top of page

Deepexi Seeks Fresh Capital as a Second H-Share Placement Tests Its AI Expansion

Aug 13
11 min read

Deepexi Technology has conditionally agreed to issue 14.286 million new H shares, targeting gross proceeds of about HK$500.01 million. The deal creates a clear tension for shareholders. Deepexi is raising equity for the second time in less than three months, even though earlier fundraising proceeds remain unspent.

The Hong Kong-listed enterprise AI company says the new money addresses a different and more immediate need. It wants to accelerate its AI agent platform, acquire computing resources, hire technical staff, and prepare for strategic investments.

That explanation rests on a recognizable growth story. Deepexi reported that first-half revenue more than doubled, while sales from its main enterprise AI platform more than tripled. However, the company remained unprofitable, and the latest shares would dilute existing investors further.

This is therefore more than a routine capital raise. It is a test of whether Deepexi can convert fast enterprise AI growth into durable returns before repeated financing changes the risk-reward balance for shareholders.

Deepexi’s Second Placement Arrived Less Than Three Months After the First

Deepexi is returning to the equity market before it has deployed the proceeds from its previous placement.

The company signed the new placing agreement before Hong Kong trading hours on August 13, 2026. According to the placement filing, agents will seek at least six independent professional or institutional investors.

The proposed 14.286 million shares equal about 4.27% of Deepexi’s existing issued shares. They would represent approximately 4.095% of the enlarged share capital after issuance, assuming no other changes occur before completion.

The company expects approximately HK$487.51 million in net proceeds after commissions and estimated expenses. Completion remains conditional, including on Hong Kong Stock Exchange listing approval and required regulatory processes.

Several conditions must be satisfied or waived by August 17, unless Deepexi and the sole overall coordinator agree to an extension. The filing explicitly warns that the transaction might not proceed.

The timetable matters because Deepexi completed another placement on May 28. That earlier deal involved 7.942 million new H shares and produced approximately HK$395 million in net proceeds.

None of those earlier net proceeds had been used by June 30, according to the latest filing. Deepexi assigned that pool mainly to overseas expansion and the development of localized products, sales operations, and customer service capabilities.

The company also had unused proceeds from its October 2025 initial public offering. Deepexi said those funds remained allocated to longer-term research, domestic sales expansion, overseas operations, acquisitions, and working capital.

Management draws a firm distinction among the three pools. IPO proceeds support its long-term plan, the May placement supports international expansion, and the August placement targets near-term AI product deployment.

That accounting distinction is important, but shareholders ultimately see one capital structure. Each new issuance spreads their ownership across more shares, regardless of how management labels the proceeds.

The August transaction also contains a 90-day lock-up commitment. Deepexi agreed not to undertake another equity issuance during that period without the coordinator’s written consent, subject to the agreement’s exceptions.

That restriction offers investors only a short window of certainty. The more important question is whether Deepexi deploys the fresh capital fast enough to produce measurable commercial gains.

Why Deepexi Says the AI Agent Window Requires More Cash

The company argues that enterprise AI has moved from model experimentation toward operational agents, increasing the cost of product delivery.

Deepexi sells software designed to help organizations use their data and knowledge in AI applications. Its central product is DeepexiOS, an enterprise platform developed from the company’s earlier FastAGI offering.

An AI agent is software that can plan and execute multi-step tasks with limited human intervention. In an enterprise setting, those agents also require access controls, data governance, system integrations, and predictable behavior.

Deepexi says customer demand increasingly centers on agents capable of handling longer tasks across business systems. These deployments require more than access to a large language model.

The platform must connect with internal applications and governed data. It must also coordinate several specialized agents, preserve organizational permissions, and operate reliably during extended workflows.

Deepexi plans to devote approximately 70% of the new net proceeds to research, product development, cloud services, computing resources, and technical recruitment. That makes product execution the central justification for the placement.

One planned investment concerns what the company calls its Harness collaboration architecture. Deepexi says this layer coordinates multi-agent teams, repeated task cycles, and enterprise permission controls.

The company also plans further work on its proprietary enterprise models and ontology architecture. An ontology organizes business concepts and their relationships so software can interpret data within a specific operational context.

Deepexi wants to expand ontology datasets and reusable skills for manufacturing, retail, scientific research, and public utilities. It also plans to improve long-horizon task planning, semantic reasoning, and automated data governance.

Those goals carry direct infrastructure costs. Training, fine-tuning, and running models require computing capacity. Deploying agents for customers creates continuing cloud expenses, especially when tasks involve multiple systems or lengthy execution cycles.

The company also intends to upgrade DeepWorks, its enterprise agent platform, into a daily operational portal combining cloud and device-side capabilities. That ambition expands both the product surface and the delivery burden.

Approximately 20% of the proceeds is earmarked for potential investments or acquisitions. Deepexi says it will look for targets with complementary technology, industry solutions, customer access, channels, or delivery capabilities.

No binding agreement with a specific acquisition target existed when Deepexi issued the announcement. The acquisition allocation therefore represents strategic flexibility, not a disclosed transaction pipeline.

The remaining 10% is designated for working capital and general corporate purposes. Possible uses include staffing, rent, professional services, and other operating expenses.

Deepexi expects to use the new placement proceeds by the end of 2028. That timeline gives management room to allocate funds, but it also stretches the period before investors can judge the entire program.

The rationale is internally coherent. Enterprise agents need product development, computing capacity, implementation expertise, and access to customer systems.

The unanswered question is whether this funding reflects unusually strong demand or a business model that consumes capital faster than reported revenue can replenish it.

Revenue Growth Gives the Placement a Stronger Case

Deepexi is raising capital against genuine revenue momentum, not merely a product roadmap.

The company reported first-half 2026 revenue of RMB283.99 million, up 115% from RMB132.10 million one year earlier. Its interim results attribute most of that increase to DeepexiOS.

Revenue from the enterprise operating system platform reached RMB225.95 million. That represented 79.6% of total revenue and a 209.2% increase from the comparable period.

FastData, Deepexi’s more established data intelligence offering, generated RMB58.04 million. That business declined 1.7%, suggesting the AI platform supplied essentially all of the company’s incremental growth.

The shift is strategically significant. In 2025, the predecessor FastAGI business generated RMB254.47 million, or 61.3% of annual revenue.

By the first half of 2026, the upgraded platform had already approached that full-year figure. Deepexi says rising computing investment and platform development helped support wider adoption.

Gross profit also grew faster than revenue. It increased 120.5% to RMB160.33 million, while gross margin rose from 55% to 56.5%.

DeepexiOS produced a 57.2% gross margin, compared with 55.2% for the corresponding product line one year earlier. Management attributes the improvement partly to more efficient delivery.

These numbers strengthen the argument for investing during a period of demand. A company with accelerating product revenue can reasonably decide that delayed capacity would cost more than fresh capital.

The customer trajectory provides additional context. Deepexi reported 130 customers for 2025, up from 89 in 2024. Customers using FastAGI increased from 20 to 70 during that period.

The 2025 business also showed less concentration risk than many young enterprise software companies. Deepexi said no single customer accounted for more than 10% of annual revenue.

However, revenue recognition complicates the standard subscription-software comparison. Most first-half revenue came from goods and services transferred at a point in time, rather than services recognized steadily over time.

Deepexi recorded RMB281.77 million at a point in time and only RMB2.22 million over time. This suggests that contract delivery and customer acceptance can materially influence the timing of reported growth.

The distinction does not invalidate the revenue. It means investors should avoid treating the results like recurring cloud subscriptions without further evidence.

Enterprise projects can be uneven. Contract size, acceptance schedules, deployment complexity, and payment collection can all create volatility between reporting periods.

Deepexi must therefore demonstrate that its growth represents a repeatable operating pattern. Sustained customer additions and expansion within existing accounts would provide stronger evidence than one exceptional half-year.

The placement gives the company resources to pursue that proof. It also raises expectations because management is asking shareholders to fund faster execution while the strongest operating indicators are still relatively new.

The Real Contest Is Growth Capital Versus Dilution

Deepexi’s placement asks investors to accept immediate dilution in exchange for an uncertain acceleration of future growth.

The company’s position is straightforward. Demand for enterprise agents is changing quickly, and Deepexi needs to secure talent, cloud infrastructure, and computing resources within the current market window.

Waiting for internally generated cash could slow development. It could also leave the company behind larger rivals with established cloud platforms, broader sales channels, and substantially greater research budgets.

Equity financing avoids adding the same repayment burden as new borrowing. It can be a practical choice for a growing company whose near-term spending supports products that will take time to commercialize.

Yet repeated placements transfer part of the execution risk to existing shareholders. Deepexi completed one issuance in May and proposed another in August, shortly after its public listing in October 2025.

The two placements together involve more than 22 million new shares. The latest issuance alone would enlarge the post-placement share count by roughly 4.1%.

Dilution is not automatically destructive. If new capital produces proportionally greater earnings and cash generation, each remaining ownership unit can still become more valuable.

The problem is that Deepexi has not yet demonstrated that outcome. The company remains in a phase where fast revenue growth, continuing losses, and substantial investment all coexist.

Its first-half loss and total comprehensive loss was RMB31.92 million. That was an 89.6% reduction from the prior-year period, but it was not a profit.

Research and development expenses climbed 23.8% to RMB72.13 million. Selling and marketing expenses increased 33.2% to RMB65.68 million as the company expanded its sales workforce and related activity.

These investments might support future scale. They also show why revenue growth alone cannot settle the financing debate.

Deepexi’s cash position makes the second placement more debatable. It reported RMB947.09 million in cash and cash equivalents at June 30, alongside time deposits, pledged deposits, and restricted cash.

The group had RMB102.40 million in short-term interest-bearing borrowings. Net current assets stood at approximately RMB1.10 billion.

Those figures do not suggest an immediate liquidity emergency. Instead, the placement appears designed to preserve existing allocations while creating another dedicated pool for accelerated AI investment.

Management says the funds are not interchangeable because the IPO and May proceeds serve different plans and timelines. That is a strategic allocation argument, not a claim that cash is unavailable.

Investors must decide whether ring-fencing the earlier funds creates discipline or masks an increasingly capital-intensive expansion plan. Both interpretations fit the disclosed facts.

The strongest case for Deepexi is that product revenue and margin improved before the raise. The strongest skeptical case is that management wants additional equity despite holding substantial cash and unused fundraising proceeds.

That is the article’s primary contest. Deepexi promises faster product development and commercial deployment, while shareholders absorb dilution before those benefits are independently visible.

What the Growth Numbers Do Not Yet Prove

Deepexi’s latest results establish momentum, but they do not establish durable profitability, recurring demand, or efficient use of capital.

First, the company’s platform terminology makes longitudinal comparison harder. FastAGI was upgraded and renamed DeepexiOS, while management describes the newer product as an AI-level enterprise operating system.

The strategic evolution may be real, but investors need consistent reporting across periods. Customer retention, contract renewal, deployment size, and expansion revenue would make adoption easier to evaluate.

Second, point-in-time revenue recognition can create lumpy comparisons. A project accepted shortly before a reporting deadline can lift one period, while delayed acceptance can depress the next.

Readers should therefore compare several periods rather than annualizing the first half mechanically. Revenue quality depends on repeatability, collection, and customer expansion as much as headline growth.

Third, receivables deserve attention. Deepexi recorded RMB17.59 million in impairment losses on financial and contract assets, compared with a gain one year earlier.

The company said higher sales led to a larger expected-credit-loss provision. That explanation is plausible, but the relationship between rapid sales and cash collection remains important.

Fourth, commercial success is not guaranteed by higher research spending. Deepexi itself lists the ability to generate returns from research investment among its material business risks.

The company also identifies competition, customer retention, product performance, cybersecurity, data protection, intellectual property, and trade restrictions as relevant uncertainties.

These risks are particularly significant for enterprise agents. Such systems interact with sensitive company data and may take actions across operational software.

A failed consumer chatbot response can annoy a user. An enterprise agent applying an incorrect permission, changing a record, or executing an unreliable workflow can create financial and compliance consequences.

Deepexi says it is investing in organizational permissions, data governance, reliability, and complex task execution. Those are necessary capabilities, but the filing does not provide independent technical benchmarks.

The placement announcement also describes potential acquisition benefits without identifying a target. Acquisitions can add technology and customer access, but integration creates another execution burden.

Deepexi would need to evaluate valuation, regulatory compliance, technical fit, financial performance, and integration feasibility. Holding capital for a possible transaction does not guarantee that an attractive target will appear.

The company’s international plans create a related uncertainty. Proceeds from the May placement remain assigned to overseas expansion and localization, while the August funds support near-term product scaling.

Running both programs simultaneously will test management capacity. Product development, domestic delivery, foreign market entry, and possible acquisitions each demand different operating skills.

The company’s reported results support cautious optimism, not certainty. Revenue growth, a higher gross margin, and a sharply narrower loss are meaningful improvements.

Still, none proves that Deepexi can fund future development from operations. That proof requires sustained cash conversion and reduced dependence on external equity.

Three Signals Will Decide Whether the New Capital Works

The next test is not whether Deepexi completes the placement, but whether deployment produces measurable operating leverage.

The first signal is the transaction’s completion. Investors should look for confirmation that all conditions were satisfied and that the shares reached at least six independent placees.

Completion would validate near-term institutional demand for the issuance. A delay, termination, or material change would weaken the claim that Deepexi can access growth capital on acceptable terms.

The identities and concentration of investors also matter where disclosure permits analysis. Deepexi does not expect any participant to become a substantial shareholder immediately after completion.

The second signal is use of proceeds. Future filings should show whether Deepexi actually spends the new capital on computing, cloud resources, technical hiring, platform development, and selected investments.

Progress should be assessed against the company’s stated allocation. The most important category is the 70% planned for research and product capabilities.

Investors should also compare deployment across all three capital pools. Continued inactivity in the May funds, combined with slow use of the August proceeds, would undermine the urgency argument.

Rapid spending would not automatically be positive. The company must connect spending with product releases, customer deployments, improved reliability, or stronger financial performance.

The third signal is operating leverage in the next results. Operating leverage occurs when revenue grows faster than the expenses required to generate it.

Deepexi should sustain DeepexiOS growth while preserving gross margin and narrowing its loss. Cash collection should also keep pace with recognized revenue.

A falling impairment charge would strengthen the quality of the sales expansion. A rising charge could suggest that rapid growth carries greater collection risk.

Product mix will matter as well. DeepexiOS now drives most revenue, while FastData has become comparatively stable.

That concentration can improve strategic focus. It also makes the group more dependent on continued enterprise AI demand and successful execution of one central platform strategy.

Customer evidence would provide another useful check. Growth in the number of deployed organizations, repeat purchases, larger contracts, and expansion within existing customers would support the company’s adoption claims.

Independent performance evidence would help too. Deepexi discusses long-horizon planning, multi-agent collaboration, controlled access, and reliable execution, but these capabilities need validation in real operating environments.

The decisive question is simple: does the new funding help Deepexi reach a point where customers, rather than new shareholders, finance its next stage?

If revenue remains strong, margins hold, losses narrow, and cash collection improves, the placement will look like timely growth capital. Continued fundraising without those outcomes would shift the interpretation toward structural capital dependence.

Deepexi has presented a detailed use-of-proceeds plan and reported encouraging growth. Now it must connect the two with verifiable results.

Readers following enterprise AI should watch the HKEX disclosure record, the company’s next financial report, and evidence of repeatable customer adoption. Those signals will show whether this second placement accelerates a scalable business or merely extends an expensive buildout.

Get your work moving in remio

See what needs your attention and take the next step in Workspace.

remio currently supports Windows 10+ (x64) and Macs with Apple silicon.

Your AI Partner at Work
Get more done with remio

Plan. Create. Deliver.
All in one place.

bottom of page