the pattern indicate round-tripping of revenue? (The $13.5M Zhizhen Guorui loan is disclosed; the rest is not fully attributed.)
2. What is the current, fully-diluted ADS count? Conversions are continuous and the post-split count is unstable; the true per-ADS claim on the business is a moving target.
3. What are the exact maturities, balances, and default status of the outstanding convertible notes? A default could trigger 18% rates and balloon redemptions the company cannot fund.
4. Will the Apple infringement case produce a ruling — and if so, will any damages be collectible given enforcement realities and likely appeals?
5. What is the founder’s ultimate intent given his super-voting control, his $7.6M loan to the company, and the management overhaul — recapitalize, sell, wind down, or continue diluting?
6. How will the two securities class actions resolve, and what is the contingent liability?
7. Is the VIE structure stable under current PRC policy, and is cash trapped in China?
8. What does the FY2026 interim trajectory look like — has revenue stabilized at the post-realignment base, or continued to decline?

14. What Must Be True

Bull case — what must be true, and its falsification test. For the equity to be a winner from here, the company must (1) survive the going-concern threat without a terminal dilution/delisting event, (2) stabilize and begin to grow collectible revenue, (3) secure non-dilutive or self-generated capital, and/or (4) convert the Apple patent into actual cash. Falsification test: if the next financing again prices off the lowest VWAP (confirming continued death-spiral dependence), or a third reverse split is required, or revenue in the next reported period fails to stabilize above the ~$12M base, the bull case is broken. A favorable Apple validity ruling does not count — only a collected cash award does.

Bear case — what must be true, and its falsification test. The bear case holds that Xiao-I is a value-destroying going-concern shell that will dilute and/or delist current holders toward near-total loss. Falsification test: the bear case is broken if the company (a) executes a genuinely non-dilutive recapitalization that extinguishes the death-spiral structure and restores positive working capital, (b) reports two consecutive periods of stabilized, collectible revenue with a path to operating cash generation, or © receives an actual cash Apple damages payment large enough to recapitalize the business. Absent one of these, the bear thesis stands.

The asymmetry is stark: the bull case requires multiple low-probability events to align; the bear case requires only the continuation of the observed pattern.

This analysis (sections 1–15) carries no investment recommendation and no price target. The only position taken anywhere in this document is in the explicitly fenced “Claude’s Take” block at the top, which is the author’s own independent opinion and general information only — not investment advice.

APPENDIX A — Standard Diligence Questionnaire

Xiao-I Corporation (NASDAQ: AIXI) — as of 2026-06-21

Supplemental to the research memo. Fact / Interpretation / Assumption labels applied where material.

General

What thoughtful questions have other investors asked about this company? Given the absence of institutional coverage, the most thoughtful “questions” are embodied in the two securities class-action complaints (NY State Supreme Court, Jun 2024; SDNY, Oct 2024), which ask whether the 2023 IPO prospectus overstated Xiao-I’s AI capabilities and customer contracts (Fact). Beyond litigation, the questions a skeptical investor must ask are: (1) Was the historical $70M revenue real and collectible, given the $59.5M FY2025 write-off? (2) How dilutive is the convertible-note machine, and what is the true fully-diluted share count? (3) Will the Apple patent ever produce cash? (4) Can the company avoid insolvency and delisting? (5) Does the controlling founder act for public holders? These are the load-bearing questions; none has a comforting answer.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Internal or external drivers? Earnings are at a catastrophic low, driven overwhelmingly by internal factors — an 82.5% revenue collapse and a $59.5M write-off — not by a macro cycle (Fact/Interpretation). The “strategic realignment” was a company-specific contraction.

How stable are revenues? Highly unstable. Revenue swung +18% (FY24) then −82.5% (FY25). Only the $4.8M maintenance line is recurring; the rest is project/resale-based (Fact).

Outlook for products/services; how big is the market? China’s conversational/generative-AI market is large and growing, but dominated by hyperscale incumbents (Baidu, Alibaba, ByteDance, Tencent, iFlytek, DeepSeek). Xiao-I is sub-scale and losing share (Interpretation). The addressable market is large; Xiao-I’s collectible share of it is small and shrinking.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More — DeepSeek and open models are commoditizing mid-tier proprietary LLMs (Interpretation).

How profitable is the business (ROIC, ROE)? Deeply unprofitable; ROIC/ROE are negative and not meaningfully computable given negative book equity. The company has posted positive operating income in only one of the last five years (FY2021) (Fact).

How profitable is the industry; barriers to entry? The economic profit accrues to scaled platform owners; barriers favor incumbents with compute/data/distribution. For a sub-scale independent, barriers work against survival (Interpretation).

Can the business be easily understood? The product set is sprawling and buzzword-heavy, but the financial story is simple and stark: collapsing, uncollectible revenue funded by dilutive debt (Interpretation).

Undermined by foreign low-cost labor? Do brands matter? Switching costs? Not a labor-arbitrage business. Brand is weak relative to incumbents. Switching costs are claimed but disproven by the revenue collapse (Interpretation).

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The Apple patent (and the RMB 10B claim) is a contingent off-balance-sheet asset of speculative, uncashed value (Fact/Interpretation).

Off-balance-sheet liabilities? The VIE contractual structure, the securities-litigation contingencies, and the death-spiral conversion features (which create contingent dilution rather than recorded debt) are the key off-balance-sheet exposures (Fact).

How conservative is the accounting? Quality-of-earnings concerns are high: a $59.5M write-off including a related-party loan, a related-party receivable structure, two small offshore auditors in three years, and disclosed internal-control material weaknesses (Fact). Conservatism is questionable.

How CapEx-hungry is the business? Capex is minimal (~$0.04M FY25) — it is an asset-light software/integration model. The cash drain is operating losses and debt service, not capex (Fact).

Capital Allocation & Management

How much FCF; how is it used; philosophy? FCF is negative every year. There is no FCF to allocate; the company consumes externally-raised cash. The financing “philosophy” is serial dilutive convertibles plus reverse splits (Fact).

Significant acquisitions recently? None material (Fact).

Buying back shares? No — the opposite. Shares grew from 31.9M to 164M+ in ~15 months via conversions; reverse splits mask the dilution (Fact).

Issuing large amounts of new shares to insiders? The founder received 3.7M super-voting preferred (20 votes each) for no disclosed cash, plus an option on ~1.45M ADSs (Fact). Public holders bear the dilution.

Compensation policy of directors/management? Modest cash (CEO ~RMB 330,000/yr ≈ $46k) plus equity; the controlling structure, not pay, is the alignment issue (Fact).

Motivations of management? The founder controls votes via free super-voting preferred and is a $7.6M creditor of the company; the figurehead CEO change (to a non-AI operator) and board churn suggest control/optionality preservation rather than operational turnaround (Interpretation).

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? ADR (one ADS = 60 ordinary shares post the May-2026 1-for-20 split); foreign private issuer filing 20-F/6-K (Fact). PFIC status is a disclosed risk for US holders.

Dividend policy? None; none expected (Fact).

How profitable is the business? Unprofitable; net loss ~8x revenue in FY2025 (Fact).

Is net income diverging from cash from operations? Yes, massively — the $101.8M loss vs. −$3.7M operating cash reflects ~$71.7M of non-cash write-offs/SBC and working-capital relief from stretching payables (Fact). Neither figure reflects a viable business.

Risks & Downside

What factors would cause the stock to decline? Further dilution, a delisting event, a going-concern/default trigger, an unfavorable litigation outcome, or simple exhaustion of the speculative bid (Interpretation). It already fell −60% from its May-2026 peak.

Risk of catastrophic loss? Chance of total loss? Elevated and non-remote. Negative equity, $2.4M cash, going-concern doubt, death-spiral financing, and delisting risk make permanent near-total loss a plausible outcome, not a tail (Interpretation). This is the defining risk.

Recent News & Events

Has the business environment changed recently? Yes — catastrophically. FY2025 brought the revenue collapse, the write-off, going-concern doubt, two reverse splits, a management/auditor overhaul, and continued toxic financing (Fact). The one “positive” headline — the Apr-2026 Apple-patent validity ruling — is non-operational and uncashed.

Significant acquisitions / accounting-policy changes / new markets/facilities/management? No acquisitions; auditor changed (Apr 2026); new CEO and directors (Jan 2026); office consolidation and headcount cuts under the realignment (Fact). No notable new markets or facilities — the trajectory is contraction.

(Note: broad financial-media coverage of AIXI is essentially absent; this section is built from SEC EDGAR filings and the FY2025 20-F.)

APPENDIX B — Source Appendix

Xiao-I Corporation (NASDAQ: AIXI) — Research Sources, as of 2026-06-21

All figures reconciled to primary filings where possible. Third-party financial-data aggregators and a quantitative factor model are used for cross-check; the SEC filings are primary.

Primary filings (SEC EDGAR; CIK 0001935172)

1. Form 20-F, FY2025 (filed 2026-05-15) — ea0290336-20f_xiao.htm. Primary source for: revenue collapse and segmentation; $59.5M write-off and related-party (Zhizhen Guorui, 26%-owned) loan; going-concern disclosure; VIE structure; competition; Apple patent litigation (Note 18); management bios; super-voting preferred; risk factors; auditor (CHI-LLTC, Malaysia). https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001935172
2. Form 20-F/A, FY2025 (filed 2026-05-22) — ea0291729-20fa1_xiao.htm. Administrative restatement; does not amend audited financials.
3. Form 20-F, FY2024 (filed 2025-05-15) — ea0238811-20f_xiaoicorp.htm. Prior-year going-concern; FY2024 financials; 1-for-9 reverse split (Aug 2024).
4. Form 20-F, FY2023 & FY2022 (filed 2024-04-30, 2023-04-28) — historical financials, IPO context.
5. Form 424B4 IPO prospectus (Mar 2023) — 5,700,000 ADS @ $6.80; ~$38.8M gross.
6. Form 424B5 prospectuses (2024-06-17, 2024-10-31, 2025-01-07, 2025-06-18, 2026-05-26) — convertible-note terms (85% of lowest 10-day VWAP − $0.05; 8% OID; 18% default; pre-delivery ADS).
7. Form 6-K filings (2024–2026) — corporate events: reverse splits (2026-05-07 [1-for-20], FY24 20-F [1-for-9]); Nasdaq deficiency/regain notices (2025-12-23, 2026-06-03); auditor change (2026-05-05); management/board changes (2026-01-28, 2026-01-13); Apple-patent ruling (2026-04-01); convertible-note closings.
8. Form F-3 shelf (File 333-279306) — registration conveyor for note conversions.
9. Schedule 13D/13G filings — founder/affiliate ownership; ZunTian Holding (BVI) super-voting preferred.

The trailing ~60-month SEC corpus (4× 20-F, 2× 20-F/A, 34× 6-K, 5× 424B5, 424B4, F-1/F-3 shelves, S-8, F-6, NT 20-F) was mirrored locally for review.

Quantitative / aggregated sources

10. Third-party financial-data aggregator (accessed 2026-06-21) — multi-year income statement, balance sheet, cash flow, valuation multiples, enterprise value, company profile for AIXI (annual, FY2021–FY2025). Used for trend reconciliation; reconciled to the 20-F. Note: the FY2025 EV estimate (~$40M) is approximate given the unstable post-split/post-conversion share count.
11. Market-data provider price history (accessed 2026-06-21) — split/dividend-adjusted OHLCV. Price history resets 2026-04-06 (post reverse-split). Used for the price-action event map. Broad news and own-history valuation datasets return no data for AIXI (uncovered microcap).
12. Quantitative factor model (accessed 2026-06-21) — R² < 1% across all models (near-zero factor explanatory power); ~200% annualized idiosyncratic volatility; “related stocks” are leveraged ETFs; insufficient stable history for a risk-adjusted track record. Used for the factor-positioning read.

Litigation / IP sources

13. Apple v. Shanghai Xiao-I patent litigation — Patent ZL200410053749.9 (“A Chatbot System,” 2009); validity upheld by China Supreme People’s Court (2020), Beijing IP Court (Jun 2024), and SPC rejection of Apple’s appeal (Apr 2026); RMB 10B (~$1.4B) damages claim; no infringement ruling or cash to date. Source: FY2025 20-F Note 18; 6-K 2026-04-01.
14. US securities class actions — NY State Supreme Court (filed Jun 26, 2024; SAC Jun 30, 2025) and US District Court SDNY (filed Oct 15, 2024), alleging 1933/1934 Act violations re IPO disclosures. Source: FY2025 20-F Note 18.

Reconciliation

Notes on reconciliation and caveats

• All financial figures are in USD as reported in the 20-F (the operating business reports in RMB; the company presents USD).
• Reverse splits (1-for-9 Aug-2024; 1-for-20 May-2026; ~1-for-180 cumulative) make multi-year per-share and price comparisons unreliable; figures are flagged accordingly.
• Third-party financial-data aggregators and the factor model are third-party data; where they diverge from the 20-F, the filing governs.
• Management commentary (e.g., the “strategic realignment” framing of the revenue collapse) is treated as hypothesis, not evidence, and weighed against the simultaneous write-off and the securities litigation.