SenseTime’s first profit still faces a cash-conversion test | Magica
SenseTime’s first profit is real, but cash conversion is still unproven
Editorial Team
••📖6 min read
SenseTime’s first IFRS profit since its 2021 listing came with stronger generative-AI revenue and margins, but also substantial investment and disposal gains, a continuing adjusted loss, negative operating cash flow and a larger receivables balance.
SenseTime reported RMB617.3 million in first-half IFRS profit, its first since listing, after a RMB1.49 billion loss a year earlier.
Generative AI supplied 79.9% of revenue and gross margin rose to 41.4%, signs of a better sales mix.
The reported profit also included investment revaluations and disposal gains; adjusted net loss and operating cash flow remained negative.
SenseTime Group, the Hong Kong-listed AI software company behind the SenseNova model family and the SenseCore compute platform, reported its first IFRS profit since its 2021 listing: RMB617.3 million for the six months to June 30, against a RMB1.49 billion loss a year earlier. The result is a meaningful financial milestone. It is not yet a clean demonstration that the company’s model, infrastructure and agent businesses are converting demand into cash.
Dr. Xu Li, SenseTime’s co-founder, executive chairman and chief executive, is also one of the company’s two weighted-voting-rights beneficiaries. That combination gives the technical leadership unusually direct influence over a strategy that still requires heavy spending on compute and ecosystem investments. The company has no controlling shareholder, its interim results say.
The earnings improvement is substantial — and qualified
Revenue rose 23.4% to RMB2.91 billion and gross profit rose 32.9% to RMB1.21 billion. Gross margin improved from 38.5% to 41.4%. Generative AI generated RMB2.33 billion, or 79.9% of revenue, while computer vision contributed RMB496.8 million, or 17.1%; the remainder came from X Businesses. That is a materially different mix from the prior-year half, when generative AI accounted for 77.0% of sales.
Five technology groups have accumulated an estimated $3.2 trillion in future purchase and lease commitments. Filings show the funding pressure, but broad spending categories, long payment schedules and a methodology change limit what the total says about AI debt and growth.
Editorial Team
Gross margin
41.4%
38.5%
IFRS profit/(loss)
RMB617.3m
(RMB1.49bn)
Adjusted net loss
(RMB385.9m)
(RMB1.18bn)
Operating cash flow
(RMB763.7m)
(RMB639.5m)
The key qualification is in the route from revenue to statutory profit. Net other gains were RMB2.13 billion, including RMB1.25 billion in net unrealised fair-value gains on financial assets after the effect of investment-fund liabilities and RMB521.0 million of gains on subsidiary disposals. The company says higher revenue and gross profit, lower operating costs and strategic-investment revaluations together produced the turnaround.
Its non-IFRS reconciliation removes fair-value movements and share-based compensation, leaving a RMB385.9 million adjusted net loss. That measure is management’s presentation rather than an IFRS result, and the company cautions that it is not a substitute for, or necessarily comparable with, statutory accounts. Still, the gap makes the right conclusion narrower: the underlying business improved sharply, but the first reported profit was not solely produced by current-period operations.
The August 16 preliminary profit alert foreshadowed exactly this mix, citing smaller principal-business losses and fair-value changes in its AI investment portfolio. It was based on unaudited management accounts and warned that the figures could change; the later interim results supplied the final reported amount.
The commercial case depends on task delivery, not token volume
SenseTime is pitching a three-part system: SenseNova models, a “Token Factory” that provides compute and inference services, and an agent-management layer intended to carry out tasks in customer workflows. The company introduced a recurring-revenue measure — revenue from existing contracts that has recurring characteristics over the contract term — of RMB1.14 billion, or 39.3% of sales, up from RMB510.1 million and 21.6% a year earlier. It is evidence of more repeat business, not an annualised-revenue figure or proof of collection speed.
The operating scale is also company-reported. SenseTime said in its results announcement that SenseCore processed more than 2.4 trillion tokens a day in July, about 22 times the prior-year level, and that compute capacity under operation had reached 48,000 petaflops. It said the platform serves four external foundation-model providers, while its agents served more than 1,000 enterprise customers in more than 20 industries. Those figures indicate deployment breadth, but they are not an independent benchmark of model quality, task success or customer economics.
This is not an uncontested advantage in China’s model market. MiniMax and Z.ai, also known as Zhipu AI, each reported triple-digit revenue growth for the same period but net losses of US$358 million and RMB2.07 billion, respectively, a contemporaneous report said. The comparisons are not like-for-like: the companies have different business mixes and the losses are reported in different currencies. They nevertheless show why SenseTime’s statutory result is notable without establishing that it has solved the wider sector’s monetisation problem.
SenseTime’s working-capital metrics across the first halves of 2024 through 2026, company-reported. Source: sensetime.com.
Compute expansion brings a financing and collection burden
The infrastructure story has two sides. Purchases of property, plant and equipment were RMB2.29 billion in the half, and the company says higher depreciation and operating costs at AI data centres were major reasons cost of sales rose. Net cash used in investing activities was RMB3.74 billion.
SenseTime had RMB9.88 billion of cash and cash equivalents and RMB3.12 billion of term deposits at June 30, but also RMB6.08 billion of borrowings, up from RMB5.72 billion at year-end. Its financing cash flow included RMB2.83 billion in net proceeds from new share placements during the half. The December 2025 and April 2026 placings were expected to yield about HK$6.38 billion in net proceeds combined, with substantial allocations to SenseCore, generative-AI research and commercialisation. The April placement still had HK$596.7 million unspent at June 30.
That balance sheet can fund a build-out; it also makes cash conversion more consequential. Net cash used in operating activities rose to RMB763.7 million from RMB639.5 million. The company attributed the increase in working-capital needs partly to higher trade receivables and the timing of customer collections.
The receivables data warrant care rather than a blanket conclusion about uncollectibility. Net trade receivables rose to RMB3.76 billion from RMB2.51 billion at year-end. In the gross ageing schedule, RMB3.26 billion of RMB7.04 billion in trade receivables had been recognised more than three years earlier — 46.3% of that gross balance — while provisions for impairment totalled RMB3.28 billion. The filing also records a RMB35.2 million net impairment gain, after a RMB142.7 million loss a year earlier, which the company attributes to collection efforts. The measures point in different directions: collections improved enough to reverse the impairment line, yet the net receivables balance and operating cash outflow both increased.
What will settle the cash-conversion question
The filing’s forward-looking statements are not results, and it warns that actual performance may differ materially. The next test is whether management can turn the improved mix into cash without another large valuation contribution.
The next evidence should therefore be more concrete than another model release or portfolio revaluation: whether recurring contracts and external compute demand translate into lower receivables, positive operating cash flow and margins that hold after data-centre costs. If they do, SenseTime will have a stronger case that its task-delivery strategy is self-funding. If statutory profit again depends materially on fair-value gains while cash remains tied up in receivables and infrastructure, the first profitable half will look more like a financial inflection point than a completed operating turnaround.
PwC finds growth in a subset of AI-exposed junior job postings demanding senior skills, alongside pressure on entry-level opportunities. Workplace research suggests AI can support learning, but leaves employers to demonstrate how recruits acquire independent expertise.
CoreWeave bonds fell after the election of candidates favoring data center restrictions. Texas governor-elect Gina Hinojosa wants to extend an existing approvals pause until lawmakers pass protections, adding delivery uncertainty for an industry whose expansion depends on power and financing.
Walmart plans new permissions and a gateway for outside shopping agents, while Amazon blocks unauthorized tools and sells ads through its own assistant. The stakes are product discovery, customer data and whether delegated purchases add sales or displace advertising opportunities.
Apple reportedly cut some engineers’ daily AI allowances from $300 to $200 in October and is weighing deeper restrictions. The limits tighten control over model use, but unused allowances make the savings uncertain.
Factory is acquiring coding-agent collaboration platform HumanLayer after raising $200 million in September. It plans to integrate shared tasks and planning tools, while HumanLayer promises open source, refunds and at least 30 days of continued service.
A minister identifies Google-commissioned research behind the action plan’s £400 billion AI claim. The government’s central estimate is £120 billion, but different methods and adoption assumptions prevent a straightforward forecast-cut comparison.
Research has separated language models’ detection of altered internal activity from their ability to identify its content. Mechanistic studies and skeptical controls leave the interpretation disputed—and one intervention that improved detection made other measures of self-knowledge worse.
The IEA maps 134 policies governing data-center integration, arguing that credible project queues, coordinated tariffs and flexible connections can protect consumers while accommodating AI growth. Many measures remain too new to judge.
Three watchdog groups seek to halt Project Meridian, alleging that its contractor-heavy membership and closed meetings violate federal advisory committee law. MITRE’s invitation for outside ideas complicates the dispute over public participation.
Meta is seeking up to €9 billion in its first euro bond sale, drawing record orders while offering wider spreads than Alphabet. Its advertising business and large liquid reserves cushion an AI investment push that has overtaken quarterly operating cash flow.
Pangram is expanding its text-detection business into images and video. Browser checks and developer access broaden its reach, while company benchmark scores leave real-world error rates and costs unresolved.
About two dozen restrictive data center measures appeared to pass in initial returns. Texas governor-elect Gina Hinojosa’s promised construction pause adds a state-level constraint, though the capacity ultimately delayed remains uncertain.
Riot Platforms says its 15-year Corsicana lease with Anthropic will generate about $28 billion in revenue. The tenant owns and operates the compute, while Riot develops power infrastructure and funds module reimbursements through debt or credits against future rent.
Sonos plans limited U.S. early access for its conversational assistant, with a reported November 6 start. Users can return to older voice controls—a choice that also separates local audio processing from the new assistant’s cloud processing.
Tesla, SpaceX and AI infrastructure shares fell after Democratic election wins, while broader stocks rallied. Texas governor-elect Gina Hinojosa’s proposed data center pause puts a major construction pipeline in focus, but Republicans retain control of the Texas Legislature and federal agencies.
Capital One Ventures leads Sentra’s $150 million Series C and gains a board seat as the company expands AI-agent data controls. Sentra reports roughly 100 major enterprise customers and revenue in the tens of millions, but exact revenue, valuation and comparative costs remain undisclosed.
Google agreed to support Anthropic’s lease obligations at TeraWulf’s Kentucky campus, strengthening its bid for project financing. The support starts after delivery milestones, while debt terms and a guaranteed construction price remain unsettled.
OpenAI plans to acquire Pierre Computer and integrate its Code Storage service into Codex. The purchase would extend an existing supplier relationship into code infrastructure, but its terms, customer economics and relationship to OpenAI’s own Pando system remain unresolved.
OpenAI says India leads ChatGPT’s markets by weekly active users. Earlier free offers, strong app engagement and planned enterprise deployments explain the stakes, but the company has not disclosed comparable paid-user or revenue figures.
