Loading today’s stories
Invest2 distinct publishers3 min readPublished
Enterprise inference sales grew about eightfold to $73.9 million, and the reported loss came in $114 million inside consensus while the adjusted loss more than doubled. Both are true.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
Gross profit for the half was $20.8 million against $3.7 million a year earlier, with margin at 17.9% from 12.1% [10]. Hold that beside the adjusted net loss and the proportion is unflattering: gross profit covers roughly 7% of the adjusted loss, or put the other way, the adjusted loss is about 14 times the margin the revenue actually threw off [8]. MiniMax says its objective is "increasingly capable intelligence with compelling unit economics, making frontier capabilities affordable at scale" [18]. On this half's arithmetic that is an ambition rather than a description.
The segment detail is where the paying-inference argument stands up. Open Platform and other enterprise AI services multiplied about eight times year on year [13], which implies total revenue of roughly $30.4 million in the comparable half of 2025 [5]. The enterprise line on its own is now about 2.4 times the size of the whole company twelve months ago [14], and the half's revenue was about 1.5 times all of full-year 2025, which MiniMax reported at $79 million [c11b][3]. Consumer-facing AI-native products roughly doubled [9]. Only one of those numbers changed the shape of the business.
The geography disclosure reads like a retreat and is not one. Revenue from outside the Chinese mainland fell to around 61% of the total from 72% [12]. On the implied base, that is about $71 million of offshore revenue against roughly $22 million a year earlier [6], while mainland revenue went from about $8.5 million to about $45.5 million, more than five times [7]. Offshore did not slow. Domestic caught up faster.
The two consensus comparisons pull in opposite directions. Revenue landed about 19% below the roughly $144.5 million FactSet-compiled consensus, while the reported net loss came in about $114 million inside the roughly $471.7 million analysts expected [4][1][2]. The reported loss narrowed by $44.2 million; the adjusted loss widened by $154.3 million. That $199 million divergence sits entirely in the items the company strips out, namely share-based payment expenses, fair-value loss on financial liabilities and listing expenses [11]. Meanwhile research and development expenses more than doubled and selling expenses fell 18% [6], which is a company buying capability and no longer buying users.
To reach the $363.77 million that Bloomberg-compiled estimates put on 2026, the second half has to produce about $247 million, roughly 2.1 times the first [c11a][4]. That is not out of reach for a segment growing eightfold, but it is the specific test Citi set when it said investors would want evidence that the H3 model translates into revenue contribution even if the model itself is well received [15]. Jefferies took the other side in August, keeping MiniMax among its best buy ideas on cost efficiency and the H3 release [16]. At an implied market value near $12 billion against the near-$60 billion peak of March [c13b][10], the stock is priced closer to Citi's question than to Jefferies' answer.
Ranked by verification strength, evidence, and original report placement.
MiniMax reported revenue of $116.6 million for the first half of 2026, nearly quadrupling year on year, in its second earnings report since its January Hong Kong listing.
MiniMax posted a 283% surge in first-half revenue to US$116.6 million for the six months ended June 30, 2026.
Net loss narrowed to $358.0 million from $402.2 million, the Shanghai-based company said on Wednesday.
Adjusted net loss, excluding share-based payment expenses, fair-value loss on financial liabilities and listing expenses, was $293.03 million in the six months ended June, widening from $138.74 million a year earlier.
Gross profit grew more than five-fold to US$20.8 million from US$3.7 million, and gross profit margin increased to 17.9% from 12.1% a year earlier.
MiniMax shares closed 1.1% higher at HK$303 ahead of the results, nearly 80% off the peak.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Company-reported interim figures, independently reported by two outlets
Every headline number — revenue, reported and adjusted loss, gross profit and margin, segment and geographic mix — comes from MiniMax's own first-half report, corroborated in the same direction and magnitude by SCMP and Dow Jones/Morningstar, with two external consensus benchmarks (FactSet, Bloomberg) as reference points. Strength is limited by the absence of a linked filing in the cluster and by reliance on a company-defined adjusted metric.
Paid enterprise inference revenue scaling from a small base
Adoption is evidenced in revenue dollars rather than logos: US$73.9 million of enterprise/Open Platform revenue, about eight times the prior year, now 63.4% of the top line, plus doubling of other AI-native products and a fresh H3 multimodal release. That is real, paid usage, but the base is small, no named customers, seat counts or token volumes are disclosed, and total revenue still fell about 19% short of the first-half consensus while tracking only ~32% of the full-year forecast.
Growth headlines outrun the unit economics
The dominant framing — revenue nearly quadrupled, loss narrowed versus consensus, reiterated as a best buy idea — is accurate but selective. The same report shows revenue missing the first-half consensus by roughly 19%, the adjusted loss more than doubling to $293.03 million, gross profit covering only about 7% of that adjusted loss, and roughly $199 million of the reported-versus-adjusted divergence sitting in excluded items. The gap is moderate rather than severe because both publishers do surface the adjusted-loss expansion and the off-pace revenue trajectory.
Issuer disclosure plus sell-side and pre-listing incentives
The figures originate with a newly listed issuer that defines its own adjusted metric by excluding share-based payments, fair-value losses and listing costs, and that is actively courting mainland capital through Stock Connect inclusion and a planned STAR listing — a context that rewards a growth-forward presentation. The analyst voices quoted (Jefferies' best-buy reiteration, Citi's revenue-contribution caveat) are covering brokers with their own positioning interests, though the Citi caution and both outlets' reporting of the adjusted-loss expansion partially offset the pull.
High confidence in the numbers, lower in the trajectory
The arithmetic core of this story is solid: two independent reports of the same interim disclosure agree, and the derived ratios follow directly from the disclosed figures. Confidence is held below the top band because the second-half trajectory depends on roughly $247 million of unreported revenue, the H3 contribution is unquantified, and no primary filing or customer-level evidence is present in the cluster.
invest
The chips never move: Washington's fix for the Southeast Asia compute loophole1 distinct publisher
product
Baidu's AI line grew 25 percent and still lost the arithmetic1 distinct publisher
invest
The $20,000 red team: US guardrails push Bitcoin's defenders onto Chinese models1 distinct publisher
build
1.5% of Hugging Face repos take 99.2% of downloads, and the ceiling is Chinese1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
morningstar.com
1 article · August 26, 2026
scmp.com
1 article · August 26, 2026